JP2003136440A - Brake control device of mobile robot - Google Patents

Abstract

PROBLEM TO BE SOLVED: To save the energy consumption in achieving the mission by optimizing the use of the friction force in a mobile robot having articulates capable of generating the friction force. SOLUTION: The mobile robot having at least one articulate which couples two links of a first link and a second link with each other in a relatively movable manner by a motor comprises a friction braking means capable of bridging the two links to generate the braking force in the relative motion, and a control means capable of selectively controlling the friction braking means in a braked condition or a released condition.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a joint device for a mobile robot, and more particularly, when the robot stops moving.
A robot that can save energy consumption when the leg joints are held in a desired posture even while walking, or when the motion direction of the joints matches the direction in which gravity acts, such as when descending stairs Related to the braking control device.

[0002]

2. Description of the Related Art A legged mobile robot having a plurality of joints has a characteristic similar to an animal in that it consumes a considerable amount of energy to support its own weight while stationary. This mobile robot does not consume energy while stationary like a mobile robot of the wheel type, but the above-mentioned characteristics are a major drawback, and the advantages of the leg type that can move even if there are steps or steps are However, it is one of the causes that prevent its practical application. In the first place, a stationary robot does not have to do any external work, so it is not necessary to consume these energies.
Some kind of improvement was required.

Consider a case where a legged mobile robot is walking on a flat ground.

When the leg that has completed the swing phase lands and shifts to the stance phase, from the viewpoint of energy consumption, a reasonable knee joint angle that supports weight is a straight line between the thigh link and the snee link. The angle is such that This is because as the degree of bending of the knee increases, the load torque of the knee joint due to the weight increases and the motor torque for controlling the joint also needs to be large. If the braking force that fixes the knee joint in a substantially straight line before and after landing can be applied to the knee joint, the load torque of the knee joint due to the landing impact force and the weight can be covered by this braking force. No motor torque is required, or very little if required.

FIG. 12 shows changes in the angle of the knee joint with respect to the time axis when a person walks naturally. As shown in FIG. 12, the knee joint begins to bend greatly immediately after the toes have left the bed, and if the legs are swung forward, the knees start to extend and land when the knees reach the end. Immediately after this landing, the knees are slightly bent to reduce the impact of the landing, but they are re-extended to support the weight and then moved to bed again.

The above is the data showing the relationship between the joint angle and the time during natural walking. On the other hand, when the army walks in steps, such as when marching, it can be seen that even when landing, the knee joints are fully extended and the weight is supported by mechanical locking means (stoppers) without relying on muscles.

Although there are legged robots that walk in this way, in a position control type robot, the method of holding a position by hitting a stopper is not compatible with position control, and therefore, a robot that hits a stopper is used. In the present situation, the method of changing the color is not yet adopted, and an alternative method has been required.

Next, let us consider a case where a legged mobile robot descends stairs.

The leg joint attached to the ground moves in the direction in which gravity acts while supporting the weight. At this time, the motor rotates in the direction opposite to the torque while generating torque against the weight. ing. As in the case of humans, conventional mobile robots consume a considerable amount of energy when climbing down stairs.

Further, legged robots usually have joints that control the direction change, but in normal movement, there are almost no scenes where these joints are used, and most of them are moving. During this time, a servo lock is applied to this joint to fix it. However, this servo book is useless because it consumes a lot of energy by itself.

[0011]

In a conventional robot, a structure in which a braking device is provided at a joint is already known. This braking device is provided for the purpose of, for example, braking an arm that holds a heavy object to prevent the heavy object from falling when an unexpected power failure occurs, thereby preventing personal injury. The principle is based on an electromagnetic clutch. Usually, a spring is provided in the direction in which the brake is applied, and the electromagnetic force opposes the spring force to open the clutch.

However, since such a braking device is made up of a member such as a coil or a steel plate that generates an electromagnetic force, it is heavy and the load on the joints is large.

By the way, a bipedal walking robot is a typical example of the mobile robot. Regarding the joint configuration of this walking robot, for example, there is a structure disclosed in Japanese Patent Publication No. 2592340. This joint uses an electric motor in combination with a speed reducer that can increase the speed of its output.

In such a configuration, when the robot maintains a stationary state, it is necessary to apply a so-called servo lock in which current is continuously supplied to the motor of the leg joint to support the weight of the robot. This is because the weight constantly works to bend the leg joints.

In view of such a problem, the applicant of the present invention has previously filed Japanese Patent Application No. 2001-195004 for a joint braking device in which the joint portion is mechanically locked at an arbitrary position when the robot is in a stationary posture. I have already applied for the issue.
According to the braking device of this prior application, the load torque of the joint portion due to the weight is borne by the braking device, so that the robot can maintain the posture at that time without the servo lock. That is, the robot can stand even if the supply of electric current to the motor is stopped or reduced.

As another example, there is an arm joint in which the elbow is bent to maintain a posture with a load. The same applies to the case where the robot has a leg joint that maintains a middle waist posture.

Here, the operation mode when the mobile robot is performing its mission will be examined as follows.

First, consider the case where the robot is walking on a flat ground.

As described above, in the gait (figure and shape when the robot is walking) in which the military personnel are marching, the knees remain fully extended before and after landing, and the impact of landing is specifically absorbed. I am walking without. In this case, the landing impact is certainly large, but walking is not difficult.
When the robot walks in this marching mode, when the knee just before landing is in a fully extended state, the knee joint is braked by the braking device so that the motor bears all the load moment of the joint due to the landing impact. Since there is no need to walk with less motor torque, it is possible to save energy consumption during walking.

When the free leg has not landed yet, the difference between the commanded angle and the actual angle of the joint, which will be described in detail later, is small. Therefore, if the difference can be maintained even when a landing impact is applied, the time of landing The motor current value of is also small.

Next, consider the knee joint when the robot is descending stairs.

Considering, for example, a knee joint of a leg in the stance phase, the joint is largely bent, and the distance from the vertical line lowered from the center of weight to the center of rotation of the knee joint is the largest. Since the value obtained by multiplying this distance by the weight is the moment to further bend the joint, the knee joint needs to gradually bend the knee while generating a torque capable of counteracting this large moment.

Therefore, in this state, the largest current flows through the motor of the knee joint, and the energy consumption is maximized. If, at this time, the knee joint can be braked by the braking device, a large part of the joint torque due to the weight can be borne by the frictional force.
The torque that the motor of the joint part bears, in other words, the motor supply current can be reduced accordingly, and energy consumption can be saved.

