JP2005335034A - Motion evaluating method for robot device - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To correctly check or evaluate the kinetic performance of a robot device composed of static characteristics and dynamic characteristics. <P>SOLUTION: When the robot device stands on one leg, a trunk tilts toward a raised leg in a frontal plane. It is ideal that the tilt of the robot device is the same when either one of the legs is raised. Further, when the robot device stands on either the right leg or the left leg, the trunk tilts frontward in a sagittal plane. It is also ideal that the tilt of the robot device is the same when either one of the legs is raised. The bilateral symmetry of the robot device is evaluated whether a difference in a tilting quantity in each direction in rolling and pitching of the trunk generated when each of the right or left leg is raised, is less than a predetermined value. <P>COPYRIGHT: (C)2006,JPO&NCIPI

Description

  The present invention relates to a motion evaluation method for evaluating characteristics of an operation or motion of mechanical devices, and more particularly to a motion evaluation method for a robot device having a plurality of movable joints.

  More particularly, the present invention relates to a motion evaluation method for a robot apparatus that correctly inspects or evaluates the motion performance of a robot apparatus including a large number of joint axes, such as a legged mobile robot. The present invention relates to a motion evaluation method for a robot apparatus that correctly inspects or evaluates motion performance including characteristics and dynamic motion characteristics.

  A mechanical device that uses an electrical or magnetic action to perform a movement resembling human movement is called a “robot”. It is said that the word “robot” comes from the Slavic word “ROBOTA (slave machine)”. In Japan, robots started to spread from the end of the 1960s, but many of them are industrial robots such as manipulators and transfer robots for the purpose of automating and unmanned production operations in factories. Met.

  Currently, research on legged mobile robots, including biped upright walking, is being actively conducted. This is because the legged mobile robot can cope with uneven walking surfaces such as rough terrain and obstacles and discontinuous walking surfaces such as up and down stairs and ladders. It depends on the reason that it can be realized. On the other hand, since this type of robot is unstable in posture and difficult to control walking, walking control that maintains the stability of the posture is positioned as one of the most important technical issues. Stable “walking” as used herein is defined as “moving with legs without falling down”.

  In many cases, ZMP (Zero Moment Point) is used as a norm for determining the stability of walking for posture stability control of a legged mobile robot. The standard for discriminating the stability by ZMP is the principle of d'Alembert that gravity and inertia force from the walking system to the road surface, and these moments balance with the floor reaction force and the floor reaction force moment as a reaction from the road surface to the walking system. based on. As a result of mechanical reasoning, there is a point where the pitch axis and roll axis moments become zero inside the support polygon formed by the sole contact point and the road surface, that is, ZMP (for example, see Non-Patent Document 1). ).

  Target ZMP control has been successful on a real machine by planning the motion to be in dynamic balance at every moment. The biped walking pattern generation based on the ZMP norm has advantages such that a foot landing point can be set in advance and it is easy to consider the kinematic constraint conditions of the foot according to the road surface shape. In addition, using ZMP as a stability determination criterion means that a trajectory is treated as a target value in motion control instead of force, and thus technically feasible.

"Migir Vokobratovic" "LEGGED LOCATION ROBOTS" (Ichiro Kato's "Walking Robot and Artificial Feet" (Nikkan Kogyo Shimbun))

  In mechanical devices, it is necessary to test or inspect performance to determine whether the command value can be accurately executed.

  In general, a robot apparatus is provided with a rotary actuator for each joint axis, and realizes a desired apparatus motion based on these position controls. For this reason, the individual joint actuators operate according to the command values, and furthermore, the desired posture and motion pattern can be generated as a whole robot device by combining the operations of the joint actuators. You must check your performance. However, at present, there is no motion evaluation method in the robot apparatus.

  In a robot apparatus that aims to replace human work as one major purpose, it is extremely important to ensure the performance that can realize movement according to the command command. In particular, in the case of a legged mobile robot device, in order to more suitably realize posture stability control, as a premise, the static motion characteristics and dynamic motion characteristics of the robot device are accurately inspected or evaluated, It is necessary to perform maintenance and operation of the device based on the motion evaluation result.

  However, even in a robot apparatus composed of the same parts, there may be an unexpected difference in motion performance due to individual differences among parts such as actuators or variations during assembly. For example, even if a robot device configured with the same specifications executes the same walking command, one robot can walk straight, but another robot moves in the left or right direction or changes its standing posture. There is a possibility that it cannot be maintained and falls.

  In addition, since the legged mobile robot has a larger number of joints than other mechanical devices, there is a problem that it is difficult to evaluate and verify whether or not a desired motion can be realized. For example, even if the robot apparatus cannot realize the motion according to the command, it is difficult to identify the cause due to the complexity of the joint configuration. For example, when an accurate walking motion cannot be realized, which part or joint axis is defective, whether it is an error based on dynamic characteristics generated when the robot apparatus is operated, or is due to static characteristics in the first place It is not easy to find out.

  As the robot industry develops in the future, the number of robotic devices that will spread will increase. Therefore, it is desired to establish a motion evaluation method that can be easily and reliably performed not only by experts who are skilled in handling robot devices but also by beginner engineers who are not sufficiently skilled. .

  The present invention has been made in view of the technical problems as described above, and its main purpose is an excellent robot capable of correctly inspecting or evaluating the motion performance of a robot apparatus composed of a large number of joint axes. An object is to provide a motion evaluation method for a device.

  It is a further object of the present invention to provide an excellent motion evaluation method for a robot apparatus that can correctly inspect or evaluate the motion performance of the robot apparatus including static characteristics and dynamic characteristics.

  A further object of the present invention is to provide excellent robot apparatus motion evaluation that can be easily and reliably performed not only by experts who are skilled in handling robot apparatuses but also by beginner engineers who are not sufficiently skilled. It is to provide a method.

The present invention has been made in consideration of the above problems, and the first aspect of the present invention is a motion evaluation method for a robot apparatus including a plurality of movable parts.
The robot apparatus is configured substantially symmetrically, and controls the driving unit adaptively using the driving unit that drives the movable unit, the measuring unit that measures the state quantity of the moving unit, and the measured state quantity, or A drive control unit that performs non-adaptive control is provided,
Forming a symmetrical posture; and
Performing the non-adaptive control in the drive control unit when the symmetric posture is formed, and evaluating the drive unit;
When the symmetric posture is formed, adaptive and non-adaptive control is alternately executed in the drive control unit, and the measurement unit or the drive is based on the measurement result of the measurement unit at the time of each control. A step of evaluating the control unit;
It is the movement evaluation method of the robot apparatus characterized by comprising.

  For example, in a legged mobile robot device that performs bipedal walking, it is generally a structure that has equal limbs on the left and right sides and is basically symmetrical, and the key pose that is the basis for the transition of movement is. The robot device is designed and controlled to have symmetrical characteristics. Therefore, in the first aspect of the present invention, a condition for symmetry is added as an evaluation criterion for static motion characteristics in a key pose which is a transition of a robot apparatus, and an evaluation method thereof is provided.

  When the robot system's control system performs posture stability control based on mass properties, that is, mass distribution information, the basic posture is the same as the actual mass distribution and is symmetrical in the ideal sensor output. When the robot is taken, the robot apparatus should maintain the same posture when the adaptive control system is turned on and when it is turned off. Therefore, by turning off the adaptive control, it is possible to evaluate the static characteristics of the hardware of the robot apparatus, that is, the mechanical system and the electrical system. Further, the measurement system and the control system can be evaluated by comparing the measurement results when the adaptive control is turned on and when the adaptive control is turned off.

The second aspect of the present invention provides a motion evaluation method for a robot apparatus having a plurality of movable parts.
The robot apparatus is configured substantially symmetrically, and controls the driving unit adaptively using the driving unit that drives the movable unit, the measuring unit that measures the state quantity of the moving unit, and the measured state quantity, or A drive control unit that performs non-adaptive control is provided,
Forming an asymmetric posture; and
Performing a left-right asymmetry evaluation of the drive unit when the left-right asymmetric posture is formed;
It is the movement evaluation method of the robot apparatus characterized by comprising.

  In the second aspect of the present invention, by utilizing the fact that the robot apparatus is designed and controlled so as to have left-right symmetric characteristics, two poses including the left and right reversed are executed as a set of key poses. As a static motion characteristic, the left-right asymmetry is evaluated.

  Specifically, the mechanical hardware can be evaluated by setting the actuator position gain high while adaptive control is off. Further, by setting the actuator position gain low while adaptive control is off, the low-frequency torque of the actuator can be evaluated. Also, by turning on adaptive control, the measurement system and control system can be evaluated.

  Here, in the step of forming the left-right asymmetric posture, the left-right reversed postures are alternately executed as one set.

  In the step of evaluating the left-right asymmetry, when the left-right asymmetry posture is formed, non-adaptive control is executed in the drive control unit and a position gain in the drive unit is set high. The left-right asymmetry of the drive unit is evaluated. Alternatively, non-adaptive control is executed in the drive control unit, and the position gain in the drive unit is set low, and the left-right asymmetry of the low-frequency torque in the drive unit is evaluated. Alternatively, adaptive control is executed in the drive control unit, and the left-right asymmetry in the measurement unit or the drive control unit is evaluated based on the measurement results of the measurement unit at the time of each control.

  The robot apparatus is, for example, a legged mobile robot that performs bipedal walking, and includes at least a trunk and left and right legs connected to the trunk. In such a case, in the step of evaluating the left-right asymmetry, whether the difference in the amount of tilt in the roll direction of the trunk that occurs when the left leg is raised and the right leg is raised is within a predetermined value. Depending on how, the left-right symmetry of the robot apparatus with respect to the roll direction can be evaluated. Similarly, the left-right symmetry of the robot apparatus with respect to the pitch direction is evaluated based on whether or not the difference in the pitch direction inclination amount of the trunk that occurs when the left leg is raised and the right leg is raised is within a predetermined value. can do. Moreover, the posture which protrudes a waist part ahead is performed, and the left-right symmetry of the said robot apparatus regarding a yaw-axis direction can be evaluated in the attitude | position which protrudes a waist part ahead.

The third aspect of the present invention provides a motion evaluation method for a robot apparatus having a plurality of movable parts.
The robot apparatus includes a drive unit that drives a movable unit, a measurement unit that measures a state quantity of the movable part, and a drive control unit that controls the drive unit using the measured state quantity, and at least a part of the movable device is movable. The part touches the road surface,
Calculating a support polygon formed by a contact point between the movable part and the road surface;
Evaluating the motion characteristics of the robotic device based on at least one of the support polygon area or the rate of change of the area of the support polygon;
It is the movement evaluation method of the robot apparatus characterized by comprising.

  For example, the robot apparatus can perform posture stability control during walking and other exercises / behaviours using a ZMP stability determination criterion. In this case, the ZMP trajectory is planned so that the pitch axis and roll axis moments become zero inside the support polygon formed by the sole contact point of the robot device and the road surface, that is, the ZMP exists, or the leg or other Adaptively control the movements in the region.

  Therefore, in the third aspect of the present invention, the dynamic motion performance of the robot apparatus is evaluated based on the area of the support polygon and the speed at which the area changes.

According to a fourth aspect of the present invention, in the motion evaluation method for a robot apparatus including a plurality of movable parts,
The robot apparatus is controlled using a drive unit that drives a movable part, a measurement unit that measures a state quantity of the movable part, a measurement part that measures a state quantity of the movable part, and a state quantity measured by the drive unit. A drive control unit,
Causing the movable part to execute a motion pattern for evaluation;
Recording a measurement result in the measurement unit during execution of the motion pattern for evaluation;
Analyzing the recorded measurement results and extracting features of the robot device;
It is the movement evaluation method of the robot apparatus characterized by comprising.

  According to the fourth aspect of the present invention, dynamic motion characteristics can be evaluated for the robot apparatus after the static motion evaluation described above. As the dynamic motion characteristics of the robot apparatus, the accuracy of the origin of each actuator and motor, the output of each sensor, and the dynamic characteristics of the actuator and motor are mainly evaluated.

  In the fourth aspect of the present invention, the robot apparatus is caused to execute a walking pattern (or movement pattern) for dynamic motion evaluation or diagnosis on a known (ideal) road surface, and at the time of executing the movement. Log information such as sensor output is recorded. Then, by analyzing this log record, evaluation and diagnosis are performed by extracting characteristic points (or individual differences) unique to each robot apparatus. That is, a known motion is reproduced, and the dynamic characteristics can be quantitatively evaluated based on how much the actual operation matches the theoretical operation.

  In addition to basic walking motions, the present inventors perform “one-step system” motions and turning motions as dynamic motion evaluation or diagnosis motion patterns.

  The “one-step system” mentioned here refers to the operation of repeating one step with either the left or right leg (that is, the standing leg and the free leg are not switched and the supporting leg is fixed) or limited to a specific part. It refers to repeating one operation. In the case of an operation in which several steps are combined, errors in individual steps interfere with each other and inherit the influence of the error, so it is difficult to specify the cause. On the other hand, according to the one-step system, it is easy to specify the cause by accumulating (that is, emphasizing) errors due to the specific cause.

  Further, in the log record analysis, feature points are extracted from the walking log that is reproducing the walking pattern for diagnosis. For example, when the basic walking pattern is executed, it is diagnosed whether the entire feet of the left and right legs are evenly and simultaneously landing or leaving (that is, evaluation of horizontal landing and horizontal leaving).

  ADVANTAGE OF THE INVENTION According to this invention, the movement evaluation method of the outstanding robot apparatus which can test | inspect or evaluate correctly the movement performance which consists of a static movement characteristic and a dynamic movement characteristic which a robot apparatus has can be provided.

  Further, according to the present invention, not only an expert skilled in handling a robot apparatus but also an elementary engineer who is not sufficiently proficient can easily and surely perform the movement of an excellent robot apparatus. An evaluation method can be provided.

  Other objects, features, and advantages of the present invention will become apparent from more detailed description based on embodiments of the present invention described later and the accompanying drawings.

  Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.

A. Robot Configuration FIGS. 1 and 2 show a state in which the “humanoid” or “humanoid” robot device 100 used for carrying out the present invention is viewed from the front and rear, respectively. Yes. As shown in the figure, the robot apparatus 100 includes a trunk, a waist, a head, left and right upper limbs, and left and right lower limbs that perform legged movement, and is built in, for example, the trunk. The operation of the robot apparatus is comprehensively controlled by a control unit (not shown).