In this case, since the moment created by the weight increases in accordance with the bending angle of the knee joint, it is desirable that the frictional force should be variable according to the joint angle.

The same applies to the ankle joint connecting the shin link and the foot link.

In the above, the walking robot goes down the stairs, but the same can be said when trying to sit down.

A device for variably generating such a frictional force is also disclosed in detail in Japanese Patent Application No. 2001-195004 of the prior application.

Further, conventionally, even in the joint portion of a legged robot capable of changing direction, the joint is servo-locked when making a straight advance, whether on a level ground or when climbing stairs. Even for such a robot, if the robot can use a frictional force instead of the servo lock when walking without changing the direction, the joint can be fixed without energy consumption, and the direction needs to be changed. At this point, if the frictional force is released and joint movement is controlled, the purpose of direction change should be achieved.

The present invention has been made by paying attention to the above points, and when a frictional braking means capable of generating a braking force is provided at the joint portion of the mobile robot and the frictional braking means performs the mission. An object of the present invention is to provide a braking control device for a mobile robot, which can reduce energy consumption during the mission of the mobile robot by operating it appropriately.

Here, a brief description will be given of how the energy generated in the joints of the walking robot is used.

In the current joint control technique, when the difference between the target joint angle and the current joint angle is calculated, the difference is multiplied by a constant proportional constant K to obtain an operation amount, and the proportional constant K is applied with a maximum load. Even so, it was set to a large size to ensure good followability.

Normally, the target joint angle and the current joint angle do not coincide with each other, and there is a slight difference between the two even when the joint is stationary. A current corresponding to this difference is supplied to the motor. A moment is generated in the joint to counter the falling moment due to gravity. If the external load moment increases and the difference between the target joint angle and the current joint angle increases, the amount of operation also increases, and by increasing the motor current according to the amount of operation, it is possible to respond to the increased load moment. Generating motor torque. That is, the external load moment and the motor torque generated by the motor are physically equal at all times, and the two are balanced. Therefore, if the proportional constant K is constant, the current consumption value of the motor is proportional to the difference between the target joint angle and the actual joint angle.

Here, if a frictional force exists in the joint, the frictional force also generates a moment in the joint. Therefore, the balance equation of these three moments changes depending on which direction the frictional force acts. .

If the moment due to the frictional force is in the opposite direction to the load moment, the motor must generate a torque equal to the sum of the load moment and the moment of the frictional force. In other words, the energy consumed by the motor is smaller when there is no frictional force in this case.

On the other hand, when the moment due to the frictional force opposes the load moment, the sum of the moment of the frictional force and the motor torque should be balanced with the load moment, so that the motor torque changes from the load moment to the moment of the frictional force. Energy savings can be achieved by subtracting the small amount. Especially when the robot is stationary, is the moment of friction equal to the load moment?
Alternatively, if large, the robot can remain stationary without supplying current to the motor. In other words, the same is true when the robot keeps its standing or bent posture.

Also, the smaller the load moment,
Since the difference between the target joint angle and the actual joint angle is small, during the swing phase during walking, and when acceleration / deceleration has subsided to some extent,
The difference between the target joint angle and the actual joint angle is the minimum. If braking is applied in this state, the increase in the difference can be suppressed even if the load moment increases due to landing next.

Further, when the user is descending stairs, the actual joint angle is ahead, and the control is executed so that the target joint angle follows this. In this situation,
When the braking force is generated in the joint, the amount of advance of the actual joint angle decreases, so the difference between the actual joint angle and the target joint angle becomes small, and energy consumption can be saved.

In the above, the walking robot goes down the stairs, but the same applies when trying to sit down.

Therefore, according to the present invention, in a mobile robot having joints capable of generating a frictional force and braking, the utilization of the frictional force is optimized as described above, so that the robot consumes energy during the mission. Is to save money.

[0040]

In order to achieve the above object, the present invention constitutes a braking control device for a mobile robot by the following means.

The invention according to claim 1 is a mobile robot having at least one joint part in which two links, a first link and a second link, are coupled by a motor so that they can move relative to each other. And a friction braking means capable of generating a braking force in the relative motion by bridging and a control means capable of selectively controlling the friction braking means to a braking state or a releasing state.

According to a second aspect of the present invention, in the braking control device for a mobile robot according to the first aspect of the invention, the control means performs the friction braking when relative movement is not scheduled between two links. It is configured to control the means to a braking state.

The invention according to claim 3 is the braking control device for a mobile robot according to claim 1, wherein the control means is arranged such that when the relative movement direction of the two links is opposite to the movement direction due to the motor torque. And is configured to control the friction braking means to a braking state.

The invention according to claim 4 is the braking control device for a mobile robot according to claim 3, wherein the control means is configured to variably control the braking force of the friction braking means. .

According to a fifth aspect of the present invention, in a mobile robot having at least one joint part in which two links, a first link and a second link, are connected by a motor so that they can move relative to each other, the two parts between the two links are And a friction braking means for providing a controllable relative movement restraining force to the relative movement of the link, and the friction braking means is controlled at a time necessary to hold the joint of the mobile robot at a desired joint angle. And a control unit that operates to generate the relative motion restriction force.

According to a sixth aspect of the invention, there is provided a mobile robot having at least two legs that are moved by being alternately moved by a motor and a joint that is used to change the direction,
Friction braking means that is provided at a portion corresponding to the joint and that generates a frictional force that suppresses the movement of the joint, and that applies braking to the joint for at least one step when the robot is in a straight advancement state. And a control means for controlling the friction braking means.

The invention corresponding to claim 7 is the braking control device for a mobile robot according to any one of claims 2, 5, and 6, wherein the friction braking means is in a braking state. An energy supply control means for reducing or stopping the energy supplied to the motor is provided.

The invention corresponding to claim 8 is the braking control device for a mobile robot according to claim 7, wherein the energy supply control means applies a braking control to the motor after the braking means is controlled by the control means. It is configured to reduce or stop the energy supply.

The invention corresponding to claim 9 is a walking robot having at least one joint part in which two links of a first link and a second link are coupled by a motor so that they can move relative to each other. And a friction braking means which is provided so as to be capable of cross-linking and provides a relative movement restraining force that can control the relative movement of the link, and one cycle of walking of the walking robot (until one leg lands after landing, and then one floor lands again) During the period), the control means controls the friction braking means so as to generate the relative motion restriction force at least once.

The invention according to claim 10 is the braking control device for a walking robot according to the invention according to claim 9,
The walking robot has at least one leg composed of three joints connected in series in the order of a hip joint, a knee joint, and an ankle joint, and connects a foot link to the lower end of the ankle joint. The joint portion is at least one of a knee joint and an ankle joint of the three joints.

The invention corresponding to claim 11 is the braking control device for a walking robot according to the invention according to claim 9,
The walking robot includes at least two legs having a structure including a hip joint portion, a knee joint portion, and an ankle joint portion, and a thigh link and a shin link connecting these joints, like a human.
Moreover, these legs are alternately moved to walk, and the control means is provided in at least one joint part of the three joint parts of the leg at the time of landing when the one leg shifts from the swing phase to the stance phase. The friction braking means is controlled so as to generate the relative motion restriction force.

The invention according to claim 12 is the braking control device for a walking robot according to the invention according to claim 9,
The walking robot includes at least two legs having a structure including a hip joint portion, a knee joint portion, and an ankle joint portion, and a thigh link and a shin link connecting these joints, like a human.
Moreover, these legs are alternately moved to walk, and the control means is the friction braking means provided at at least one joint part of the three joint parts of the leg in the stance phase when the robot descends the stairs. Is controlled so as to generate the relative motion restriction force.

[0053]

BEST MODE FOR CARRYING OUT THE INVENTION Embodiments of the present invention will be described below with reference to the drawings.

FIG. 1 is a perspective view showing the skeleton of a bipedal walking robot to which the present invention is applied. The walking robot is on the left,
Since the right leg is in a mirror image relationship and has a symmetrical shape, the symbols of the respective parts are shown with suffixes L and R in the figure.

In FIG. 1, the walking robot has six joint axes on its left and right legs. These six joint axes are
From top to bottom, a joint shaft 10 for rotating a leg, a joint shaft 12 in the pitch direction of the hip joint, a joint shaft 14 in the same roll direction, a joint shaft 16 in the pitch direction of the knee joint, a joint shaft 18 in the pitch direction of the ankle joint, It is the joint axis 20 in the same roll direction, and the lower part of the joint axis 20 has a 6-axis force sensor 22 on it.
24 are provided.

A waist plate link 26 corresponding to the pelvis of the human body is provided at the uppermost portion. The waist plate link 26 is equipped with a power source and an amplifier (amplifier) necessary for joint control, and an inclinometer 28 for detecting the inclination of the robot with respect to the walking space and outputting it as an electric signal by means not shown.

As shown, the hip joint has three joint axes 10,
12 and 14, the hip joint and the knee joint are connected by a thigh link 30, and the ankle joint is also composed of two joint shafts 18 and 20, as shown in the drawing, and the ankle joint and the knee joint are connected by a tibial link 32. Are combined.

Between the ankle joint and the foot portion 24, a vertical component Fz, a traveling component Fx, and a lateral component F of the floor reaction force F.
A 6-axis sensor 22 is provided which can separately detect y, the vertical direction component Nz, the traveling direction component Nx, and the lateral direction component Ny of the moment N generated by the floor reaction force.

This 6-axis force sensor is publicly known, and it is also publicly known that it is mounted between the foot portion and the ankle joint, and therefore its explanation is omitted here.

FIG. 2 shows the prior application No. 2001-195004.
FIG. 3 is a diagram schematically showing a braking device provided on the joint shaft 16 in the pitch direction of the knee joint as one of the joint braking devices described in No. 6, and the operating principle thereof will be briefly described here.

In FIG. 2, when an appropriate level of air pressure is introduced from the air compression source 226 into the cylinder 207 connected to the side of the thigh link 30, it is slidably inserted into the cylinder 207. Piston 22
2 is pushed (to the right in the drawing) toward the rotary disk 224 connected to the side of the shin link 32. Then, a frictional force is generated between the thigh link 30 and the shin link 32, a moment for suppressing relative movement between the two is generated, and braking is applied.

This braking device is provided for each of the 12 joints existing in the robot shown in FIG.

FIG. 3 is a pneumatic pressure control system diagram for sending compressed air to a braking device provided in each joint.

In FIG. 3, the compressor 200 is driven by a DC motor 201 and stores energy in the form of pressure in a surge tank 204 via a check valve 202. These motor, compressor, surge tank, etc. are fixed to the waist plate link of the robot.

The cylinder and piston of the braking device described above are set in each joint. That is, since the right leg has six joints, six sets of braking devices are provided. The three braking devices of the hip joint are shown surrounded by a dotted line 210 in the drawing.

The broken line 210 in the drawing includes three braking devices 207-1R, 207-2R, 207-3R, and is surrounded by another broken line 212 in the drawing.
R and 207-6R are for the ankle joint, and the braking device 207-4R on the lower side of the drawing is for the knee joint.

Compressed air is introduced from the surge tank 204 into the air pressure chamber of the above-mentioned hip joint and ankle joint braking device via a common solenoid valve 205-R, and from the surge tank 204 to the air pressure chamber of the knee joint portion, servo control is performed. The compressed air is introduced through a solenoid valve 206-R of the mold. When any of the above solenoid valves is demagnetized, the pneumatic chamber of the braking device is opened to the atmosphere, and when excited, it is connected to the surge tank 204.

Since such a solenoid valve is publicly known and used for public use, its explanation is omitted here.

The above is the braking device provided at each joint on the right leg side, but the same pneumatic control system as the above is formed for the braking device provided at each joint on the left leg side. .

FIG. 4 is a block diagram showing a first embodiment of a braking control device for a mobile robot according to the present invention.

In FIG. 4, when receiving a command signal from the command device 304, a central processing circuit (CPU) 302 for executing processing necessary for controlling the braking device stores gait data suitable for walking at that time. Read from 306. The memory 306 stores gait data (angle command value) _D Θ
(t) and the braking control signal _P E (t) are stored as a function command value _D J (t) which is a vector of time series data.
Here, the gait is the figure / shape of the robot walking, which can be mathematically expressed as time series data Θ (t) of the angle of each joint. Therefore, since the leg structure of the robot has 12 joints, the gait is Θ (t) = [θ _1L (t), θ _2L (t), θ _3L (t), θ _4L (t), θ
_5L (t), θ _6L (t), θ _1R (t), θ _2R (t), θ _3R (t), θ _4R (t), θ _5R
(t), θ _6R (t)] ^T vector. Each element of this vector is a function of time.