  Each of the left and right lower limbs is composed of a thigh, a knee joint, a shin part, an ankle, and a foot, and is connected to the lowermost part of the waist by a hip joint. The left and right upper limbs are composed of an upper arm, an elbow joint, and a forearm, and are connected to the left and right side edges above the trunk by shoulder joints. The head is connected to the substantially uppermost center of the trunk by a neck joint.

  The robot apparatus 100 configured as described above can realize bipedal walking by whole body cooperative operation control by a control unit (not shown in FIGS. 1 and 2). Such biped walking is generally performed by repeating a walking cycle divided into the following operation periods. That is,

(1) Single leg support period with left leg lifted right leg (2) Both leg support period with right leg grounded (3) Single leg support period with right leg lifted with left leg (4) Both legs with left leg grounded Support period

  The control unit includes a controller (main control unit) that processes drive control of each joint actuator constituting the robot device 100 and external input from each sensor (described later), and a housing in which a power supply circuit and other peripheral devices are mounted. Is the body. In addition, the control unit may include a communication interface and a communication device for remote operation.

  The walking control in the robot apparatus 100 is realized by planning a target trajectory of the lower limb in advance and correcting the planned trajectory in each of the above periods. For example, in the both-leg support period, the correction of the lower limb trajectory is stopped, and the waist height is corrected at a constant value using the total correction amount with respect to the planned trajectory. In the single leg support period, a corrected trajectory is generated so that the relative positional relationship between the corrected ankle and waist of the leg is returned to the planned trajectory.

  ZMP is used as a norm for determining the stability of walking for posture stability control of the robot apparatus including correction of walking motion trajectory. For this reason, interpolation calculation using a fifth-order polynomial is performed so that the position, speed, and acceleration for reducing the deviation from ZMP are continuous. The standard for discriminating the stability by ZMP is the principle of d'Alembert that gravity and inertia force from the walking system to the road surface, and these moments balance with the floor reaction force and the floor reaction force moment as a reaction from the road surface to the walking system. based on. As a result of the dynamic reasoning of the principle, there is a point where the pitch axis and roll axis moments become zero, that is, ZMP, inside the support polygon formed by the sole contact point and the road surface.

  FIG. 3 schematically shows the joint degree-of-freedom configuration of the robot apparatus 100. As shown in the figure, the robot apparatus 100 includes an upper limb that includes two arms and a head 1, a lower limb that includes two legs that realize a moving operation, and a trunk that connects the upper limb and the lower limb. And a structure having a plurality of limbs composed of the waist.

  The neck joint (Neck) that supports the head has four degrees of freedom: a neck joint yaw axis 1, first and second neck joint pitch axes 2 a and 2 b, and a neck joint roll axis 3.

  Each arm portion has a degree of freedom as a shoulder joint pitch axis 4 at the shoulder, a shoulder joint roll axis 5, an upper arm yaw axis 6, an elbow joint pitch axis 7 at the elbow, and a wrist ( Wrist) is composed of a wrist joint yaw axis 8 and a hand portion. The hand part is actually a multi-joint / multi-degree-of-freedom structure including a plurality of fingers.

  The trunk (Trunk) has two degrees of freedom: a trunk pitch axis 9 and a trunk roll axis 10.

  Further, each leg part constituting the lower limb includes a hip joint yaw axis 11 at the hip joint (Hip), a hip joint pitch axis 12, a hip joint roll axis 13, a knee joint pitch axis 14 at the knee (Knee), and an ankle (Ankle). ), An ankle joint pitch axis 15, an ankle joint roll axis 16, and a foot.

  However, the robot apparatus 100 does not have to be equipped with all the above-described degrees of freedom or is not limited to this. It goes without saying that the degree of freedom, that is, the number of joints, can be increased or decreased as appropriate in accordance with design / manufacturing constraints and required specifications.

  Each degree of freedom of the robot apparatus 100 as described above is actually mounted using a rotary actuator, and motion control is performed based on these rotational position controls. These joint actuators must be small and light because of the need to eliminate extra bulges in appearance and approximate human body shape, and to perform posture control on unstable structures such as biped walking. Is preferred.

  In this embodiment, a small AC servo actuator of a gear direct connection type and a servo control system of a single chip built in a motor unit is mounted. This type of AC servo actuator is disclosed in, for example, Japanese Patent Laid-Open No. 2000-299970 already assigned to the present applicant. Each joint actuator is provided with a torque sensor for detecting motor torque and an angle / position sensor for detecting a rotational position or a joint position.

  Further, in this embodiment, by adopting a reduced speed gear as the direct connection gear of the actuator / motor, the passive characteristic of the drive system required for the robot 100 of the type that places importance on physical interaction with humans is obtained. ing.

  FIG. 4 schematically shows a control system configuration of the robot apparatus 100. As shown in the figure, the robot apparatus 100 is adapted to each mechanism unit 30, 40, 41, 50R / L, 60R / L expressing human limbs, and adaptive control for realizing a cooperative operation between the mechanism units. (Where R and L are suffixes indicating right and left, respectively, and so on).

  The overall operation of the robot apparatus 100 is comprehensively controlled by the control unit 80. The control unit 80 exchanges data and commands between a main control unit 81 configured by main circuit components (not shown) such as a CPU (Central Processing Unit) and a memory, and each component of the power supply circuit and the robot 100. The peripheral circuit 82 includes an interface (none of which is shown).

  The peripheral circuit 82 referred to here is an interface for connecting external peripheral devices, charging stations (not shown), and other peripheral devices connected through cables and radio, in addition to peripheral devices mounted on the robot apparatus.・ Include connectors.

  In realizing the present invention, the installation location of the control unit 80 is not particularly limited. Although it is mounted on the trunk unit 40 in FIG. 4, it may be mounted on the head unit 30. Alternatively, a control unit 80 may be provided outside the robot apparatus 100 to communicate with the robot apparatus 100 main body by wire or wirelessly.

Each degree of freedom of joint in the robot apparatus 100 shown in FIG. 3 is realized by a corresponding actuator. That is, the head unit 30 includes a neck joint yaw axis actuator M ₁ , neck joint pitch axis actuator M ₂ , neck joint roll representing the neck joint yaw axis 1, neck joint pitch axis 2, and neck joint roll axis 3. axis actuator M ₃ is disposed.

The trunk unit 40 is provided with a trunk pitch axis actuator M ₉ and a trunk roll axis actuator M ₁₀ that represent the trunk pitch axis 9 and the trunk roll axis 10, respectively.

Further, the arm unit 50R / L is subdivided into an upper arm unit 51R / L, an elbow joint unit 52R / L, and a forearm unit 53R / L, but a shoulder joint pitch axis 4, a shoulder joint roll axis 5, an upper arm Shoulder joint pitch axis actuator M ₄ , shoulder joint roll axis actuator M ₅ , upper arm yaw axis actuator M ₆ , elbow joint pitch axis actuator M ₇ representing each of yaw axis 6, elbow joint pitch axis 7, and wrist joint yaw axis 8. , wrist joint yaw axis actuator M ₈ is arranged.

The leg unit 60R / L is subdivided into a thigh unit 61R / L, a knee unit 62R / L, and an ankle unit 63R / L, but the hip joint yaw axis 11, hip joint pitch axis 12, hip joint roll Hip joint yaw axis actuator M ₁₁ , hip joint pitch axis actuator M ₁₂ , hip joint roll axis actuator M ₁₃ , knee joint pitch axis representing the axis 13, knee joint pitch axis 14, ankle joint pitch axis 15, and ankle joint roll axis 16. An actuator M ₁₄ , an ankle joint pitch axis actuator M ₁₅ , and an ankle joint roll axis actuator M ₁₆ are provided.

The actuators M ₁ , M ₂ , M ₃ ... Used for each joint are more preferably a small AC servo actuator of the type directly mounted on a gear and mounted in a motor unit with a servo control system integrated into a single chip. ).

  For each mechanism unit such as the head unit 30, the trunk unit 40, the arm unit 50, and each leg unit 60, sub-control units 35, 45, 55, and 65 for actuator drive control are disposed.

  An attitude sensor 95 and an acceleration sensor 96 are disposed on the waist 41. The acceleration sensor 96 is arranged in each XYZ axial direction. Further, by providing an acceleration sensor 96 on the waist 41, the waist 41, which is a part with a large mass operation amount, is set as a control target point, and the posture and acceleration at that position are directly measured, and the posture based on ZMP Stable control can be performed. The attitude sensor 95 and the acceleration sensor 96 are configured as a gyro sensor G1 and an acceleration sensor A1, respectively, in FIG.

  In addition, ground check sensors 91 and 92 and acceleration sensors 93 and 94 are disposed on the legs 60R and 60L, respectively. The ground contact confirmation sensors 91 and 92 are configured by, for example, mounting a pressure sensor on the sole, and can detect whether the sole has landed based on the presence or absence of a floor reaction force. The acceleration sensors 93 and 94 are arranged at least in the X and Y axial directions. By disposing the acceleration sensors 93 and 94 on the left and right feet, the ZMP equation can be directly assembled with the feet closest to the ZMP position. In FIG. 3, sensors A2 and A3 for measuring acceleration at the foot and gyro sensors G2 and G3 for measuring the posture of the foot are arranged on the left and right ankles, respectively. In addition, force sensors F1 to F4 and F5 to F8 for measuring ground contact and floor reaction force are disposed at the four corners of the left and right soles.

  Here, when the acceleration sensor is arranged only on the waist 41 that is a part where the mass operation amount is large, only the waist 41 is set as the control target point, and the state of the foot is relative to the calculation result of the control target point. Therefore, the following condition must be satisfied between the foot and the road surface.

(1) The road surface does not move no matter what force or torque is applied.
(2) The friction coefficient with respect to translation on the road surface is sufficiently large, and no slip occurs.

  On the other hand, in the present embodiment, a reaction force sensor system (such as a floor reaction force sensor) that directly applies a force to the ZMP is provided on a foot that is a contact portion with the road surface, and local coordinates used for control and the coordinates thereof are used. An acceleration sensor for directly measuring is provided. As a result, the ZMP balance equation can be assembled directly at the foot closest to the ZMP position. Therefore, more rigorous posture stabilization control can be realized at high speed. As a result, even on a gravel where the road surface moves when force or torque is applied, on a carpet with long bristle feet, or on a residential tile that cannot easily secure a sufficient translational friction coefficient, it can easily slide. The stable walking (motion) of the device can be guaranteed.

  The main control unit 80 can dynamically correct the control target in response to the outputs of the sensors A1 to A3, G1 to G3, and F1 to F8. More specifically, adaptive control is performed on each of the sub-control units 35, 45, 55, and 65 to realize a whole body movement pattern in which the upper limbs, trunk, and lower limbs of the robot apparatus 100 are driven in cooperation. To do. In the present embodiment, static motion characteristics and dynamic motion characteristics of adaptive control in these control systems are evaluated, but details will be given later.

The main control unit 81 generates a whole body motion pattern of the robot apparatus 100 by setting a foot motion, a ZMP trajectory, a trunk motion, an upper limb motion, a waist height, and the like, and instructs an operation according to these settings. The command to be transferred is transferred to each sub-control unit 35, 45, 55, 65. Then, in each of the sub-control section 35, 45 ... interprets the command received from the main control unit 81 outputs a drive control signal to each actuator _{M 1, M 2, M 3} .... Here, “ZMP” refers to a point on the floor where the moment due to floor reaction force during walking is zero, and “ZMP trajectory” refers to, for example, ZMP during the walking motion period of the robot 100. It means the trajectory that moves.

  Legged mobile robots use ZMP as a norm for determining the stability of walking. The stability criterion for ZMP is that the system forms an appropriate ZMP space, and if there is ZMP inside the support polygon, the system does not generate rotational or translational motion, and solves the equations of motion related to rotation and translation. There is no need. On the other hand, when there is no ZMP inside the support polygon or when there is no support action point for the outside world, it is necessary to solve the equation of motion instead of the ZMP equation. For example, since a support polygon does not exist at the time of getting off, such as when jumping or jumping off a hill, the equation of motion may be solved instead of the ZMP equation.

The ZMP equation describes the balance relationship of each moment on the target ZMP. In the control system according to this embodiment has a mass properties relates to a robot device (i.e. distribution information of the mass), represents the robot in a number of mass points m _i. Then, when these mass points m _i and the control target point, in all expressions ZMP equilibrium equation for obtaining the orbit of each control point total is zero moment on the target ZMP generated at the control target point m _i is there.

  The ZMP balance equation described in the world coordinate system (O-XYZ) and the ZMP balance equation described in the local coordinate system (O-X′Y′Z ′) of the robot apparatus are as follows.

Each of the above equations represents the sum of moments around the ZMP (radius r _i −P _zmp ) generated by the acceleration component applied at each mass point (or control target point) m _i and the applied j-th external force moment. the sum of M _j, describes that the sum of the moments of ZMP around produced by an external force F _k (a point of action of k-th external force F _k and S _k) is balanced.

  This ZMP balance equation includes a floor reaction force moment (moment error component) T in the target ZMP. By suppressing this moment error to zero or within a predetermined allowable range, the posture stability of the robot apparatus is maintained. In other words, correcting the movement (such as the foot movement and the trajectory of each part of the upper body) so that the moment error is zero or less than the allowable value is the essence of posture stabilization control using ZMP as a stability criterion. It is.

  As described above, when the acceleration sensor is provided on the foot sole that is the contact portion with the road surface, the local coordinate system of the real robot with respect to the world coordinate system is set, and the foot sole as the origin is set. And the ZMP balance equation can be derived directly. Furthermore, by arranging an acceleration sensor at a control target point having a large mass manipulated variable including the waist 41, the moment amount around the ZMP for each control target point can be directly derived using the acceleration sensor output value. Can do.