The symbol T shown on the right side of the vector is a permutation symbol, and when this symbol is added to a horizontal vector as described above, it is converted into a vertical vector. Conversely, when this symbol is added to a vertical vector, the horizontal vector is added. Is converted to.

Further, the real angle information _R Θ (t) is fed back to the CPU 302 from the encoder 308 provided coaxially with the drive motor of each joint. However, in this embodiment, since the encoder 308 uses an incremental type that simply outputs a pulse signal in response to rotation, a counter 310 that counts the number of pulses and converts it into angle information.
Are provided between the encoder 308 and the CPU 302.

CPU 302 reads from memory 306
Angle command value_DΘ (t) and real angle information _RThe difference from Θ (t)
Calculate and multiply the difference by the proportional constant K to obtain the speed again.
Command value_DΩ (t) to the first amplifier 312 of the joint
Output. This first amplifier 312 has a counter 310
Signal input from_RReal angular velocity information by differentiating Θ (t)
_RMake Ω (t). And the speed command value_DΩ (t) and real angle
Speed information_RCalculate the difference of Ω (t) and set proportional to the difference
The motor 31 is set as the current command value I (t) by multiplying the number K by
Output to 4. Then, the motor 314 generates the torque T (t).
To drive the joint 316, resulting in encoder 3
A pulse change occurs at 08.

As the first amplifier 312, a speed control type is used, and this manipulated variable is the speed command value of the motor 314, and it is preferable to form a speed feedback loop inside the control loop. Achieves excellent followability.

Now, the joint command value read by the CPU 302
Command information regarding the solenoid valve 322 _DContains E (t)
ing. CPU 302 uses these solenoid valve command values_DE
(t) is input to the second amplifier 320 to excite the solenoid valve 322.
Magnetize or degauss. The solenoid valve of the knee joint is applied to the applied voltage
Servo type that can change its air pressure level according to
Is used, so the solenoid valve command given to the solenoid valve of the knee joint
value_DE (t) is a simple on / off signal (signal of 1 or 0)
No.), but a grayscale signal. That is,
In addition to 0 and 1, at least an intermediate signal such as 0.5
It shall be included. On the other hand, electromagnetic waves for other joints
Since the valve is on / off control,_DE (t) is 1 or 0
It shall be an on-off signal.

Although not used in this embodiment, the tilt angles α, β, and γ of the robot, which are the output signals of the inclinometer 28 described above, are input to the CPU 302, and the 6-axis force sensor attached to the left and right ankles is input. 22L, the force from the 22R information _F R, _{F L} and moment information _N R, also _{N L} is input.

Here, an example of the form of the joint command value _D J (t) stored in the memory 306 will be described with reference to FIG. The joint command value _D J (t) is a command value of the joint angle _D
Θ (t) and the solenoid valve command value _D E (t) are expressed in the form of a vertical vector. Further, the joint angle command value _D Θ (t) is a vertical vector in which the angle command values θ _I (t) of each joint are arranged vertically, and the solenoid valve command value _D E (t) is also given to each solenoid valve. The power level e _I (t) is a vertical vector that is vertically aligned.

Although each element of the vector is a function of time, it must be expressed as a discrete number because the computer performs digital signal processing. FIG. 6 shows the joint command value _D J (t) expressed as a discrete number.

That is, the information of _D J (t) when t = 1 is
It is given by _D J (1) and the information when t = 2 is _D J
It is given in (2). A series of information in which these are arranged in order eventually takes the form of a matrix. For example, in the case of walking on a flat ground, first, gait data for acceleration from the standing state is required. Therefore, the contents of _{D J} (1), all the elements of the joint angle command value _D theta (t) (angle)
Is 0 (in the standing posture), and the solenoid valve command value _D E
The element (voltage level) of (t) is also zero. A vector with zero elements can be written mathematically as a vector 0.
Therefore, _D J (1) = [0 ^T , 0 ^T ] ^T.

Next, gait data for steady walking is required. Finally, gait data until decelerating and stopping (finishing in a standing posture) is required. Therefore, the final joint command value _D J (n) is also _D J (n) = [0 ^T ,
0 ^T ] ^T. By the way, in the normal walking, the same gait data is repeatedly used for a predetermined number of steps (to be exact, the left and right data are interchanged for each step).

The present invention can be applied to walking when going up and down stairs as well as level ground walking. The gait data is naturally different depending on where you walk, and joint command values are used to distinguish these.
_A suffix is added after DJ (t) to distinguish them. That is,
_D J (t) _W is the gait when walking on a level ground, _D J (t)
_SU is the gait when ascending stairs and _{D J} (t) _SD is assumed to show a gait when descending the stairs respectively.

Next, the operation of the braking control device for the robot configured as described above will be described with reference to the control algorithms shown in FIGS. 7 and 8.

The CPU 302 is activated at regular intervals to perform processing, but here it is assumed to be activated every 2 msec (0.002 seconds).

Further, when the robot is first made to stand up and to be activated, although not shown in the figure, all the Flag bits to be described later are set to 0 by initialization.

In FIG. 7, since the command signal for climbing the stairs is not yet inputted at the beginning, step S410,
The determination in step S421 is No, and since the command signal for descending the stairs has not been input, step S4 is performed.
The determination of 40 is also No, and since the person is not walking further, the determination of Step S446 is also No. Then, since the flag for identifying when standing is not standing, step 4
The determination of 50 is also No.

However, since the robot is actually standing up, the bit of Flag-stand is set to 1 in step S451. Succeedingly, in a step S452, a timer for braking is set to C _BR = 200. With this timer setting, if 1 is subtracted from the timer at each activation time, the timer becomes C _BR = 0 at the 200th time. That is, the 200th time means that it takes 400 msec, that is, 0.4 seconds since it is activated every 2 msec.

Next, proceeding to step S453, the flag-timer bit, which means that the timer has been activated, becomes 1. Next, in step S454, since the body is in the standing posture, the value of _D Θ (t) is set to 0, and the proportional constant K is set to a value K _W suitable for level ground walking in step S455.

Next, in step S456, a joint angle follow-up control flow is executed. That is, the actual joint angle _R Θ
(t) is read, and in step S457, the joint angular velocity command value _D Ω (t) is calculated based on the determined calculation formula, and in step S458, the joint angular velocity command value _D Ω is calculated.
(T) is input to the first amplifier 312. In this case, since the number of joints is 12, naturally the joint angular velocity command value _D Ω
The number of elements ω _j (t) in (t) is 12 and the number of output amplifiers is 12. Finally, in step S459, the time index is advanced by 1, and the process returns to the basic flow of the CPU.

When it is started next time, step S45
Since the determination at 0 is Yes, the process proceeds to step S461,
It is determined whether Flag-Freeze is standing. Since it is the first time, of course, Flag-Freeze is 0, and the determination is No, so the flow advances to step S462 to perform the work of subtracting 1 from the timer. Then, in step S463, it is determined whether C _BR has become smaller than 100. Since C _BR is larger than 100 at the beginning, the determination is No, and the process proceeds to step S464 to determine whether C _BR has become 0 or not. Since this result is of course also No, step S45
Returning to step 5, the processing similar to the above is executed from now on.

When this cycle is repeated, the determination result in step S463 becomes Yes, and in this case, the process proceeds to step S465 and the value of _D E (t) is fixed to _D E ₀ . It is assumed that the value of _D E ₀ is such a signal that the excitation signal is 1 for the braking device for the hip joint and the ankle joint, and that the knee joint has a relatively strong braking force.
When this excitation signal is received, all joints are fixed, and the robot is at a level where it can maintain its standing posture even if power is not supplied.

Next, in step S466, the newly set solenoid valve control signal _D E is set in the second amplifier 320.
give (t). Then, it returns to step S464 and it is verified whether the timer is 0. Since the determination is No at first in the beginning, the process returns to step S455 to execute the follow-up control flow. However, if the activation is repeated 100 times thereafter, the determination in step S464 becomes Yes,
It proceeds to step S467. In this step S467,
Set the proportionality constant to a very small K _FZ . This K _FZ
The value of is ideally 0.

By making the proportional constant small in this way,
Although there is a difference between the joint angle command value _D Θ (t) and the actual joint angle _R Θ (t), the joint angular velocity command value _D Ω (t) given to the motor becomes an extremely small amount, and the energy saving system is ready. Is made. Now that the timer has finished its role, go to step S468 to set Flag-timer to 0,
Fl indicating that the whole body is braking in step S469
The bit of ag-Freeze is set to 1 and the joint follow-up control flow is executed.

In the next startup cycle, step S4
Since the determination in 61 is yes, the flow proceeds to step S456, the previously determined solenoid valve control signal _D E (t) is output to the second amplifier 320, and the flow proceeds to the follow-up control flow. From now on, the robot is fixed in the standing state until the next new command signal is input.

In such a state, it is assumed that a command signal for climbing the stairs is input to the CPU. Here, in order to simplify the explanation, it is assumed that the command signal is issued only once.

In step S410, it is determined whether or not a command signal for climbing the stairs has been input. The determination at this time is Yes.
Since it is s, the process proceeds to (a) in FIG. 8 and the value of the proportional constant K is set to a value K _SU suitable for climbing stairs in step S411. Then, in Step S412, Flag-Freeze
The bit of is set to 0 and the bit of Flag-SU is set to 1. When the Flag-SU bit is 1, it indicates that the stairs are being climbed.

Next, in step S413, the time index is set to 1. Succeedingly, in a step S414, a command value _D J (t) used when going up the stairs as a function command value
Select _SU , read it, and proceed to step S415.
In step S415, the joint angle command value _D Θ (t) _SU is separated from the joint command value _D J (t) _SU and replaced with the angle command value _D Θ (t) used in the subsequent equation. At the same time, the solenoid valve command value _D E in _D J (t) _SU
(t) The _SU is separated and set again as the solenoid valve command value _D E (t).

When the joint is braked while climbing the stairs, energy consumption basically increases, so that in this case, _D E (t) _SU = 0 is set in principle.

Here, the value of _D Θ (t) is set to 0 (the standing posture) when the first t is small (for example, the first 0.15 seconds).
Leave. [Theta] This is because there is a slight delay in the operation of the braking device, and by doing so, the joint that has been braked in the upright posture is completely released. After that, the process proceeds to step S416, and the electromagnetic valve command value _D E (t) is given to the second amplifier 322 to demagnetize the electromagnetic valve. In this case, since there are four amplifiers as the output destinations, it goes without saying that they are output to each of them.

Next, in step S417, it is determined whether the gait data has ended. For example, this determination method is t
By checking whether or not there is data of = t + 1, it is possible to know whether or not the current gait data is the final data. Since it is No at first in the beginning, step S4 from FIG.
In step 56, the joint angle follow-up control flow is executed. That is, the actual joint angle _R Θ (t) is read and step S45
In 7, the command value of the joint angular velocity is calculated based on the determined calculation formula.
_D Ω (t) is calculated, and this joint angular velocity command value _D Ω (t) is given to the first amplifier 312 in step S458. Finally, in step S459, the time index is advanced by 1, and C
Return to the basic flow of PU.

At the next starting time, step S4 in FIG.
In 10, the command signal has already disappeared, so it is determined as No and the process proceeds to step S421. Since the bit of flag-SU is 1 in this step S421, the process proceeds to (b) of FIG. 8 and, in step S414, when t = 2,
_Read DJ (t) _SU . From now on, (g) of FIG. 7
After going through the above-described processing flow, the process proceeds to step S456 again, and the follow-up control flow is executed. When such a process is repeated, it is determined that _D Θ (t) has reached the stationary standing posture and the mission to climb the stairs is completed at some point in step S417 of FIG. 8, and the process proceeds to step S418.

In step S418, all Flag bits are set to 0, and in step S419, a proportional constant K _w suitable for walking on a level ground is selected, and this is set again as a proportional constant.
After that, execute the follow-up control flow from (g) of FIG.
Return to the basic flow of the CPU.

Next, the case where a command to go down the stairs is input while standing up will be described. Also at this time, the command is only 1
It is supposed to be input to the CPU only once.