  Here, since a real robot as a control target is a mobile device, it is difficult to obtain a position vector on the world coordinates at each control target point. As an alternative, the position vector of the control target point on the local coordinates is obtained by a relatively easy calculation method such as inverse kinematics calculation. Therefore, the actual posture stabilization process may be performed using the latter ZMP balance equation described in the local coordinate system (O-X′Y′Z ′) of the robot apparatus. (Since it is difficult to accurately measure the ZMP trajectory in the world coordinate system with the state detector mounted on the robot, an external measuring instrument fixed to the world coordinate system is generally required. On the other hand, the ZMP trajectory can be obtained accurately and directly in local coordinates by measuring the ZMP trajectory by placing the load sensors F1 to F8 on the ground contact portion. In addition, in the present embodiment, in the local coordinate system, it is possible to directly measure information that is dominant in high-speed motion by arranging an acceleration sensor near the local coordinate origin. ZMP balance equation is used.)

  Note that the robot apparatus 100 according to the present embodiment has a center of gravity set at the waist position, is an important control target point for posture stability control, and constitutes a “base” of the apparatus.

B. Evaluation of the motion performance of a robot device In a robot device that is intended to replace human work as one major purpose, it is extremely important to ensure the performance that can realize motion according to the command command. In particular, in the case of a legged mobile robot device, it is necessary to accurately inspect or evaluate the static characteristics and dynamic characteristics of the robot device, and to evaluate the motion, as a precondition for realizing a more stable posture control. It is necessary to perform maintenance and operation of the device based on the result.

  Even robot devices composed of the same parts may cause unexpected differences in motion performance due to individual differences among parts such as actuators and variations during assembly. Since the legged mobile robot has a larger number of joints than other mechanical devices, it is difficult to inspect whether a desired movement can be realized.

  For example, even if a robot device configured with the same specifications executes the same walking command, one robot can walk straight, but another robot moves in the left or right direction or changes its standing posture. There is a possibility that it cannot be maintained and falls. In such a case, it is difficult to identify the cause due to the complexity of the joint configuration.

  Therefore, the present inventors propose a method for correctly evaluating the motion performance by dividing the robot device into static characteristics and dynamic characteristics.

B-1. Evaluation of static motion characteristics of the robot apparatus As the static characteristics of the robot apparatus, the origin accuracy of each actuator and motor, the output of each sensor, and the static characteristics of the actuator and motor are mainly evaluated.

  The principle of evaluation related to the static motion performance of a robotic device is that the local coordinate origin is set at the site where the target floor reaction force action point exists, and the static or dynamic position or posture of the control target site with respect to that origin is changed. The measurement is performed in the key pose (static or dynamic) which is the basis of the above, and the evaluation is performed with the deviation from each target value.

  As can be seen from FIG. 1 and FIG. 2, the robot apparatus according to the present embodiment is a structure that is basically symmetrical with the left and right limbs. The key pose that is the basis of the transition described here utilizes the fact that the robot apparatus is designed and controlled so as to have left-right symmetric characteristics. The key pose conditions that are the basis of transition and the evaluation methods are listed below.

(1) Key pose that is symmetrical in the left and right (including the target action point of force) In this case, the adaptive control that maintains the key pose is turned on and off, and the difference between the two is measured for evaluation. In this case, by turning off the adaptive control, the static characteristics of the hardware of the robot apparatus, that is, the mechanical system and the electrical system can be evaluated. Further, the measurement system and the control system can be evaluated by comparing the measurement results when the adaptive control is turned on and when the adaptive control is turned off.

(2) Key poses that are asymmetric in the left and right (including the target action point of force) Here, two poses including the left and right reverses are executed as a set of key poses. In this case, the left-right asymmetry is evaluated by turning off the adaptive control and setting the position gain of the related actuator high, then setting the position gain of the actuator low, and turning on the adaptive control. The machine hardware can be evaluated by setting the actuator position gain high while adaptive control is off. Further, by setting the actuator position gain low while adaptive control is off, the low-frequency torque of the actuator can be evaluated. Also, by turning on adaptive control, the measurement system and control system can be evaluated.

  As described above, the evaluation of the static motion characteristics of the robot apparatus is composed of the evaluation of the asymmetry of the airframe whose left and right are the targets, and the evaluation of adaptive control. Each will be described below.

B-1-1. Asymmetry Evaluation The robot apparatus according to this embodiment is designed and controlled so as to have symmetrical characteristics in principle, and quantitatively evaluates it. Specifically, using the fact that both legs are designed and manufactured almost symmetrically, the left and right legs stand on one leg respectively, and the posture at that time is symmetrical in the roll direction and the same in the front-rear direction. Evaluate whether the posture is achieved using the acceleration sensor of the trunk.

  When standing on one leg, the following three patterns are used, and the symmetry is evaluated in each.

(1) Set the actuator position gain low.
(2) Set the actuator position gain high.
(3) The position gain of the actuator is set high, and the waist 41 is projected forward, thereby exerting the influence of the waist 41 yaw axis.

  In the asymmetry evaluation, two poses obtained by adding the left and right reverses to the robot apparatus are set as a set of key poses, and each is executed alternately to check the influence on the roll, pitch, and yaw axis directions.

  FIG. 5 illustrates the influence on the key pose formed by the robot apparatus and the roll direction when evaluating the left-right asymmetry as the static motion characteristic evaluation of the robot apparatus. As shown in the figure, when the biped robot device stands on one leg, the waist 41 tilts toward the raised leg (free leg) in the frontal plane. Ideally, this inclination is the same when the left and right legs are raised. Therefore, left-right symmetry depends on whether or not the difference between the tilt amount of the waist 41 in the roll direction or the tilt amount of the absolute acceleration vector with respect to the roll direction when the left leg is raised and the right leg is raised is within a predetermined value. Can be evaluated.

  FIG. 6 illustrates the effect on the key pose and pitch direction formed by the robot apparatus when evaluating the left-right asymmetry as the static motion characteristic evaluation of the robot apparatus. As shown in the figure, the waist 41 tilts forward in the sagittal plane when the biped robot device stands on one leg on either the left or right side. Ideally, this inclination is the same when the left and right legs are raised. Therefore, left-right symmetry depends on whether or not the difference between the tilt amount of the waist 41 and the tilt amount of the absolute acceleration vector with respect to the pitch direction when the left leg is raised and the right leg is raised is within a predetermined value. Can be evaluated.

  As described above, the left-right asymmetry as the static motion characteristic of the robot apparatus can be evaluated in the roll direction and the pitch direction. On the other hand, the yaw axis is perpendicular to the gravitational acceleration in a normal standing posture and is not affected by the gravitational acceleration, and therefore has little influence on the left-right asymmetry. Therefore, as shown in FIG. 7, for example, the waist 41 is inclined and the yaw axis is also subjected to the gravitational acceleration so that the asymmetry is also evaluated with respect to the yaw axis. As shown in FIG. 7, by tilting the waist 41, a left-right asymmetric component that is affected by gravitational acceleration due to an error about the yaw axis can be extracted (see FIG. 8).

B-1-2. In the control system of the adaptive control embodiment robot apparatus according to, it has a mass properties relates to a robot device (i.e. distribution information of the mass) (described above), and the mass points m _i control target points. For example, the trajectory of each control point where the sum of moments on the target ZMP generated at all the control target points m _i becomes zero can be obtained based on the ZMP balance equation, and the posture stability of the robot apparatus can be adapted. Can be controlled.

  Here, when the mass property managed in the control system and the actual mass distribution of the robot device match and the sensor output is ideal, the adaptive control system is The robotic device should maintain the same posture when turned on and when turned off. Therefore, by measuring the posture with adaptive control turned off, it is possible to evaluate the static motion characteristics of the hardware of the robot apparatus, that is, the mechanical system and the electrical system. Further, the measurement system and the control system can be evaluated by comparing the measurement results when the adaptive control is turned on and when the adaptive control is turned off.

  FIG. 9 illustrates a method for evaluating the static motion characteristics in the roll direction in the adaptive control system of the robot apparatus. In a state where the actual mass distribution of the robot apparatus matches the sensor output and the sensor output is ideal, when taking a basic posture in which left and right are symmetrical, when the adaptive control system is turned on and off, Ideally, the inclination of the trunk is the same within the frontal plane. Therefore, when the adaptive control system is turned on and when it is turned off, the inclination of the waist 41 is measured by, for example, the posture sensor 95 or the acceleration sensor 96, and the roll direction inclination amount of the waist 41 or the roll of the absolute acceleration vector is generated. The static motion characteristic in the roll direction of the adaptive control system can be evaluated depending on whether or not the difference in the amount of inclination with respect to the direction is within a predetermined value.

  FIG. 10 illustrates a method for evaluating the static motion characteristics in the pitch direction in the adaptive control system of the robot apparatus. In a state where the actual mass distribution of the robot apparatus matches the sensor output and the sensor output is ideal, when taking a basic posture in which left and right are symmetrical, when the adaptive control system is turned on and off, Ideally, the inclination of the waist 41 is the same in the sagittal plane. Therefore, when the adaptive control system is turned on and when it is turned off, the inclination of the waist 41 is measured by, for example, the posture sensor 95 or the acceleration sensor 96, and the pitch direction inclination amount of the waist 41 or the pitch of the absolute acceleration vector is generated. The static motion characteristic in the pitch direction of the adaptive control system can be evaluated depending on whether the difference in the amount of inclination with respect to the direction is within a predetermined value.

B-2. Evaluation of Dynamic Motion Characteristics of Robot Device In the present embodiment, the stability determination standard by ZMP is used for posture stability control during walking and other motions. That is, the trajectory of each control point at which the sum of moments on the target ZMP generated at all the control target points m _i becomes zero is obtained based on the ZMP balance equation, and the support ground formed by the foot contact point and the road surface is formed. The ZMP trajectory is planned so that the point where the pitch axis and roll axis moments become zero inside the square, that is, the ZMP exists, or the motion in the legs and other parts is adaptively controlled.

  Therefore, the principle of the evaluation of the dynamic motion performance of the robot apparatus is based on the area of the support polygon and the changing speed of this area. Specifically, the following items can be listed as evaluation items.

(1) Number of grounding polygons (islands) formed in the non-grounded state of each leg (2) Number of grounding polygons (islands) in which the area of the support polygon shows the maximum value (3) Area of the grounding part Number of transitions from a certain value A to a certain value B (4) The position of the center of gravity of the supporting polygon (5) The time of the grounding polygon (island) in the continuous non-grounding state of each leg (6) The shape of the supporting polygon

It should be noted that the ground contact surface and its area are obtained by inverse kinematics, for example, sensor outputs from the ground contact confirmation sensors 91 and 92 and the posture sensor 95 provided on the sole and joint angles of the joint actuators M ₁ , M ₂ ,. It can be calculated by calculating.

  In addition, as the dynamic motion characteristics of the robot apparatus, the accuracy of the origin of each actuator and motor, the output of each sensor, and the dynamic motion characteristics of the actuator and motor are mainly evaluated. However, as a premise for evaluating the dynamic motion characteristics, it is preferable that the robot apparatus has sufficient static motion characteristics evaluated in the previous section B-1.

  The processing procedure of the dynamic motion evaluation of the robot apparatus is schematically as follows. That is, the robot apparatus after the static motion evaluation is executed on the known (ideal) road surface, and the robot apparatus is caused to execute a dynamic motion evaluation or diagnosis walking pattern (or motion pattern). Record log information such as sensor output. Then, by analyzing this log record, evaluation and diagnosis are performed by extracting characteristic points (or individual differences) unique to each robot apparatus.

  Here, the dynamic motion evaluation or diagnosis walking pattern (or motion pattern), in addition to the basic walking motion, the present inventors propose to perform the “one-step system” motion and the turning motion. .

  The “one-step system” mentioned here refers to the operation of repeating one step with either the left or right leg (that is, the standing leg and the free leg are not switched and the supporting leg is fixed) or limited to a specific part. It refers to repeating one operation. In the case of an operation in which several steps are combined, errors in individual steps interfere with each other and inherit the influence of the error, so it is difficult to specify the cause. On the other hand, according to the one-step system, it is easy to specify the cause by accumulating (that is, emphasizing) errors due to the specific cause.

  Further, in the log record analysis, feature points are extracted from the walking log that is reproducing the walking pattern for diagnosis. For example, when the basic walking pattern is executed, it is diagnosed whether the entire feet of the left and right legs are evenly and simultaneously landing or leaving (that is, evaluation of horizontal landing and horizontal leaving). A dedicated log analysis tool can be built to display and output diagnostic results and save files. Details of the log analysis tool will be described later.

B-2-1. The dynamic motion characteristic evaluation of a one-step robot apparatus is basically performed by reproducing a known motion and quantitatively evaluating how much the actual operation matches the theoretical operation.

  In the robot apparatus according to the present embodiment, sensors F1 to F4 and F5 to F8 for measuring ground contact and floor reaction force are arranged at the four corners on the soles of the left and right legs. At the time of evaluating the dynamic motion characteristic, a motion for evaluation or diagnosis is theoretically reproduced in which the four corner points of the sole are theoretically lifted (leaving) and landing (landing) at the same time. Then, using the sensors F1 to F4 and the F5 to F8 at the four corners of the soles, the getting-off and landing timings on the actual machine are measured. Data obtained from this is analyzed, and a quantitative evaluation value is calculated for each item.

  FIG. 11 illustrates a basic operation for evaluating the motion characteristics of the robot apparatus. In the example shown in the figure, the robot apparatus performs an operation of raising the right leg and stepping forward one step from the basic posture that is evenly supported by the left and right legs.

  In the diagnostic pattern, the entire feet of the left and right legs are evenly and simultaneously getting off and landing, that is, horizontal leaving and horizontal landing are basically used. Therefore, when performing an ideal motion, the sensors disposed at the four corners of the right leg sole should be turned off at the same time when leaving the floor and turned on at the same time when landing. In other words, when leaving the floor and landing on the actual machine, simultaneous landing and simultaneous landing cannot be performed on the sensors at the four corners of the sole, and if there is a difference in detection timing between the sensors, this is the actual machine operation. Deviation from theoretical behavior. In the evaluation of dynamic motion characteristics, statistical processing is performed on the deviation between this kind of actual machine operation and theoretical operation, and the cause of the deviation is investigated.

  In the one-step operation, a step with one of the left and right legs (that is, the standing leg and the free leg are not switched and the supporting leg is fixed) is repeatedly performed. In the case of an operation (walking pattern) that combines several steps, errors in individual steps interfere with each other and inherit the influence of the error, making it difficult to identify the cause. On the other hand, according to the walking pattern consisting of a one-step system, it is easy to specify the cause by accumulating (that is, emphasizing) errors due to the specific cause.

  12 to 14 exemplify types of one-step system operations.

  In the pitch system, the left and right legs repeat the steps of stepping forward and returning, stepping once on the spot, stepping forward and stepping back after one step.

  In the roll system, the left and right legs repeat the operation of stepping outward → one step back. In the roll system 2, the roll system is executed at half speed.

  In the yaw system, the left and right legs repeat the operation of stepping outward and moving back one step while turning the toes outward. In the yaw system 2, the yaw system is executed at half the speed.

  In the special yaw system, in order to see the influence of the yaw axis, the yaw system operation is executed with the posture in which the waist 41 is inclined. In the normal standing posture, the elements around the yaw axis are not affected by gravity, and therefore, the influence on the contact timing of the sole is difficult to appear. Therefore, the influence of the yaw axis on the ground contact is evaluated by inclining the waist 41 so that the yaw axis is also affected by gravity (see FIG. 8).

  By executing these one-step motions and analyzing the ground contact information of the soles obtained by the detection outputs of the sensors F1 to F4 and the F5 to F8 at the four corners of the soles, Properties can be evaluated. The evaluation items obtained as dynamic motion characteristics by analysis of the ground contact information will be exemplified below.

  It should be noted that the minimum time for transition to the non-grounded state of the bottom of the foot is that the first point of the lifted foot leaves the floor, and all points have finished leaving the state where the remaining three points are grounded (surface grounded state). Is the time t (time indicated by the arrow in the figure) (see FIG. 15), and the minimum bottom surface non-ground state transition average time is the average of the total number of steps on one leg. Is.

  In addition, in the case where the bottom of the foot composed of sensors F1 to F4, F5 to F8 arranged in the four corners of the sole, F5 to F8, and the foot panel is formed as a flat surface with high accuracy, FIG. As shown, the sole diagonal ground contact where only two points on the diagonal line are grounded is not ideal. Therefore, the number of times that a grounding pattern as shown in FIG. 16 occurs during motion reproduction is counted as the number of ground diagonal contact, and the number of times is limited. For example, when the heights of the sensors F1 to F4 and the alignments F5 to F8 are uneven or uneven, the assembly of the joint shaft is poor, or the frame is distorted, the bottom diagonal contact is likely to occur.

  Based on each evaluation item of the above [Table 1], the dynamic motion characteristic of the robot apparatus can be determined. Examples of determination criteria are shown below. However, in this example, the determination is made only for the first three (pitch system, roll system, yaw system) of the six motions. The rest is used for determining the cause when a defect (NG) occurs and for obtaining data.

B-2-2. The dynamic motion characteristic evaluation of the turning motion robot apparatus is basically performed by reproducing a known motion and quantitatively evaluating how much the actual motion matches the theoretical motion.

  In the previous item B-2-1, the dynamic motion characteristics were evaluated by the operation of a one-step system in which the support legs are fixed. Here, the dynamic motion characteristics are evaluated by causing the robot apparatus to reproduce a known turning motion and quantitatively evaluating how much the movement amount in the actual machine matches the theoretical movement amount.

  For example, as shown in FIG. 17, the reference line is placed on the right side surface of the right sole, and the robot device is rotated N times (for example, 3 times) by repeatedly performing the rotation of the posture as shown in the table below. Ideally, the turning operation is finished at the same place as the start of turning, but after turning, the angle error and the movement amount are measured with reference to the point outside the heel of the right sole.

  Although the dynamic motion characteristics can be evaluated based on both items of the angle error and the movement amount, the evaluation / determination may be performed only on the angle error. The table below shows the angle error (turning error) and its determination contents.

C. The motion characterization procedure Figure 18, the robot apparatus, shows a schematic procedure for performing motion characterization in the form of a flowchart.

  First, an arbitrary motion and behavior data of a control target model corresponding to the arbitrary motion are created by the motion creation system (step S1).

  For motion creation, for example, a motion editing apparatus described in Japanese Patent Laid-Open No. 2003-266347 already assigned to the present applicant can be used. According to this motion editing device, whether or not the motion can be operated while stabilizing the posture on the actual machine after performing motion interpolation by performing motion interpolation between the poses according to the pose of the aircraft input from the operator. Can be verified based on the ZMP stability criterion, and the motion pattern capable of posture stabilization can be corrected. In addition, the behavior data of the control target model corresponding to the corrected motion pattern is created using a motion simulator or the like.

  Next, the created motion is input to the control system of the robot apparatus, and the motion is reproduced on the actual machine (step S2). At this time, during motion playback, the grounding information obtained from the grounding confirmation sensors F1 to F4 and F5 to F8 disposed at the four corners of the left and right soles, other sensor information, and the like are recorded as a walking log. .

  And in the diagnostic system which evaluates the motor performance of a robot apparatus, a walk log is acquired from a robot apparatus, and these are analyzed and motor performance and feature points, such as a behavior of a sole, are extracted. Then, these analysis results are compared with ideal behavior data of the control target model, and it is determined whether or not the deviation falls within a certain allowable range (step S3).

  As a result of the motion reproduction on the robot apparatus, when the motion data that matches the ideal behavior data of the control target model or within the allowable range can be obtained, the entire processing routine is terminated. On the other hand, if the motion performance of the robot apparatus does not fall within the allowable range, after the motion performance of the robot apparatus is readjusted (step S4), the process returns to step S2, and the motion performance related to the motion is repeatedly evaluated.

  FIG. 19 shows a processing procedure for performing motion characteristic evaluation in the form of a flowchart in a robot apparatus having a plurality of movable legs.

  First, an arbitrary motion related to the robot apparatus is created by the motion creation system (step S11). Regarding motion creation, for example, a motion editing apparatus described in Japanese Patent Laid-Open No. 2003-266347, which has already been assigned to the present applicant, can be used (same as above). The motion created in this step is hereinafter referred to as “motion M”.

  Next, the motion creation system creates a motion for operating one or more of the plurality of legs in the robot apparatus (step S12). Regarding motion creation, for example, a motion editing apparatus described in Japanese Patent Laid-Open No. 2003-266347, which has already been assigned to the present applicant, can be used (same as above). The motion created in this step is hereinafter referred to as “motion N”.

  Furthermore, a symmetric motion is created with respect to the created motion with reference to the control unit (step S13). The motion created in this step is hereinafter referred to as “motion N ′”.

  Next, in the robot apparatus, after the adaptive control is turned off, the motion M is reproduced. And the absolute acceleration of the control part at the time of motion reproduction | regeneration is measured (step S14). The absolute acceleration obtained in this step is hereinafter referred to as “acceleration A”.

  Next, in the robot apparatus, after the adaptive control is turned on, the motion M is reproduced. And the absolute acceleration of the control part at the time of motion reproduction | regeneration is measured (step S15). The absolute acceleration obtained in this step is hereinafter referred to as “acceleration A ′”.

  Next, the motion N is reproduced in the robot apparatus. And the absolute acceleration of the control part at the time of motion reproduction | regeneration is measured (step S16). The absolute acceleration obtained in this step is hereinafter referred to as “acceleration B”.

  Next, the motion N ′ is reproduced in the robot apparatus. And the absolute acceleration of the control part at the time of motion reproduction | regeneration is measured (step S17). The absolute acceleration obtained in this step is hereinafter referred to as “acceleration B ′”.

  Here, the asymmetry with reference to the control unit in the robot apparatus is verified (step S18). This asymmetry can be evaluated based on whether or not the difference between the acceleration B obtained when the motion N is reproduced and the acceleration B 'obtained when the motion N ′ is reproduced is equal to or less than an allowable value.

  If sufficient asymmetry of the robot apparatus is not obtained, after the movement performance of the robot apparatus is readjusted (step S20), the process returns to step S14 to repeatedly evaluate the movement performance related to the motion.

  On the other hand, if a sufficient asymmetry of the robot apparatus is obtained, the adaptive control is further verified (step S19). Regarding the adaptive control system, the difference between the acceleration A obtained by reproducing the motion M with the adaptive control turned off and the acceleration A ′ obtained by reproducing the motion M with the adaptive control turned on is less than the allowable value. It can be evaluated whether or not there is.

  If sufficient performance is not obtained with respect to the adaptive control system of the robot apparatus, the movement performance of the robot apparatus is readjusted (step S20), and then the process returns to step S14 to repeatedly evaluate the movement performance related to the motion.

  On the other hand, when sufficient performance is obtained for the adaptive control system of the robot apparatus, the entire processing routine is terminated.

  FIG. 20 shows, in the form of a flowchart, a processing procedure for evaluating a static motion characteristic of a robot apparatus that is basically designed and manufactured symmetrically, such as having symmetrical legs. .

  First, in the robot apparatus, an arbitrary key pose that is symmetrical is created. Furthermore, a motion that transitions from the basic posture to the key pose to the basic posture is created using a motion creation system or the like (step S21). The motion created in this step is hereinafter referred to as “motion M”. In addition, the left-right symmetry mentioned here includes the target action point of force.

  In the robot apparatus, an arbitrary key pose that is asymmetrical is created. Furthermore, a motion that transitions from basic posture → key pose → basic posture is created using a motion creation system or the like (step S22). The motion created in this step is hereinafter referred to as “motion N”. Note that the left-right asymmetry mentioned here includes the target action point of force.

  Furthermore, a motion N ′ that is symmetrical with respect to the motion N is created (step S23). For example, if the motion N is a motion for stepping on the right leg, the motion N ′ is a motion for stepping on the left leg.

  In this way, a motion group for static motion evaluation is generated. Using this motion evaluation motion, the robot apparatus is evaluated. Here, the local coordinate origin is set at the site where the target floor reaction force action point exists, and the static or dynamic position and posture of the control target site with respect to the origin are changed to the key pose (static or dynamic) that is the basis of the transition. And evaluate with the deviation from each target value.

  First, in the robot apparatus, when adaptive control is turned off (a high gain and a low gain that is 75% or less of the high gain) and turned on, the motion M is reproduced. Then, the local coordinate origin is set at the part where the target floor reaction force action point exists in the key pose, and the static or dynamic position and posture of the predetermined control target part with respect to the origin are measured (step S24). In this step, A1, A2, and A3 are handled as target values and measurement data.

  Next, in the robot apparatus, when the adaptive control is turned off (a high gain and a low gain that is 75% or less of the high gain) and turned on, the motion N is reproduced. Then, the local coordinate origin is set at the part where the target floor reaction force action point exists in the key pose, and the static or dynamic position and posture of the predetermined control target part relative to the origin are measured (step S25). In this step, B1, B2, and B3 are handled as target values and measurement data.

  Furthermore, in the robot apparatus, when the adaptive control is turned off (a high gain and a low gain that is 75% or less of the high gain) and turned on, the motion N ′ is reproduced. Then, the local coordinate origin is set at the part where the target floor reaction force action point exists in the key pose, and the static or dynamic position and posture of the predetermined control target part with respect to the origin are measured (step S26). In this step, it is assumed that B′1, B′2, and B′3 are handled as target values and measurement data.

  Based on the data thus obtained, the static motion characteristics of the robot apparatus are evaluated.

  First, the hardware is inspected based on whether the difference between A1, B1, B'1 and each target value is less than the allowable value (step S27).

  If sufficient hardware performance of the robot apparatus cannot be obtained, the movement performance of the robot apparatus is readjusted (step S30), and then the process returns to step S24 to repeatedly evaluate the static movement performance of the robot apparatus.

  On the other hand, if the robot apparatus has sufficient hardware performance, then a low torque inspection is performed (step S28). This test is based on the difference between A2, B2, B′2 and each target value, and whether each difference between A2 and A1, B2 and B1, and B′2 and B′1 is less than the allowable value. To do.

  If sufficient low torque performance of the robot apparatus is not obtained, the robot apparatus is readjusted (step S30), and then the process returns to step S24 to repeatedly evaluate the static movement performance of the robot apparatus.

  On the other hand, if the robot apparatus has sufficient low torque performance, then the measurement and control system are inspected (step S29). This inspection is based on the difference between A3, B3, B′3 and each target value, and whether each difference between A3 and A1, B3 and B1, and B′3 and B′1 is less than the allowable value. To do.

  If sufficient performance is not obtained with respect to the measurement system and control system of the robot apparatus, the robot apparatus is readjusted (step S30), and the process returns to step S24 to evaluate the static movement performance of the robot apparatus. Repeat.

  On the other hand, when sufficient performance is obtained with respect to the measurement system and control system of the robot apparatus, the entire processing routine is terminated.

  In the performance evaluation shown in FIG. 20, the pose that is symmetrical (including the target action point of force) is the key pose that is the basis of the transition, and the motion is measured by measuring the difference when adaptive control is turned on and off. The performance is evaluated by evaluating the hardware (mechanical system and electrical system) based on the measurement results when the adaptive control system is off, and the measurement system based on the measurement results when the adaptive control system is on and off. And for evaluating the control system.

  In addition, the adaptive control system is turned off (actuator position gain is high) using a pose that is asymmetrical (including the target action point of force) and a key pose that is a combination of two poses that are reversed to the left and right. Evaluate left-right asymmetry when the system is off (actuator position gain is low) and the adaptive control system is on. Machine hardware can be evaluated by the left-right asymmetry when the adaptive control system is turned off and the actuator position gain is set high. Further, the low-frequency torque of the actuator can be evaluated by the left-right asymmetry when the adaptive control system is turned off and the actuator position gain is set low. Further, the measurement system and the control system can be evaluated by the left-right asymmetry when the adaptive control system is turned on.

  FIG. 21 shows, in the form of a flowchart, a processing procedure for evaluating a static motion characteristic of a legged mobile robot device configured by connecting both symmetrical legs to the waist portion 41.

  First, in a state where the adaptive control system of the robot apparatus is turned off, a predetermined basic posture is taken, and the inclination amount ss1 of the waist 41 or the absolute acceleration vector is measured (step S31).