The determination result in step S430 in FIG. 7 is Y.
Since it becomes es, the process proceeds from step (c) of FIG. 8 to step S431, a proportional constant K _SD suitable for descending the stairs is selected as a new proportional constant K, and the bit of Flag-Freeze is set to 0 in step S432. Set the Flag-SD bit to 1 to indicate that you are descending the stairs. Step S4
33, the time index is set to t = 1, and step S434
At, the joint command value _D J (t) _SD suitable for getting down the stairs is read from the memory, and the process proceeds to step S435. In step _{S435, D} theta from the _D J (t) _SD
(t) _SD is read, and this is determined again as _D Θ (t).
At the same time, the solenoid valve command value _D E (t) in _D J (t) _{SD
The SD} is read, and this is read anew and the solenoid valve command value _D E (t)
I will keep it. Also in this case, as in the above example, _D E (t) = 0 is set at the beginning when t is small, and the gait for continuing the static posture is set as _D Θ (t) _SD . After that, the process proceeds to step S416, and the second valve 320 is controlled by the solenoid valve command value _D.
Since the flow after outputting E (t) is the same as the flow for climbing the previous stairs, its description is omitted here.

At the time of the next startup cycle, No is determined in step S430 of FIG. 7, but step S436.
Since it is determined to be Yes, the process proceeds to (d) in FIG. 8 and the process according to the same flow as above can be executed. When the stairs are finished, the determination result of step S417 is Yes.
Therefore, the flow is switched to the stand-up control flow via step S418.

Here, the important thing is that the solenoid valve command value_D
E (t)_SDPattern. As shown in Figure 5_D
There are four elements of E (t). Of this,_De_2L(t) and_D
e_2R(t) is the command value that determines the braking of the left and right knee joints.
It on the other hand,_DThe element of Θ (t) that corresponds to the knee joint
_Dθ_4L(t) and_Dθ_4R(t). Solenoid valve command value_De
_2L(t) and_De_2RThe following relation is established with (t).
That is, for the landing leg, the solenoid valve command value
The value changes according to the joint angle command value. For example on the left
If the leg is the one supporting the weight,_De_2L
(t) is in principle_Dθ_4LIt increases as (t) increases.
To set. By doing this, the knee joint will break.
The knee joint model due to gravity increases as it bends.
You can also increase the friction moment against the
The moment the motor bears (motor torque) is small
Energy saving effect. The degree of association is the book
As in the embodiment, it is held as a map in advance.
, You can change the map values freely, so
Optimized easily.

Of course, since the right foot bears the weight immediately after landing on the right foot, the load moment of the knee joint of the left leg is drastically reduced. Therefore, at the same time as landing, the value of _D e _2L (t) is set to 0 or a small value It is necessary to change to (In reality, there is a time delay in pneumatic cylinders, so please allow for that delay.
It is necessary to decide the map value of _D e _2L (t)),
Information on when to land is also required to be included in the gait data in advance.

Next, the case where a command to walk on a level ground is issued, not the case of going down the stairs, will be described.

At this time, in step S440 of FIG.
Since it is determined to be Yes, proceed to (e) of FIG.
Proportional constant optimized when walking on flat ground by S441
Number K _WIs selected again as the proportional constant K. And from now on
Since the mode of ground walking is entered, Flag-F is entered in step S442.
Set the reeze bit to 0 and the Flag-walk bit.
Set to 1.

Next, in step S443, the time index is t = 1.
To the joint command value for level ground walking in step S444.
data_DJ (t)_WIs read from memory. this_DJ
(t) _WThe data of is as long as t is small_DE (t) = 0
With_DΘ (t) is also a vector in the static standing posture
Keep it.

In step S445, the joint angle command value _D Θ (t) _W is extracted from this data and is set again as the target joint angle _D Θ (t). At the same time, _D E (t) _W is extracted and this is re-established. Set the solenoid valve command value _D E (t).
In step S416, the solenoid valve command value _D E (t) is output to the second amplifier 320, and in step S417, it is verified whether the gait has ended. Since it is the first time, the result of the determination is No. After that, from step (g) of FIG. 7, the process proceeds to step S458 along the flow of the joint angle tracking control, and at step S459, the time index is advanced by 1 and the original basic Return to flow.

When the next activation time comes, the result of the determination in step S440 of FIG. 7 is No, but the determination in the next step S446 is Yes. Returning to the flow of the mode, the same as the above is repeated. If the gait data becomes the last data during such repetition, the determination result of step S417 becomes Yes, the process proceeds to step S418, all Flags are set to 0, and the last joint angle follow-up control is performed. After going, enter the standing posture mode.

Here, the important thing is that the solenoid valve command value _D
It is a pattern of E (t) _W. In creating a flat ground gait, after the legs have left the bed, bend the knee joint greatly and swing forward,
Extend your knee joint. If the gait is designed so that when the knee is fully extended, the landing is completed without bending the knee and the weight-bearing phase is advanced, the bending moment applied to the knee can be made relatively small. Is larger, it acts as a compressive force on the knee, but does not increase the bending moment). Focusing on this fact, after the knee is fully stretched, the braking device for the knee joint is activated to fix the knee during the swing leg.

In order to realize this idea, the solenoid valve command value
The pattern of the element related to the knee joint braking in _D E (t) _W is set to 0 when leaving the bed, and is designed to be a relatively high constant value when the knee joint angle command value becomes 0. It is designed to return to 0 immediately before leaving the bed. By creating such an angle command value _D J (t) _W , most of the knee joint moment when supporting the weight is carried by the frictional force, and the motor torque of the knee joint is suppressed to substantially zero. Energy saving can be achieved.

Next, a second embodiment of the braking control device for a mobile robot according to the present invention will be described with reference to the drawings.

In the second embodiment, the bipedal walking robot and the knee joint braking device are the same as those in FIGS. 1 and 2, and therefore their explanations are omitted here.

FIG. 9 shows an air pressure control circuit corresponding to FIG. 3 used in this embodiment. In FIG. 9, a friction braking device 50 provided at the joint that controls the direction change
On-off control type solenoid valves 505-L and 505-R are used for 7-1L and 507-1R, respectively, so that braking force can be obtained independently. The friction braking devices for other joints are independent on the left and right, but a pair of solenoid valves 506-
It is configured to be controllable by L and 506-R. That is, the braking device surrounded by the dotted line 510 is for the hip joint, and the braking device surrounded by the dotted line 512 is for the ankle joint. As for the right leg, by exciting the solenoid valve 505-R, the joint for direction change is fixed by frictional force.

The control block diagram of the pneumatic control circuit having such a configuration is almost the same as that shown in FIG.
Is all on / off type, and the way of connecting them is only changed as shown in FIG. The vector display of the joint command values is also the same as that in the first embodiment (FIG. 5), and the matrix display of the joint command values is also the same as that in the first embodiment.