  Next, with the adaptive control system of the robot apparatus turned off, take a posture with emphasis on the hip joint yaw axis (see FIGS. 7 and 8) and measure the tilt amount ss1y of the waist 41 or the absolute acceleration vector. (Step S32).

  Next, in a state where the adaptive control system of the robot apparatus is turned on, a predetermined basic posture is taken, and the inclination amount ss2 of the waist 41 or the absolute acceleration vector is measured (step S33).

  Next, in a state where the adaptive control system of the robot apparatus is turned off, the position feedback gain is lowered, the right leg is stood, and the inclination 41 of the waist 41 or the absolute acceleration vector is measured (step S34).

  Next, in a state where the adaptive control system of the robot apparatus is turned off, the position feedback gain is lowered, the left leg is stood, and the tilt amount ss15 of the waist 41 or the absolute acceleration vector is measured (step S35).

  Next, in a state where the adaptive control system of the robot apparatus is turned off, the position feedback gain is increased, the right leg is stood, and the tilt amount ss16 of the waist 41 or the absolute acceleration vector is measured (step S36).

  Next, in a state where the adaptive control system of the robot apparatus is turned off, the position feedback gain is increased, the left leg is standing, and the inclination 41 of the waist 41 or the absolute acceleration vector is measured (step S37).

  Next, with the adaptive control system of the robot apparatus turned off, the position feedback gain is increased, the right leg is stood in a posture (same as above) with the yaw axis of the waist 41 emphasized, and the tilt or absolute acceleration of the waist 41 is The vector inclination amount ss16y is measured (step S38).

  Next, with the adaptive control system of the robot apparatus turned off, the position feedback gain is increased, the left leg is placed in a posture (same as above) with the yaw axis of the waist 41 emphasized, and the tilt or absolute acceleration of the waist 41 is The vector inclination amount ss17y is measured (step S39).

  Next, the robot stands on the right leg while the adaptive control system of the robot apparatus is turned on, and measures the tilt amount ss3 of the waist 41 or the absolute acceleration vector (step S40).

  Next, the robot stands on the left leg with the adaptive control system of the robot apparatus turned on, and measures the inclination of the waist 41 or the absolute acceleration vector inclination amount ss4 (step S41).

  Then, based on these measurement results, an evaluation on the asymmetry of the robot apparatus and the adaptive control system is performed.

  First, the left-right asymmetry in the low-frequency torque of the actuator is evaluated based on whether the difference between ss14 and ss15 is equal to or less than a predetermined allowable value (step S42). If sufficient left-right symmetry of the robot apparatus is not obtained, the robot apparatus is readjusted for the kinematic performance (S47), and the process returns to step S31 to repeatedly evaluate the static kinematic performance of the robot apparatus.

  On the other hand, when sufficient left-right symmetry of the robot apparatus is obtained in the low-range torque of the actuator, subsequently, in the machine hardware, whether or not the difference between ss16 and ss17 is less than a predetermined allowable value. The left-right asymmetry is evaluated (step S43). If sufficient left-right symmetry of the robot apparatus is not obtained, the robot apparatus is readjusted for the kinematic performance (S47), and the process returns to step S31 to repeatedly evaluate the static kinematic performance of the robot apparatus.

  On the other hand, when sufficient left-right symmetry is obtained in the mechanical hardware of the robot apparatus, the influence of the waist 41 yaw axis is subsequently determined based on whether the difference between ss16y and ss17y is less than a predetermined allowable value. The left-right asymmetry in the machine hardware is taken into consideration (step S44). If sufficient left-right symmetry of the robot apparatus is not obtained, the robot apparatus is readjusted for the kinematic performance (S47), and the process returns to step S31 to repeatedly evaluate the static kinematic performance of the robot apparatus.

  On the other hand, if sufficient left-right symmetry is obtained in the mechanical hardware of the robot apparatus in consideration of the influence of the waist 41 yaw axis, subsequently, is the difference between ss3 and ss4 less than a predetermined allowable value? Thus, the left-right asymmetry in the measuring instrument and control system of the robot apparatus is evaluated (step S45). If sufficient left-right symmetry is not obtained in the measurement system and control system of the robot apparatus, the robot apparatus's kinematic performance is readjusted (S47), and the process returns to step S31 to determine the static kinematic performance of the robot apparatus. Repeat the evaluation.

  On the other hand, when sufficient left-right symmetry is obtained in the measuring instrument and control system of the robot apparatus, the robot apparatus is adapted according to whether the difference between ss1 and ss2 is less than a predetermined allowable value. The control system is evaluated (step S46).

  If sufficient performance is not obtained for the adaptive control system of the robot apparatus, the movement performance of the robot apparatus is readjusted (S47), and then the process returns to step S31 to repeatedly evaluate the static movement performance of the robot apparatus. . On the other hand, if sufficient performance can be confirmed for the adaptive control system, the entire processing routine is terminated.

  FIG. 22 shows a schematic processing procedure in the form of a flowchart for evaluating dynamic motion characteristics in a robot apparatus having a contact point with the road surface and using ZMP as a stability criterion.

  First, an arbitrary target motion pattern having a target support polygon is generated by the motion generation system (step S51). Regarding motion creation, for example, a motion editing apparatus described in Japanese Patent Laid-Open No. 2003-266347, which has already been assigned to the present applicant, can be used (same as above).

Next, the created motion is input to the control system of the robot apparatus, and the motion is reproduced on the actual machine (step S52). At this time, the behavior of the support polygon of the motion pattern at the time of motion playback is determined based on the sensor outputs from the ground check sensors F1 to F4 and the F5 to F8 installed at the four corners of the sole, and the respective joint actuators. The joint angles of M ₁ , M ₂ ,... Are calculated by inverse kinematics calculation.

  The behavior of the support polygon here refers to the area of the support polygon, the speed at which the area changes, and the like. Specifically, the following items can be listed as evaluation items.

(1) The number of grounding surfaces (islands) in the non-grounded state of each leg (2) The number of islands where the area of the support polygon shows the maximum value (3) The area of the grounding part from a certain value A to a certain value B Number of transitions (4) Position of the center of gravity of the support polygon (5) Time of islands in the continuous non-contact state of each leg (6) Shape of the support polygon

  Then, in the diagnostic system that evaluates the motion performance of the robot device, the behavior of the obtained support polygon is compared with the target value, and the dynamic motion performance is determined based on whether or not the deviation falls within a certain allowable range. Is evaluated (step S53).

  When the behavior of the support polygon obtained by the motion execution matches the target value or is within the allowable range, the entire processing routine is terminated. On the other hand, when the desired dynamic motion performance is not obtained, the motion performance of the robot apparatus is readjusted (step S54), and the process returns to step S52 to repeatedly evaluate the motion performance related to the motion.

  FIG. 23 shows, in the form of a flowchart, a processing procedure for evaluating dynamic motion characteristics in a robot apparatus having a waist portion 41 and a plurality of movable legs and using ZMP as a stability criterion.

  First, the robot apparatus is placed on a predetermined walking platform (not shown), and adaptive control is turned on (step S61).

  Next, the stepping motion 1 in the pitch direction is reproduced, and the sole behavior is recorded as Pitch 1 (step S62).

  Next, the stepping motion 1 in the roll direction is reproduced and the sole behavior is recorded as Roll1 (step S63).

  Next, the stepping motion 1 in the yaw direction is reproduced, and the sole behavior is recorded as Yaw1 (step S64).

  Next, the stepping motion 2 in the roll direction is reproduced and the sole behavior is recorded as Roll2 (step S65).

  Next, the stepping motion 2 in the yaw direction is reproduced, and the sole behavior is recorded as Yaw2 (step S66).

  Further, with the posture in which the hip joint yaw axis is emphasized (see FIGS. 7 and 8), the stepping motion 1 in the yaw direction is reproduced, and the sole behavior is recorded as Yaw0 (step S67).

  Based on these measurement results, the dynamic motion characteristics of the robot apparatus are evaluated.

  First, it is determined whether or not the number of observed steps of each motion is the same as the planned value (step S68). If the observed number of steps is different from the planned value, the motion performance of the robot apparatus is readjusted (S75), and the process returns to step S61 to repeatedly evaluate the dynamic motion performance of the robot apparatus.

  Next, it is determined whether or not the number of times of both-leg support for each motion is the same as the planned value (step S69). If the number of times of both-leg support is different from the planned value, the robot apparatus is readjusted (S75), and the process returns to step S61 to repeatedly evaluate the dynamic movement performance of the robot apparatus.

  Next, it is determined whether or not the number of leaving / landing of each motion is the same as the planned value (step S70). If the number of times of getting off and landing is different from the planned value, the movement performance of the robot apparatus is readjusted (S75), and the process returns to step S61 to repeatedly evaluate the dynamic movement performance of the robot apparatus.

  Next, it is determined whether or not the number of times of leaving the sole inside of each motion is zero (step S71). If the number of times of floor leaving the plantar is not zero, after the robot device's motion performance is readjusted (S75), the process returns to step S61 to repeatedly evaluate the dynamic motion performance of the robot device.

  Next, it is determined whether or not the floor leaving time of the sole of each motion is within an allowable range (step S72). When the floor leaving time of the sole exceeds the allowable range, the motion performance of the robot apparatus is readjusted (S75), and the process returns to step S61 to repeatedly evaluate the dynamic motion performance of the robot apparatus.

  Next, it is determined whether or not the number of times of sole diagonal contact (see FIG. 16) of each motion is equal to or less than an allowable value (step S73). If the number of times that the plantar diagonal contact exceeds the allowable value, the robot apparatus is readjusted (S75), and then the process returns to step S61 to repeatedly evaluate the dynamic movement performance of the robot apparatus.

  Next, it is determined whether or not the number of separations in the single leg support period of each motion is zero (step S74). If the number of separations during the single leg support period is not zero, after the robot apparatus's movement performance is readjusted (S75), the process returns to step S61 to repeatedly evaluate the dynamic movement performance of the robot apparatus.

  If it is confirmed in steps S68 to S74 that sufficient dynamic motion characteristics can be obtained, the entire processing routine is terminated.

D. In this embodiment, the robot apparatus is obtained from the sensors F1 to F4 and F5-F8 for ground contact confirmation installed at the four corners of the sole while the robot apparatus reproduces a known diagnostic walking pattern. Record grounding information and other sensor information as a walking log. Then, by analyzing this log record, it is possible to perform evaluation and diagnosis by extracting characteristic points (or individual differences) unique to the robot apparatus. In this section, a robot motion characteristic diagnosis system will be described.

  FIG. 24 shows an example of an operation form of the motion evaluation of the robot apparatus by the diagnostic system.

  In order to evaluate the motion of the robot apparatus, motion evaluation patterns such as basic walking, one-step motion, and turning motion (described above) are prepared in advance. A walking log is collected as evaluation data for performing motion evaluation by inputting and reproducing these motion evaluation patterns into the robot apparatus.

  Diagnosis system diagnoses by analyzing walking log consisting of ground contact information obtained from sensors F1 to F4 for ground contact, F5 to F8, and other sensor information installed at the four corners of the sole. A characteristic point (or individual difference) unique to each target robot apparatus is extracted to obtain a diagnosis result. Of course, gait logs and their analysis results can be accumulated to allow statistical interpretation of diagnosis.

  Then, using the diagnosis result, the motion pattern for the normal operation of the robot apparatus is designed.

  The motion evaluation itself is performed on the actual device of the robot apparatus. However, other motion evaluation patterns, diagnosis systems, and diagnosis results may be inherent in the actual device, or may be stored on a server system outside the robot apparatus. You may make it manage.

  In the present embodiment, the diagnostic system basically checks whether the dynamic characteristics in the operation are appropriate based on the plantar contact determination. Based on the log records collected while the robotic device is executing a diagnostic walking pattern on the known road surface where the ideal plantar landing / leaving is known, feature points to be noted are extracted, The situation of the robot device is easily diagnosed by considering and judging.

  For example, when playing a walking pattern that leaves the bottom of the foot horizontally and lays down horizontally, the ground contact detection sensors at the four corners of the bottom of the foot are turned off at the same time when the aircraft is in the best condition. It will turn on (see FIG. 25).

  When performing motion evaluation in the diagnostic system, it is assumed that the static motion characteristics of the robot apparatus are appropriate. Then, the robot apparatus performs a walking road surface that is close to an ideal environment as a walking execution road surface, and the robot apparatus executes a known walking pattern (diagnostic walking pattern) in which the soles leave the floor horizontally and land horizontally.

  The diagnostic system extracts feature points from the walking log during the reproduction of the diagnostic walking pattern, and automatically makes a pass / fail judgment for the reference value. The diagnostic system also has various displays regarding processing results and a file output function.

  FIG. 27 shows a configuration example of a GUI main screen for determining the ground contact of the sole in the diagnostic system. In the sole grounding GUI main screen, the following three types of usage are prepared.

(1) Single log drag and drop A determination result is displayed by dragging and dropping an arbitrary log file to the main screen shown in the figure.

(2) A plurality of log files can be continuously processed by inserting a check box for automatically saving a plurality of log drag-and-drop results.

(3) Command line automatic execution

  FIG. 28 shows functions prepared on the main screen of the sole ground contact determination GUI. As shown in the figure, a pass / fail judgment belt window, a result auto-save check box, a log window, a log operation button, an evaluation details display button, a feature point list display button, and a landing graph display button are arranged on the main screen. It is installed.

  In the pass / fail judgment belt window, the judgment display is basically completed only by the contents of the window.

  By checking the result auto-save check box, the processing results are automatically and sequentially saved with the file name corresponding to the walking log after the sole contact determination processing by the system. When multiple log files are dragged and dropped, no result will be left unless this function is used (only the last processing result is displayed).

  The log window displays the internal processing status (processing status log). In order to check the algorithm at the time of introducing a new pass / fail judgment formula, etc., it has redundancy for outputting judgment midway information.

  The processing status log is deleted by operating the log operation button. Also, the specified processing status log is copied to the clipboard by operating the log copy button.

  The evaluation detail display button is a button for inputting an instruction to display a detailed breakdown of the determination result displayed on the pass / fail determination belt window in another window.

  The feature point list display button is a button for inputting an instruction so that a list of feature points extracted from the walking log by the diagnostic system is displayed in a separate window.

  The landing graph display button is a button for inputting an instruction to be displayed in a separate window in order to make it easy to understand the ground contact state as a visual state, which is noticed in the diagnostic system.

  29 and 30 show a configuration example of an evaluation detail screen displayed as a separate window by operating the evaluation detail display button. As shown in the figure, the screen is configured as a table in which an entry is provided for each evaluation item. Each entry includes a name, a diagnostic determination value, a pass / fail determination conditional expression, an item description, and a determination result for the corresponding determination item. Contains a field that stores

  In the determination field, “OK” or “NG” is displayed as “OK” or “NG”, and the operator is immediately glanced at the impossible items.

  In the determination item name field, the determination item name described in the diagnosis pass / fail setting file is described.

  In the diagnosis determination value field, a value for each determination item analyzed or calculated in the diagnosis system from the walking log is described.

  In the pass / fail judgment expression field, the pass / fail judgment formula is described together with the judgment item name in the diagnosis pass / fail setting file.

  In the diagnosis determination value field, a diagnosis determination value related to the determination item is described. That is, if this pass / fail judgment formula is true, it is “OK”, and if it is false, it is “NG”.

  In the item description, a comment text stored in the diagnosis pass / fail setting file together with the determination item name is described. Here, it is presented to the operator for each item whether it is “NG” and how to deal with it.

  31 and 32 show a configuration example of a feature point list display screen displayed as a separate window by operating the feature point list display button. As shown in the figure, the screen is configured as a table in which an entry is provided for each feature point name, and each entry includes fields for storing a feature point name, a feature point value, and a feature point description.

  A corresponding feature point name is described in the feature point name field. In the diagnostic system, a value extracted from the walking log is called a feature point as a determination focus point. This feature point can be used as a variable of an arbitrary determination formula.

  In the feature point value field, the value of the feature point extracted from the walking log currently being processed is described.

  In the feature point explanation field, what kind of feature point is, what kind of intention is focused on, and the like are presented to the worker in an easy-to-understand manner.

  FIG. 33 and FIG. 34 show configuration examples of the landing graph screen displayed as a separate window by the operation of the landing graph display button. As shown in the figure, this screen displays a graph of the temporal transition of the distance to the floor surface of each of the left and right soles together with the number of grounding points. In the illustrated screen, a graph scale button, a ground state display button, and a current display time are arranged.

  The right foot landing graph / left foot landing graph is a graph of the transition of the landing state of each of the soles. In these graphs, the horizontal axis represents time, the vertical axis represents the number of contact points of the sole, and the height of the sole when the diagnostic walking is performed (the rectangle is the number of contact points and the curve is the height of the sole).

  By operating the graph scale button, the time step width of the right foot landing graph / left foot landing graph changes according to the amount of operation, and the entire graph and a part of the graph are visually observed in detail. Is possible.

  When the right foot landing graph / left foot landing graph is pointed with the mouse, the landing state of both soles at the time is displayed in the ground state display / current time display area. Since only the number of contact points is expressed in the right foot landing graph / left foot landing graph, the detailed ground contact state can be confirmed here.

  The distance from the bottom of the leg to the floor can be calculated by, for example, inverse kinematics calculation with log information of each joint angle constituting the leg. Further, the number of grounding points referred to here is the number of on-states of the grounding confirmation sensors F1 to F4 and F5 to F8 provided at the four corners of the sole of each leg. When performing ideal horizontal landing and horizontal leaving, the number of grounding points can be either 4 or 0.

  In addition, it is possible to start the ground contact determination processing by the diagnostic system on the character-based command line without using the main GUI screen for determining the ground contact as shown in FIG. 27 (not shown). ). In this case, by specifying a desired walking log file and starting the corresponding application, the processing result can be output as a file without going through the main screen. By using such a command line execution function, use from various script languages and integration processing of other management systems can be performed.

  For example, in the manufacturing process of the robot apparatus, the diagnosis system that performs the sole contact determination is used by the command line function from the integrated processing mechanism as part of its own processing.

  In the use stage of the robot apparatus, a main screen as shown in FIG. 27 is used. At this time, the portability of the diagnostic system can be realized by carrying only the functional module that determines the contact of the sole. For example, even a lightweight notebook PC that can be carried can sufficiently execute the function of determining the base of the foot of the diagnostic system.

  FIG. 35 shows an example of a usage pattern of the diagnostic system according to the present embodiment. In the example shown in the figure, the robot apparatus itself (actual machine), a computer system (PC) such as a personal computer that realizes an integrated processing function, and a database that accumulates diagnosis results are included.

  The robot apparatus and the PC are connected via a wireless LAN, a wired cable, or other transmission media. The robot apparatus executes a diagnostic walking having a known walking pattern in response to an instruction to perform a diagnostic walking from the PC, and records a log of a ground contact confirmation sensor and other sensor information during the execution.

  When the PC obtains the log record from the robot apparatus, the PC activates the diagnostic system by, for example, a command line execution function, and performs diagnostic processing such as foot sole contact determination. Then, the obtained diagnosis result is registered in the database. Thereafter, the PC also functions as a database retrieval terminal, and can perform database output and confirmation work by searching the database for diagnosis results obtained in the past.

  Further, on the database side, diagnosis results are accumulated, but statistical interpretation of diagnosis such as calculation of lot tendency may be performed. Then, by using the result of such statistical processing, a new pass / fail judgment conditional expression relating to the sole contact determination can be sequentially generated, or a guideline for the diagnosis of the next lot can be presented. Therefore, the diagnostic results are accumulated, and the diagnostic results can be effectively used for the design and production of the next robot apparatus.

  FIG. 36 shows another example of the usage pattern of the diagnostic system according to this embodiment. In the illustrated example, the robot apparatus itself (actual machine) and a computer system (PC) such as a personal computer that realizes an integrated processing function are configured. Unlike the configuration example shown in FIG. 35, the database is not included. However, the PC may be able to access a remote database server via a transmission medium (not shown) such as the Internet, and the diagnosis results may be accumulated and statistically interpreted as described above.

  The robot apparatus and the PC are connected via a wireless LAN, a wired cable, or other transmission media. On the PC, the pass / fail criterion for each operation content of the robot apparatus is edited and the robot is instructed to execute a diagnostic walk.

  The robot apparatus executes a diagnostic walking having a known walking pattern in response to an instruction to perform a diagnostic walking from the PC, and records a log of a ground contact confirmation sensor and other sensor information during the execution.

  Then, on the PC side, when a log record is acquired from the robot apparatus, the diagnostic system is activated and diagnostic processing such as sole contact is performed. Although not shown, the obtained diagnosis result may be registered in a database.

  The usage mode as shown in FIG. 36 is effective, for example, when various demonstrations are performed by the robot apparatus, or when the user performs self-diagnosis of the robot apparatus.

  In addition, the feature point said in this specification is the attention point which the said diagnostic system analyzes or extracts from log records, such as a walk log. What feature points are generated is determined in advance. In addition, names that do not overlap are assigned to the feature points, and can be used in pass / fail judgment formulas in the diagnostic pass / fail setting file.

  The diagnosis pass / fail setting file is a script-format file describing the diagnosis pass / fail conditions for each determination item, and is basically described according to a pass / fail determination format of “(item name) = (determination formula); comment”.

  FIG. 37 shows an example of a diagnosis pass / fail setting file. As shown in the figure, the pass / fail judgment is basically described in the form of one item and one line.

  In the same file, the feature points can be used as reserved words in the pass / fail judgment formula. After “;”, it is treated as a comment, and the comment behind the pass / fail judgment formula is displayed together with the pass / fail judgment specially on the evaluation details screen. The determination formula can specify the determination method by the following three types of notation.

(A) Describe one coincidence constant or feature point name. "OK" if exact match
(B) Describe two range range constants or feature point names separated by “,”. OK between two values
(C) Invalid Nothing is described. Calculate but pass / fail judgment is not made as a reference value

  Hereinafter, a diagnosis flow of the robot apparatus using the diagnosis system according to the present embodiment will be described.

  FIG. 38 schematically shows the diagnosis flow.

  When the robot apparatus executes a diagnostic walking pattern in a known diagnostic walking execution environment, a walking log including a set of diagnostic data is acquired.

  Here, the diagnostic walking pattern is created based on the actual machine model information (device configuration, etc.) of the robot apparatus on, for example, a motion creation system (described above). The diagnostic walking pattern includes basic walking, one-step walking, turning motion, and the like (described above).

  The diagnostic data includes the grounding information obtained from the grounding confirmation sensors F1 to F4 and F5 to F8 provided at the four corners of the left and right soles when the left and right legs execute a walk, It consists of sensor information.

  The diagnostic system extracts feature points of the robot device by acquiring the walking log and diagnostic walking execution environment information when the diagnostic walking pattern is executed, and performing predetermined diagnostic processing such as statistical processing of the walking log. Etc. can be obtained.

  The diagnosis process is obtained by comparing the walking log with a predetermined diagnosis pass / fail condition (see FIG. 37 described above). The diagnosis process is performed by starting predetermined diagnosis software (including command line execution) on a computer system such as a personal computer.

  The diagnostic result is quantitative data. Therefore, it is possible to manage the diagnosis results obtained from a plurality of robot devices as a database and perform statistical consideration. In this case, based on the database, it is possible to construct a cycle in which the influence on the motion performance of the robot apparatus on the production line is analyzed to improve the productivity and the support system.

  Further, it is possible to deal with a problem with a bad diagnosis based on the diagnosis result of each robot. Information on how to deal with this problem can also be managed in the database, so that a cycle of improving productivity and improving the support system can be established.

  FIG. 77 shows a log file format example of the diagnostic walking pattern. In the illustrated example, the CSV (Comma Separated Value) format is basically used, the upper end is the title line, and the left end is the log recording time. A log having the planned amount information of the diagnostic walking motion and the actual sensor information when the diagnostic walking motion is performed is referred to as a “walking log”. The sensor information record includes a sole contact state record, various posture sensor information, and an output information record of all joints.

  FIG. 39 illustrates an overview of the diagnostic processing in the diagnostic system according to the present embodiment.

  The walking log is a set of diagnostic data acquired by the robot apparatus executing diagnostic walking patterns such as basic walking, one-step system, and turning motion.

  In the diagnosis system, each feature point (see FIG. 31) as a determination target point is extracted from the walking log and counted as a variable of the determination formula. Then, the feature point value for each feature point is registered in the database in association with the feature point name.

  Statistical consideration is possible on the database. That is, it is possible to perform processing for referring to the feature points from the database, adding further consideration and analysis, and registering new feature points in the database.

  In addition, it is possible to make a pass / fail determination according to the diagnosis pass / fail condition while referring to the feature point database. The diagnosis pass / fail condition is as shown in FIG. It can be described as a script-type pass / fail judgment expression.

  The pass / fail determination result according to the diagnosis pass / fail condition is output externally as a diagnosis result of the walking log by the diagnosis system. There are various output modalities such as display output on a monitor screen, printout, and file storage. In the case of display output, the form of drawing such as list display and graph display is arbitrary.

  FIG. 40 shows a procedure of the entire diagnosis process of the robot apparatus by the diagnosis system according to the present embodiment in the form of a flowchart.

  Execute the diagnostic walking pattern on the robot device, and input the actual walking model information, diagnostic walking pattern, diagnostic walking execution environment information, etc. about the robot device to be diagnosed into the diagnostic system along with the obtained walking log Is done.

  In the diagnosis system, first, analysis processing is performed to extract the transition of the ground contact point or to reproduce the body motion.

  Subsequently, extraction processing of each feature point as a determination target point is performed. Specifically, the feature point extraction process includes a step conversion process, a contact state statistical process, a gait period determination process, a gait period progress record, and the like.

  Subsequently, a diagnosis process is performed based on the extracted feature point data. For example, the following diagnosis is performed.

● Planned number of steps diagnosis ● Left and right leg number of steps diagnosis ● Number of times of support for both legs ● Number of times of getting out of bed ● Diagnosis of number of times of getting out of foot sole ● Diagnosis of time of foot leaving phase ● Diagnosis of time of foot landing phase ● Diagnosis of number of times of foot leaving failure ● Feet Diagnosis of the number of floor landing failures ● Single leg support period instability diagnosis ● Single leg support period peeling diagnosis ● Bottom sole non-contact state transition time diagnosis

  And the diagnostic result of the walk log by the said diagnostic system is output outside. There are various output modalities such as display output on a monitor screen, printout, and file storage. In the case of display output, the form of drawing such as determination list display and landing graph display is arbitrary.

  FIG. 41 shows the procedure of the step conversion process, which is one of the feature point extraction processes, in the form of a flowchart. The step conversion process is intended to extract various conversion values related to “steps” as feature points.

  The ground contact state is extracted from the walking log input to the system according to the time series.

  First, it is determined whether the sole state taken out is full grounding, surface grounding, or no installation (step S101).

  If the sole state is full contact or surface contact, it is determined whether or not the stack is ungrounded (step S102). If the stack is not grounded, the number of times of landing is added and the grounded state is stored in the stack (step S103).

  On the other hand, if the sole state is ungrounded, it is determined whether the stack is fully grounded or surface grounded (step S104). If the stack is fully grounded or surface-grounded, the number of times of bed leaving is added and the grounded state is stored in the stack (step S105).

  Such processing is repeated until the end of the walking log.

  According to the above-described processing procedure, the number of steps is calculated independently for each of the left and right legs based on the ground contact state transition of the sole, and the number of times of getting out and the number of times of landing are converted at each sole. Since the feet that have left the floor must land, the number of times of leaving the floor = the number of times of landing = the number of steps of the leg is obtained. Therefore, through the above processing procedure, as shown by a graph in FIG. 42, it is possible to calculate the number of steps with high robustness against a slight fluttering of the sole.

  FIG. 43 shows a procedure of a ground state statistical process, which is one of feature point extraction processes, in the form of a flowchart. In this process, what kind of grounding state has occurred in the transition of the ground contact state during the diagnostic walking is extracted as a feature point.

  The ground contact state is extracted from the walking log input to the system according to the time series. And it branches according to a grounding state, and an addition process is performed about the applicable grounding state. Such processing is repeated until the end of the walking log.