The proportional constant K used in the control algorithm will be described with reference to FIG. The constant of proportionality to allow setting a different value for each joint in FIG. 10 K _W
It is defined in the form of a matrix. In the first 6 rows and 6 columns, the proportional constants suitable for the 6 joints of the left leg are described as a diagonal matrix. Therefore, the diagonal matrix has 12 rows and 12 columns as a whole. In the prior art, 12 joints are controlled according to this K _W.

On the lower side of FIG. 10, another proportional constant matrix K _{W-STR T} suitable for a straight ahead row is shown.
The difference between K _W and K _W-STRT is, only the value in the first row the first row and the seventh row seventh column is different, except it is the same (* mark given by Aru values are different) . The value of K _FZ is sufficiently small or 0.

Next, the operation of the braking control device for the robot configured as described above will be described according to the control algorithm shown in FIG. It should be noted that, in the drawings, for the sake of brevity, only the portions that are the gist of the invention are shown.

In FIG. 11, when a timer interrupt occurs in step S601, it is determined in step S610 whether a flatland departure signal has been input. If the determination result is Yes, the process proceeds to step S611, and Flag-Fre
Set eze bit to 0 and Flag-Walk bit to 1
To Then, in step S612, the time index t is set to 1
Set to.

Next, at step S613, the joint command value _D J
(t) _W is read, and in step S614, the joint angle command value is separated from the joint command values, and this is again set as the target value _D Θ (t) of the joint angle. Next, proceeding to step S615, the joint angle command value in the joint command values is separated, and this is again set as the electromagnetic valve command value _D E (t).

In step S616, the solenoid valve command value _D
E (t) is output to the second amplifier 320, and the process proceeds to step S617 to determine whether the time index t is between i and k in FIG. Since the result of this determination is, of course, the first, it is No and the process proceeds to step S619. In this step S619, K _W in FIG. 10 is selected as the proportional constant, and it is determined in step S620 whether or not the gait data has ended.
After that, the joint follow-up control flow is executed as in the first embodiment. That is, the actual joint angle is read in step S622, and the target value of the joint angle is read in step S623.
The difference between _D Θ (t) and the actual joint angle _R Θ (t) is calculated, this is multiplied by the matrix K, and the result is set as the target angular velocity command value _D Ω (t) of each joint.

Since this calculation is a multiplication of a vector and a matrix, the product _D Ω (t) can be calculated by a mathematical formula,
It becomes the following vector.

[0126]

[Equation 1]

When the next activation time comes, this time step S
Since Flag-walk is 1 in 630, the determination result is Yes, and the same is repeated for the data of t = 2. If this is repeated several times, the determination result of step S617 becomes Yes. During this period, the joint for direction change is completely locked, so that the proportional constant of the joint for direction change becomes small in step S618.
_W-STRT is again selected as the proportional constant K.

The same operation is repeated thereafter, but if it is repeated several times, the result of the determination in step S617 is No. Since the walking enters the deceleration mode and eventually stops, the proportional constant K is returned to the original K _W again in step S619.
After that, the same processing is repeated after step S622. Finally, when the gait data is over, the determination result in step S620 is Yes, so the process proceeds to step S621, the flag-walk is set to 0, and the final follow-up control is performed.
Return to the basic flow of U.

If the walking command signal is not input when the next starting time comes, the process proceeds to step S640.
Proceeding to the stationary mode shown in the first embodiment (in this embodiment, a flow at rest is indicated), each joint is fixed by operating the friction braking device.

As described above, according to the second embodiment, since the joint for direction change can be usually fixed by the friction braking device, the proportional constant of the joint is decreased or changed to 0 to reduce the energy consumption. It is possible to avoid wasteful consumption. Further, it can be similarly applied to the case of moving up and down the stairs.

The present invention is not limited to the embodiments described above and shown in the drawings, but can be carried out in the same manner as above with the following configurations.

(1) The illustrated solenoid valve is shown to generate a braking force when excited, but the connection structure may be reversed so that the braking force is generated when demagnetized. Further, the case where the servo-type solenoid valve that can continuously change the braking force of the knee is used has been described, but, for example, a solenoid valve that can be changed stepwise may be used. Further, although the case where the device for generating the braking force by air pressure is used is described, a braking device by driving a wire or the like may be used. In short, regardless of the type of the braking device, any controllable device can be applied.

(2) In each of the above-mentioned embodiments, in order to create gait data, the data created in advance is stored in the memory, and this is read from the memory for each walking robot to perform walking. However, there is also a method of generating a gait in real time according to the situation as a method of causing a robot to walk, and even with such a method, the present invention can be applied and implemented as described above. That is, the computer of the robot can easily calculate the gravitational moment applied to each joint from the gait at that time, and can easily determine whether braking at that time leads to energy saving. (Even if it is difficult to make a judgment, it can be easily judged by experiment.) Therefore, considering the moment due to gravity applied to a specific joint (eg, knee joint) during walking, it is energy-saving to have a large moment and braking. If the joint braking is performed in such a case, the initial purpose can be achieved.

(3) In the above-described embodiment, the joints to be braked when going up and down stairs or walking on level ground are explained as knee joints and joints for changing direction, but the present invention is not limited to this. For example, when the ankle joint is driven in the front-rear direction (pitch direction joint), the joint may be fixed by friction when leaving the bed to support the weight in the next stance phase.

Further, in the above-described embodiment, the stationary posture is explained as the standing posture, but the present invention is not limited to this, and the posture may be one in which both legs are opened left and right to increase the stability. Monitor instruments in low places,
Alternatively, even when the reading work is performed in the middle waist posture, the braking control can be performed in the stationary posture.

(4) In the second embodiment, the joint for direction change is fixed by frictional force, and at the same time, the value of the current supplied to the motor is reduced, but the present invention is not limited to this. If not aiming at a time when the load on the joint for direction change becomes small (for example, a swing phase and a momentum when the leg is rotated is small), the joint angle at this time is between the command value and the actual value. Since the difference between the two is extremely small, if the braking is applied at this time, the energy consumption can be reduced as a result even if the supply current value to the motor is not particularly controlled to be low. According to this principle, it is not necessary to control the value of the current supplied to the motor to be low even if the braking force is applied to all joints, aiming at the time when the external load is small.

(5) In the present embodiment, the biped robot is taken as an example of the walking form, but the present invention can be applied to all legged robots other than biped.

[0138]

As described above, according to the braking control device for a mobile robot of the present invention, a braking device capable of generating a controllable braking force is provided at the joint portion of the robot, and the joint of the mobile robot is connected to a desired joint angle. Since the braking force is controlled by giving a braking command to the braking device at a time required to hold the robot, it is possible to save energy consumed when the robot walks.