  Here, the entire grounding is a state in which all of the four grounding confirmation sensors F1 to F4 or F5 to F8 arranged at the four corners of the sole detect the grounding, and there is only one pattern. Surface grounding is a state in which three of the four grounding confirmation sensors arranged at the four corners of the sole detect grounding, and there are four patterns. The line grounding is a state in which two of the four grounding confirmation sensors arranged at the four corners of the sole detect grounding, and four patterns on each edge of the sole and two patterns on the diagonal. There is. Further, no grounding is a state in which all four grounding confirmation sensors arranged at the four corners of the sole do not detect grounding, and there is only one pattern. Point grounding is a state in which one of the four grounding confirmation sensors arranged at the four corners of the sole detects grounding, and there are four patterns at each corner of the sole.

  FIG. 44 shows the procedure of gait period processing, which is one of feature point extraction processing, in the form of a flowchart. This process is similar to the above-described step conversion process, but its purpose is to extract the transitional state of the leg track during getting out of bed and landing as a feature point.

  The ground contact state is extracted from the walking log input to the system according to the time series. Then, the gait period is extracted and the transition is sequentially recorded. Such processing is repeated until the end of the walking log.

  In the gait period determination process, the transition of the gait period is recorded independently for each of the left and right legs based on the ground contact state transition of the sole. FIG. 45 shows an image of gait period transition data. The gait period here is one of the ways of dividing the leg trajectory in actual walking in stages.

  Moreover, since it is a time transition record according to a walk log, it records according to time for every division | segmentation of a period. This is extracted as, for example, a feature point “how long the transient situation has continued in time”.

  FIG. 46 shows a processing procedure for recording the gait progress, which is one of the feature point extraction processes, in the form of a flowchart. This process is similar to the above-described gait period determination process, but its purpose is to extract feature points focused on both legs.

  The ground contact state is extracted from the walking log input to the system according to the time series. Then, taking into account the gait periods of the left and right legs simultaneously, the gait transitions of both legs are sequentially recorded. Here, the process is performed based on the “gait period transition” extracted in the “gait period determination process”. Such processing is repeated until the end of the walking log. FIG. 47 shows an example of recorded data of gait progress for the left and right legs.

  From the gait period transition on the left and right, the gait period transition taking into account both legs simultaneously = “gait period progression record” is generated. Moreover, since it is a time transition record according to a walk log, it records according to time for every division | segmentation of a period. This is extracted as, for example, a feature point “how long it took to land the right leg”.

  FIG. 48 schematically shows a diagnostic processing flow performed based on each feature point data extracted in this way. As shown in the figure, the diagnosis process is the selection of the determination formula, and the “diagnosis pass / fail condition” indicating the determination threshold value, and the calculation of various feature point information calculated as substitution values for these formulas. "Is output as the answer.

  The feature points extracted directly from the diagnostic walking pattern are as follows.

● Planned steps: correct number of steps in the diagnostic walking pattern ● Right foot planned steps: right step in the right leg in the diagnostic walking pattern ● Left planted steps: correct number of steps in the left leg in the diagnostic walking pattern

  The feature points extracted from the step conversion process are as follows.

● Number of steps: Number of steps actually taken by the actual machine ● Number of steps + 1: Number of steps actually taken by the actual machine + 1 (corresponds to the number of support periods for both legs)
● Right leg steps: The number of steps of the right leg that the actual machine actually walks ● Left leg steps: The number of steps of the left leg that the actual machine actually walks

  Further, the feature points extracted from the ground state statistical processing are as follows.

● Number of right foot diagonal contact: This is the number of times the right foot is in diagonal contact, and is added if the road surface during diagnostic walking is uneven or the sole is deformed.
● Left sole diagonal contact number: This is the number of times that the left sole is in a diagonal contact state, and is added when the road surface at the time of diagnostic walking is uneven or the sole is deformed.

  Further, the feature points extracted from the gait period determination process are as follows.

● Number of support periods for both legs: Number of support periods for both legs ● Number of right foot floor leaving periods: Number of right foot floor leaving periods ● Number of left foot floor leaving periods: Number of left foot floor leaving periods ● Number of right foot floor landing periods: Number of right foot floor landing periods: ● Number of times of left foot landing period: Number of times of left foot floor landing period ● Number of times of right foot sole inner floor leaving: The number of times the foot sole leaves the inner side when the right foot sole leaves the floor, and the aircraft may flow outward Add in case.
● Number of left foot sole inside floors: This is the number of times the soles have left the inside when leaving the left sole, and is added when there is a possibility that the aircraft is flowing outside.
● Average time of right foot floor leaving period: This is the time average of right foot floor leaving period, and it becomes longer when the right leg movements for some reason in the leg trajectory plan.
● Left plantar bed removal time average: This is the time average of the left plantation bed removal period, and it becomes longer when the left leg motion stutters for some reason with respect to the leg trajectory plan.
● Right foot floor landing time average: This is the time average of the right foot floor landing phase, and becomes longer when the right leg moves for any reason against the leg trajectory plan.
● Left planting time average: This is the time average of the left plantation landing period, and becomes longer when the left leg movements for any reason with respect to the leg trajectory plan.
● Right foot bottom ungrounded state transition minimum time average: Average time from surface contact state to complete leaving floor ● Left foot bottom ungrounded state transition minimum time average: Surface time from ground contact state to complete floor leaving ● Right foot Bottom non-surface contact state transition minimum time average: average time from free leg state to surface contact (or full surface contact) Left sole non-surface contact state transition minimum time average: free leg state to surface contact (or full surface contact) Average time to complete

  The feature points extracted from the gait progress record are as follows.

● Number of errors during the support period for both legs: This is the number of errors during the support period for both legs.
● Right foot landing failure count: This is the number of times the right sole has failed to land, and is added when the ground contact state collapses immediately after the sole has completely landed.
● Left planting failure frequency: This is the number of times the left plantar has failed to land, and is added when the grounding state collapses immediately after the sole has completely landed.
● Right foot floor failure failure count: This is the number of times the right foot sole has failed to get out of the floor, and is added when the foot sole has left the floor immediately after it has completely left the floor.
● Left plantation failure count: This is the number of times the left sole has failed to get out of the floor, and is added when the plantation has just landed immediately after the sole has completely left.
● Both-leg support period instability frequency: Number of times the ground contact state changed during the both-leg support period (8-5 points contact → 4 points contact)
● Right monopod support period instability: The number of times the ground contact state changed during the right monopod support period (4-point contact → 3-point contact), which corresponds to a dangerous case.
● Left monopod support period instability: This is the number of times the ground contact state changed during the left monopod support period (4-point contact → 3-point contact), which corresponds to a very dangerous case.
● Separation frequency of both legs support period: Number of times the number of ground contact points fell below 3 points during support period of both legs ● Right frequency of support period of single leg support period: Number of times the right sole peeled during the support period of right leg (4 points contact → 2.1)・ This is a very dangerous case and may fall over.
● Left monopod support period peeling frequency: This is the number of times the left sole is detached during the left monopod support period (4-point grounding → 2, 1/0 point grounding). is there.
● Right monopod support period Left leg gait period interference frequency: The number of times the left leg gait transition and right leg gait transition overlap during the right monopod support period, while the other leg is still landing Nevertheless, it is added when the other leg starts to get out of bed.
● Left monopod support period Right leg gait period interference frequency: This is the number of times the right leg gait transition and left leg gait transition overlap in the left monopod support period, while the other leg is still on the ground Nevertheless, it is added when the other leg starts to get out of bed.

  Based on the feature point data obtained by the feature point extraction process, a diagnosis pass / fail decision is made according to the diagnosis pass / fail condition, whereby the diagnosis process is performed. Each diagnosis pass / fail condition is described as a determination formula in a script format.

  FIG. 49 shows the diagnosis pass / fail determination conditions regarding the planned number of steps. In the planned step count diagnosis, the number of steps in the diagnostic walking pattern and the number of times the robot apparatus actually walks are diagnosed. If they do not match, it can be determined that the vehicle has fallen or the foot sensor is inadequate.

  50 and 51 show the pass / fail judgment conditions for the planned step count diagnosis for the left and right legs. In these cases, when the planned number of steps is not possible, a diagnosis for narrowing down the cause of the right or left foot is performed.

  FIG. 52 shows the diagnosis pass / fail determination condition regarding the number of times of both-leg support period. In the case of a diagnostic walking pattern that starts first from the both-leg support period, the both-leg support period is the number of steps plus one. If the measured number of both-leg support periods does not match this, it can be determined that the sole is not properly grounded before the determination is started, or that the robot apparatus has fallen.

  53 and 54 show diagnosis pass / fail determination conditions regarding the number of bed leaving periods for the left and right soles. The number of right foot lift-off periods coincides with the number of right foot steps. If the measurement result does not agree with this, it can be determined that there was a significantly unstable bed leaving the right sole. The same applies to the left sole.

  55 and 56 show diagnosis pass / fail determination conditions regarding the number of times of landing for the left and right soles. The number of right foot landing periods coincides with the number of right foot steps. If the measurement result does not agree with this, it can be determined that there was a significantly unstable landing on the right sole. The same applies to the left sole.

  FIG. 57 and FIG. 58 show the diagnostic pass / fail judgment conditions related to the number of inner bed leavings for the left and right soles. When the number of floor leavings on the inside of the sole does not become zero, it is considered that the robot apparatus main body is flowing outward because the flooring is performed from the inside. In such a case, the robot apparatus is diagnosed as being in a very dangerous state.

  59 and 60 show diagnosis pass / fail determination conditions regarding the average bed leaving time for the left and right soles. If the average time required for getting out of bed is equal to or less than a predetermined time, it is judged as good, and if it takes too long, it is diagnosed that the response of the robot apparatus is not good.

  61 and 62 show diagnosis pass / fail judgment conditions regarding the average landing time for the left and right soles. If the average time required for landing is less than or equal to a predetermined time, it is judged as good, and if it takes too much time, it is diagnosed that the response of the robot apparatus is not good.

  FIG. 63 and FIG. 64 show the diagnostic pass / fail judgment conditions regarding the average time of the minimum surface non-contact state transition for the left and right soles. The determination condition is a diagnostic item that pays particular attention to the transitional state of getting out of bed. If the average time from the start of getting off to the end of getting completely off is less than a predetermined time, it is judged as good, and if it takes too much time, it is diagnosed that the response of the robot apparatus is not good.

  FIG. 65 and FIG. 66 show the diagnostic pass / fail judgment conditions regarding the minimum non-surface contact state transition time average for the left and right soles. The determination condition is a diagnostic item that pays particular attention to the transient state of landing. If the average time from the start of landing until the completion of landing is not more than a predetermined time, it is judged as good, and if it takes too much time, it is diagnosed that the response of the robot apparatus is not good.

  67 and 68 show diagnosis pass / fail determination conditions regarding the number of times of landing failure for the left and right soles. The determination condition is a diagnostic item paying attention to the landing situation of the left and right feet. If you flutter when landing, the number of landing failures increases. When the number of times exceeds a predetermined value, it is possible to determine a significant looseness of the mechanism.

  69 and 70 show diagnosis pass / fail determination conditions regarding the number of bed leaving failures for the left and right soles. The determination condition is a diagnostic item paying attention to the left and right foot leaving situations. When trying to get out of bed, the number of times of getting out of bed increases when the sole does not rise easily. When the number of times exceeds a predetermined value, it is possible to determine whether the joint output is insufficient, the response is insufficient, or the mechanism is significantly loosened.

  71 and 72 show diagnosis pass / fail determination conditions regarding the number of diagonal contact points for the left and right soles. The number increases when there is a diagonal contact of the corresponding sole, and when it exceeds a predetermined value, it can be determined that the corresponding sole may be deformed.

  73 and 74 show diagnosis pass / fail determination conditions regarding the single-leg support period instability degrees for the left and right legs, respectively. The frequency increases when the corresponding leg is staggered. When the frequency exceeds a predetermined value, it can be diagnosed that stable walking is difficult.

  75 and 76 show diagnosis pass / fail determination conditions regarding the single-leg support period peeling frequency for the left and right legs. It is possible to find a situation where the sole of the corresponding leg is detached during the single leg support period. Since the separation of the sole on the support leg side is synonymous with a fall just because the fall did not result, it is diagnosed as a fall.

  The present invention has been described in detail above with reference to specific embodiments. However, it is obvious that those skilled in the art can make modifications and substitutions of the embodiments without departing from the gist of the present invention.

  The gist of the present invention is not necessarily limited to a product called a “robot”. That is, as long as it is a mechanical device that performs an exercise resembling human movement using an electrical or magnetic action, the present invention can be applied to products belonging to other industrial fields such as toys. Can be applied.

  In short, the present invention has been disclosed in the form of exemplification, and the description of the present specification should not be interpreted in a limited manner. In order to determine the gist of the present invention, the claims section described at the beginning should be considered.

FIG. 1 is a view showing a state in which a “humanoid” or “humanoid” robot apparatus 100 used for carrying out the present invention is viewed from the front. FIG. 2 is a view showing a state in which the “humanoid” or “humanoid” robot apparatus 100 used for carrying out the present invention is viewed from the rear. FIG. 3 is a diagram schematically illustrating a joint degree-of-freedom configuration included in the robot apparatus 100. FIG. 4 is a diagram schematically showing a control system configuration of the robot apparatus 100. FIG. 5 is a diagram for explaining the influence on the key pose and the roll direction formed by the robot apparatus when evaluating the left-right asymmetry as the static motion characteristic evaluation of the robot apparatus. FIG. 6 is a diagram for explaining the influence on the key pose formed by the robot device and the pitch direction when evaluating the left-right asymmetry as the static motion characteristic evaluation of the robot device. FIG. 7 is a diagram for explaining the influence on the key pose formed by the robot apparatus and the yaw axis direction when evaluating the left-right asymmetry as the static motion characteristic evaluation of the robot apparatus. FIG. 8 is a diagram for explaining the influence on the key pose formed by the robot device and the yaw axis direction when evaluating the left-right asymmetry as the static motion characteristic evaluation of the robot device. FIG. 9 is a diagram for explaining a method for evaluating the static motion characteristics in the roll direction in the adaptive control system of the robot apparatus. FIG. 10 is a diagram for explaining a method of evaluating the static motion characteristics in the pitch direction in the adaptive control system of the robot apparatus. FIG. 11 is a diagram for explaining a procedure for evaluating the dynamic motion characteristics of the robot apparatus. FIG. 12 is a diagram illustrating an example of a one-step motion for evaluating the dynamic motion characteristics of the robot apparatus. FIG. 13 is a diagram illustrating an example of a one-step motion for evaluating the dynamic motion characteristics of the robot apparatus. FIG. 14 is a diagram illustrating an example of a one-step motion for evaluating the dynamic motion characteristics of the robot apparatus. FIG. 15 is a diagram for explaining the minimum time for transition to the bottom surface non-ground state. FIG. 16 is a diagram for explaining the number of planted diagonal contact points. FIG. 17 is a diagram illustrating an example of a turning motion for evaluating the dynamic motion characteristics of the robot apparatus. FIG. 18 is a flowchart showing a schematic processing procedure for performing motion characteristic evaluation in the robot apparatus. FIG. 19 is a flowchart showing a processing procedure for performing motion characteristic evaluation in a robot apparatus having a plurality of movable legs. FIG. 20 is a flowchart showing a processing procedure for evaluating static motion characteristics of a robot apparatus designed and manufactured symmetrically. FIG. 21 is a flowchart showing a processing procedure for evaluating a static motion characteristic of a legged mobile robot device configured by connecting both symmetrical legs to the waist portion 41. FIG. 22 is a flowchart showing a schematic processing procedure for evaluating dynamic motion characteristics in a robot apparatus having a contact point with the road surface and using ZMP as a stability criterion. FIG. 23 is a flowchart showing a processing procedure for evaluating dynamic motion characteristics in a robot apparatus having a waist portion 41 and a plurality of movable legs and using ZMP as a stability criterion. FIG. 24 is a diagram illustrating an example of an operation mode of motion evaluation of the robot apparatus by the diagnosis system. FIG. 25 is a diagram for explaining a basic idea of the diagnostic system. FIG. 26 is a diagram for explaining the basic idea of the diagnostic system. FIG. 27 is a diagram illustrating a configuration example of a GUI main screen for determining the ground contact of the sole in the diagnostic system. FIG. 28 is a diagram for explaining the function of the main screen of the sole ground contact determination GUI. FIG. 29 is a diagram showing a configuration example of the evaluation details screen. FIG. 30 is a diagram for explaining the configuration of the evaluation details screen. FIG. 31 is a diagram showing a configuration example of a feature point list display screen. FIG. 32 is a diagram for explaining the configuration of the feature point list display screen. FIG. 33 is a diagram illustrating a configuration example of the landing graph screen. FIG. 34 is a diagram for explaining the configuration of the landing graph screen. FIG. 35 is a diagram illustrating an example of a usage pattern of the diagnostic system. FIG. 36 is a diagram showing another example of the usage pattern of the diagnostic system. FIG. 37 is a diagram showing an example of a diagnostic pass / fail setting file. FIG. 38 is a diagram schematically showing a diagnosis flow. FIG. 39 is a diagram for explaining the contents of the diagnostic processing in the diagnostic system. FIG. 40 is a flowchart showing a procedure of the entire diagnosis process of the robot apparatus by the diagnosis system. FIG. 41 is a flowchart showing the steps of the step conversion process. FIG. 42 is a graph showing the result of the step conversion process. FIG. 43 is a flowchart showing a procedure of the ground state statistical processing. FIG. 44 is a flowchart showing the procedure of gait period processing. FIG. 45 is a diagram showing the results of gait period determination processing. FIG. 46 is a flowchart showing a processing procedure for recording the gait progress, which is one of the diagnosis processes. FIG. 47 is a diagram showing an example of gait phase progression recording data. FIG. 48 is a diagram schematically showing a diagnostic processing flow performed based on the feature point data. FIG. 49 is a diagram showing diagnosis pass / fail determination conditions regarding the planned number of steps. FIG. 50 is a diagram illustrating diagnosis pass / fail determination conditions regarding the planned number of steps for the right leg. FIG. 51 is a diagram showing diagnosis pass / fail determination conditions regarding the planned number of steps for the left leg. FIG. 52 is a diagram showing diagnosis pass / fail determination conditions regarding the number of times of both-leg support period. FIG. 53 is a diagram illustrating diagnosis pass / fail determination conditions regarding the number of times of right foot sole bed leaving. FIG. 54 is a diagram illustrating diagnosis pass / fail determination conditions regarding the number of times of leaving the left sole. FIG. 55 is a diagram illustrating diagnosis pass / fail determination conditions regarding the number of right foot landing periods. FIG. 56 is a diagram showing diagnosis pass / fail determination conditions regarding the number of times of landing on the left sole. FIG. 57 is a diagram showing diagnosis pass / fail judgment conditions regarding the number of times of getting out of the right sole inside the floor. FIG. 58 is a diagram illustrating diagnosis pass / fail determination conditions regarding the number of times of leaving the left sole inside the bed. FIG. 59 is a diagram showing diagnosis pass / fail determination conditions regarding the average right foot floor getting-off time. FIG. 60 is a diagram showing diagnosis pass / fail determination conditions regarding the average left foot sole leaving time. FIG. 61 is a diagram illustrating diagnosis pass / fail determination conditions regarding the right foot sole landing time average. FIG. 62 is a diagram showing diagnosis pass / fail determination conditions regarding the average left foot landing time. FIG. 63 is a diagram showing diagnosis pass / fail determination conditions regarding the minimum time average of the right foot bottom non-ground state transition. FIG. 64 is a diagram showing diagnosis pass / fail determination conditions regarding the minimum time average of the left foot bottom non-ground state transition. FIG. 65 is a diagram showing diagnosis pass / fail determination conditions regarding the minimum time average of right foot sole non-surface contact state transition. FIG. 66 is a diagram showing diagnosis pass / fail determination conditions regarding the minimum time average of left sole non-surface contact state transition. FIG. 67 is a diagram showing diagnosis pass / fail determination conditions regarding the number of times of right foot sole landing failure. FIG. 68 is a diagram illustrating diagnosis pass / fail determination conditions regarding the number of times of left sole landing failure. FIG. 69 is a diagram illustrating diagnosis pass / fail determination conditions regarding the number of times of failure to get off the right sole. FIG. 70 is a diagram illustrating diagnosis pass / fail determination conditions related to the number of times of left foot bottom bed failure. FIG. 71 is a diagram showing diagnosis pass / fail determination conditions related to the number of right foot diagonal contact points. FIG. 72 is a diagram showing diagnosis pass / fail determination conditions regarding the number of diagonal contact with the left sole. FIG. 73 is a diagram showing diagnosis pass / fail determination conditions regarding the right monopod support period instability frequency. FIG. 74 is a diagram showing diagnosis pass / fail determination conditions regarding the left monopod support period instability frequency. FIG. 75 is a diagram showing diagnosis pass / fail determination conditions regarding the right single leg support period peeling frequency. FIG. 76 is a diagram showing diagnosis pass / fail determination conditions regarding the left single leg support period peeling frequency. FIG. 77 is a diagram showing a log file format example of a diagnostic walking pattern.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 ... Neck joint yaw axis 2A ... 1st neck joint pitch axis 2B ... 2nd neck joint (head) pitch axis 3 ... Neck joint roll axis 4 ... Shoulder joint pitch axis 5 ... Shoulder joint roll axis 6 ... Upper arm yaw axis 7 ... Elbow joint pitch axis 8 ... Wrist joint yaw axis 9 ... Trunk pitch axis 10 ... Trunk roll axis 11 ... Hip joint yaw axis 12 ... Hip joint pitch axis 13 ... Hip joint roll axis 14 ... Knee joint pitch axis 15 ... Ankle joint pitch Axis 16 ... Ankle joint roll axis 30 ... Head unit, 40 ... Trunk unit, 41 ... Lumbar unit 50 ... Arm unit, 51 ... Upper arm unit 52 ... Elbow joint unit, 53 ... Forearm unit 60 ... Leg unit, 61 ... thigh unit 62 ... knee joint unit, 63 ... ankle unit 80 ... control unit, 81 ... main control unit 82 ... peripheral circuit 91, 92 ... grounding confirmation sensor 93, 94 ... acceleration sensor 95: Posture sensor 96 ... Acceleration sensor 100 ... Legged mobile robot

Claims (17)

In a motion evaluation method for a robot apparatus having a plurality of movable parts,
The robot apparatus is configured substantially symmetrically, and controls the driving unit adaptively using the driving unit that drives the movable unit, the measuring unit that measures the state quantity of the moving unit, and the measured state quantity, or A drive control unit that performs non-adaptive control is provided,
Forming a symmetrical posture; and
Performing the non-adaptive control in the drive control unit when the symmetric posture is formed, and evaluating the drive unit;
When the symmetric posture is formed, adaptive and non-adaptive control is alternately executed in the drive control unit, and the measurement unit or the drive is based on the measurement result by the measurement unit at the time of each control. A step of evaluating the control unit;
A motion evaluation method for a robot apparatus, comprising: In a motion evaluation method for a robot apparatus having a plurality of movable parts,
The robot apparatus is configured substantially symmetrically, and controls the driving unit adaptively using the driving unit that drives the movable unit, the measuring unit that measures the state quantity of the moving unit, and the measured state quantity, or A drive control unit that performs non-adaptive control is provided,
Forming an asymmetric posture; and
Performing a left-right asymmetry evaluation of the drive unit when the left-right asymmetric posture is formed;
A motion evaluation method for a robot apparatus, comprising: In the step of forming the left-right asymmetric posture, the left-right reversed posture is alternately executed as a set,
The method for evaluating a motion of a robot apparatus according to claim 2. The step of evaluating the left-right asymmetry is as follows.
When the left-right asymmetry posture is formed, first non-adaptive control is performed in the drive control unit and a position gain in the drive unit is set high, and the left-right asymmetry of the drive unit is evaluated. The evaluation steps of
When the left-right asymmetric posture is formed, non-adaptive control is executed in the drive control unit, and the position gain in the drive unit is set low, thereby evaluating the left-right asymmetry of the low-frequency torque in the drive unit A second step of performing
When the left-right asymmetric posture is formed, the drive control unit executes adaptive control, and based on the measurement results by the measurement unit at the time of each control, the left-right asymmetry in the measurement unit or the drive control unit A third step of evaluating
The motion evaluation method for a robot apparatus according to claim 2, further comprising: The robot apparatus includes at least a trunk and left and right legs connected to the trunk,
In the step of evaluating the left-right asymmetry, the roll direction depends on whether the difference in the amount of tilt in the roll direction of the trunk that occurs when the left leg is raised and when the right leg is raised is within a predetermined value. Evaluating the symmetry of the robotic device with respect to
The method for evaluating a motion of a robot apparatus according to claim 2. The robot apparatus includes at least a trunk and left and right legs connected to the trunk,
In the step of evaluating the left-right asymmetry, depending on whether or not the difference in the amount of tilt in the pitch direction of the trunk that occurs when the left leg is raised and the right leg is raised is within a predetermined value, the pitch direction is determined. Evaluating the symmetry of the robotic device;
The method for evaluating a motion of a robot apparatus according to claim 2. The robot apparatus includes at least a waist and left and right legs connected to the waist.
In the step of forming the left-right asymmetric posture, a posture of protruding the waist portion forward is executed,
In the step of evaluating the left-right asymmetry, the left-right symmetry of the robot device with respect to the yaw axis direction is evaluated in a posture in which the waist is projected forward.
The method for evaluating a motion of a robot apparatus according to claim 2. In a motion evaluation method for a robot apparatus having a plurality of movable parts,
The robot apparatus includes a drive unit that drives a movable unit, a measurement unit that measures a state quantity of the movable part, and a drive control unit that controls the drive unit using the measured state quantity, and at least a part of the movable device is movable. The part touches the road surface,
Calculating a support polygon formed by a contact point between the movable part and the road surface;
Evaluating the motion characteristics of the robotic device based on at least one of the area of the support polygon or the rate of change of the area of the support polygon;
A motion evaluation method for a robot apparatus, comprising: In a motion evaluation method for a robot apparatus having a plurality of movable parts,
The robot apparatus is controlled using a drive unit that drives a movable part, a measurement unit that measures a state quantity of the movable part, a measurement part that measures a state quantity of the movable part, and a state quantity measured by the drive unit. A drive control unit,
Causing the movable part to execute a motion pattern for evaluation;
Recording a measurement result in the measurement unit during execution of the motion pattern for evaluation;
Analyzing the recorded measurement results and extracting features of the robot device;
A motion evaluation method for a robot apparatus, comprising: In the step of executing the movement pattern, a movement consisting of driving only a specific movable part is repeatedly executed.
The motion evaluation method for a robotic device according to claim 9. The robot apparatus includes a plurality of legs as the movable part,
In the step of executing the exercise pattern, one step of exercise with a specific leg is repeatedly executed.
The method for evaluating a motion of a robot apparatus according to claim 10. The robot apparatus includes a plurality of legs as the movable part,
In the step of executing the exercise pattern, the one-step exercise with the support legs fixed is repeatedly executed.
The method for evaluating a motion of a robot apparatus according to claim 10. The robot apparatus includes a plurality of legs as the movable part,
In the step of executing the movement pattern, a turning motion using legs is executed.
The motion evaluation method for a robotic device according to claim 9. The robot apparatus includes a plurality of legs having feet as the movable part,
In the step of extracting the characteristics of the robot apparatus, it is analyzed whether the entire foot of the leg is evenly and simultaneously landing or leaving the floor.
The motion evaluation method for a robotic device according to claim 9. In the step of analyzing the recorded measurement result and extracting the characteristics of the robot device, the difference between an ideal motion pattern instructed to the robot device and a motion pattern actually executed is analyzed, and the robot device is analyzed. Extract the features of
The motion evaluation method for a robotic device according to claim 9. Further comprising the step of displaying and outputting the analysis data of the recorded measurement result or the characteristics of the robot apparatus extracted by the analysis.
The motion evaluation method for a robotic device according to claim 9. The analysis data of the recorded measurement result, or further comprising the step of saving the characteristics of the robot apparatus extracted by analysis as a file,
The motion evaluation method for a robotic device according to claim 9.

Images