In particular, when the legged robot is stationary, each joint can be braked and fixed to significantly reduce or stop the current supplied to the motor. Has a remarkable practical energy-saving effect.

When at least one of the joints of the legged robot is in a stationary state, the joints can be braked and fixed to significantly reduce or stop the current supplied to the drive motor of the joints. The energy consumed at the joint can be saved. Especially when there is a joint that is expected to be stationary for a relatively long time, such as a joint for turning direction, the present invention can be applied to this joint to save energy consumption during walking. be able to.

Further, when going down the stairs, for example, by applying a braking force according to the load moment of the knee joint to the knee joint when the load is heavy, the energy used for walking can be reduced. it can.

Furthermore, even in the case of walking on a level ground, if there is a joint that can walk even if the joint angle is kept constant for a certain period of time, a braking force is applied to the joint during that period to save unnecessary motor power. it can.

[Brief description of drawings]

FIG. 1 is a perspective view showing a skeleton of a bipedal robot to which the present invention is applied.

FIG. 2 is a diagram schematically showing a braking device for a knee joint of the bipedal walking robot.

FIG. 3 is a pneumatic control system diagram for sending compressed air to a braking device provided in each joint.

FIG. 4 is a block diagram showing a first embodiment of a braking control device for a mobile robot according to the present invention.

FIG. 5 is a diagram showing an example in which a joint command value is expressed in the form of a vector in the same embodiment.

FIG. 6 is a diagram showing an example in which a joint command value is expressed in the form of a discrete number so that it can be applied to a computer.

FIG. 7 is a diagram showing a control algorithm for explaining the operation of the embodiment.

FIG. 8 is a diagram showing a control algorithm similarly to FIG. 7.

FIG. 9 is a pneumatic control system diagram for sending compressed air to a braking device according to a second embodiment of the present invention.

FIG. 10 is a diagram showing a proportional constant of the same embodiment.

FIG. 11 is a diagram showing a control algorithm for explaining the operation of the embodiment.

FIG. 12 is a curve diagram showing an angle change of a knee joint with respect to a time axis when a human is walking.

[Explanation of symbols]

10, 12, 14, 16, 18, 20: Joint axis 22: 6-axis force sensor 24: foot 26: Waist board link 30: Thigh link 32: shin link 207: Cylinder 222: piston 224: rotating disk 226: Air compression source 302: CPU 304: Command 306: memory 308: Encoder 310: Counter 312: First amplifier 314: Motor 316: joint 320: Second amplifier 322: Solenoid valve

Claims (12)

[Claims] 1. A mobile robot having at least one joint part in which two links, a first link and a second link, are connected by a motor so that they can move relative to each other. A braking control device for a mobile robot, comprising: friction braking means capable of generating a braking force; and control means capable of selectively controlling the friction braking means in a braking state or a released state. 2. The braking control device for a mobile robot according to claim 1, wherein the control means is configured to control the friction braking means to a braking state when no relative movement is planned between the two links. A braking control device for a mobile robot, characterized in that 3. The braking control device for a mobile robot according to claim 1, wherein the control means controls the friction braking means to a braking state when the relative movement direction of the two links is opposite to the movement direction due to the motor torque. A braking control device for a mobile robot, which is configured to: 4. The braking control device for a mobile robot according to claim 3, wherein the control means is configured to variably control the braking force of the friction braking means. apparatus. 5. A mobile robot having at least one joint part in which two links, a first link and a second link, are coupled by a motor so that they can move relative to each other, wherein the two links are provided so as to be bridgeable. Friction braking means for applying a controllable relative movement restriction force to the relative movement of the link, and the relative movement restriction by controlling the friction braking means at a time required to hold the joint of the mobile robot at a desired joint angle. A braking control device for a mobile robot, comprising: a control unit that generates force. 6. A mobile robot having at least two legs that move by being alternately moved by a motor and a joint for changing direction, the movement robot being provided at a portion corresponding to the joint so as to suppress the movement of the joint. Braking means for generating various frictional forces, and at least 1 when the robot is in a straight ahead state.
A step is a braking control device for a mobile robot, comprising: a control unit that controls the friction braking unit to apply a brake to the joint. 7. The braking control device for a mobile robot according to claim 2, 5, or 6, wherein the energy supplied to the motor when the friction braking means is in a braking state. A braking control device for a mobile robot, comprising: an energy supply control means for reducing or stopping the operation. 8. The braking control device for a mobile robot according to claim 7, wherein the energy supply control means reduces or stops the energy supply to the motor after the braking means is controlled by the control means. A braking control device for a mobile robot, which is configured to: 9. A walking robot having at least one joint part in which two links, a first link and a second link, are connected by a motor so that they can move relative to each other, wherein the two links are provided so as to be bridgeable. At least one period of the walking of the walking robot (the period from the landing of one leg to the landing to the landing to the landing again) of at least one friction braking means for giving a controllable relative motion restriction force to the relative motion of the link A braking control device for a walking robot, comprising: a control unit that controls the friction braking unit to generate the relative motion restriction force once. 10. The braking control device for a walking robot according to claim 9, wherein the walking robot has at least one leg consisting of three joints connected in series in the order of a hip joint, a knee joint, and an ankle joint. And a foot link linking to a lower end of the ankle joint, wherein the joint portion is at least one of a knee joint and ankle joint of the three joints. apparatus. 11. The braking control device for a walking robot according to claim 9, wherein the walking robot has three joints of a hip joint, a knee joint, and an ankle joint, and a thigh link connecting these joints, like a human. At least two legs having a structure including a shin link,
Moreover, these legs are alternately moved to walk, and the control means is provided in at least one joint part of the three joint parts of the leg at the time of landing when the one leg shifts from the swing phase to the stance phase. A braking control device for a walking robot, characterized in that the frictional braking means is controlled so as to generate the relative motion restriction force. 12. The braking control device for a walking robot according to claim 9, wherein the walking robot has three joints, a hip joint, a knee joint, and an ankle joint, and a thigh link that connects these joints, like a human. At least two legs having a structure including a shin link,
Further, the robot walks by alternately moving these legs, and the control means is provided with at least one joint part of the three joint parts of the leg in the stance phase when the robot descends the stairs. A braking control device for a walking robot, characterized in that the control is performed so as to generate the relative motion restriction force.