JPH02188809A - Controller for avoiding obstacle of traveling object - Google Patents

Abstract

PURPOSE:To attain the flexible avoiding actions like those of human beings against a dynamic obstacle set in any state with no deadlock by calculating the avoiding direction based on the degree of danger obtained via a fuzzy recognizing part. CONSTITUTION:The present positions of a mobile robot 1, an obstacle, and a target point received from a sensor group 13 are converted into the polar coordinate display by a state input part 14 in terms of a robot coordinate system. Then a fuzzy recognizing part 16 decides a static degree of danger based on the polar coordinate display of those positions and also a dynamic degree of danger based on the polar coordinate display of speed respectively. A behavior deciding part 17 decides an avoiding vector from a decision table contained in a data base part 15 and based on the dynamic and static degrees of danger. Then the part 17 synthesizes the avoiding vector and a target arriving vector to decide a steering vector. Thus the flexible avoiding actions are attained like those of human beings.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to an obstacle avoidance control device for a mobile object such as a mobile robot that travels in an environment where dynamic obstacles exist.

[Conventional technology]

FIG. 8 is a principle diagram of a conventional obstacle avoidance control device for a moving object disclosed in, for example, Japanese Unexamined Patent Publication No. 63-5408. In the figure, 1 is a mobile robot as a moving object, and 2 is an external sensor. distance sensor (for example, consisting of an ultrasonic sensor), OB is the obstacle, R is the mobile robot 1 and the obstacle OB
Indicates the distance from

Further, P indicates a three-dimensional image (potential surface) virtually constructed from distance information.

Next, the operation will be explained. For example, consider an example in which the mobile robot 1 traveling on a two-dimensional surface shown in FIG. 8(A) is controlled to avoid colliding with an unknown obstacle OB. (Distance information, ie, distance R), the presence of the obstacle OB is detected and the distance is recognized.

If the work space (movement space) of the mobile robot 1 is two-dimensional with the X and Y axes as shown in Figure 8 (A), then the P axis is added to this as shown in Figures 8 (B) and (C). Assume a three-dimensional virtual space. Then, with respect to the output (distance R) of the distance sensor 2 obtained in the work space of Fig. 8 (A), the position R in the virtual space has an extent in the X and Y axis directions and a height in the P axis direction. A three-dimensional image P is generated. That is, the output of the distance sensor 2 in real space is converted into a three-dimensional image P in virtual space.

In this virtual space, one-dimensional distance information is represented as a multidimensional three-dimensional object as shown in FIG. 8(C).

Since this transformed image in the virtual space has a height t in the P-axis direction, we use the state quantity of the field in the P-axis direction due to the 3D image of the obstacle OB at position A in the virtual space corresponding to the mobile robot 1. The control amount can be obtained by using the control amount, and the movement of the mobile robot 1 can be adaptively controlled based on this.

That is, assuming an n+1)-dimensional virtual space relative to an n-dimensional real space (work space), a three-dimensional image is formed in the virtual space by the sensor output in the real space, and the control amount for the mobile robot 1 is calculated from the three-dimensional image. That's what you're trying to get.

In this case, since the sensor output in the real space is converted into a three-dimensional image in the (n+1)-dimensional virtual space relative to the n-dimensional real space, one-dimensional detection information can be handled as multidimensional information.

Therefore, by using this virtual space as a kind of control base and defining the stereoscopic image in an appropriate shape (for example, a multidimensional normal distribution surface as shown in FIG. 8(C)),
) at point A of mobile robot 1.

The control amount in the Y-axis direction can be obtained from the height and inclination of the stereoscopic image P at point A.

That is, in the example of obstacle avoidance shown in FIG. 8(A), in the virtual space, a control amount that does not approach the three-dimensional image P based on the sensor output is obtained, and as a result, the obstacle avoidance movement is executed. Become.

[Problem to be solved by the invention]

Since the conventional obstacle avoidance control device for a moving object is configured as described above, a sensor with a certain degree of accuracy is required as an external sensor, and it is also necessary to use a sensor with a certain degree of accuracy as an external sensor. If such exists, a phenomenon called stagnation occurs in which the positive potential and negative potential are balanced, and as a result, there is a problem in that the moving object gets stuck in a deadlock.

This invention was made to solve the above-mentioned problems; it does not require particularly high precision external sensors that provide input information, and it also enables flexible avoidance behavior similar to humans. The purpose of this invention is to obtain an obstacle avoidance control device for a moving body.

[Means to solve the problem]

The obstacle avoidance control device for a moving object according to the present invention has a database section that calculates the degree of danger consisting of static danger level and dynamic danger level for the obstacle of the moving body based on the input information from the external sensor. A fuzzy recognition unit that calculates based on fuzzy theory based on data, the static risk level and dynamic risk level calculated by this fuzzy recognition unit, an action determination unit that calculates the avoidance direction of the moving object from the data in the database unit, and this action determination unit. The apparatus includes a movement data generation part that obtains movement data of the moving object from the avoidance direction obtained by the part.

[For production]

In the obstacle avoidance control device for a mobile object according to the present invention, the fuzzy recognition section first determines the degree of danger consisting of ambiguous static degree of danger and dynamic degree of danger based on the data in the database section based on the input information from the external sensor. Based on this, the action determining section further determines the avoidance direction of the moving object from the data in the database section, and realizes an avoidance action similar to human action.

〔Example〕

An embodiment of the present invention will be described below with reference to the drawings. 1st
In the figure, l is a mobile robot as a moving body, 12 is a determination mechanism which is a main part of the present invention for realizing the obstacle avoidance control device for a mobile body according to the present invention, and 13 is a measuring mechanism for measuring the distance to an obstacle. 14 is a state input unit that converts sensor information from the sensor group 13 to obtain predetermined input information, and 15 is a static risk level and a dynamic level sensor, which will be described later. A database section 16 in the judgment mechanism 12 that stores membership functions and decision tables for determining the risk level and the avoidance direction consisting of the risk level, calculates the risk level from ambiguous input information from the state input section 14. 17 is a fuzzy recognition unit that generates roughly recognition information; 17 is an action determining unit that determines the actual avoidance direction of the moving robot); 18 is a movement determination unit that determines the avoidance direction determined by this action determination unit 17; A movement data generating section 19 that creates actual movement data of the robot 1 is a servo command section that gives servo commands to a group of actuators 20 such as motors based on the movement data.

FIG. 2 shows a judgment flowchart conceptually showing the function of the judgment mechanism 12, and the operation of the embodiment shown in FIG. 1 of this invention will be explained below.

First, to explain the basic operation, the distance information input from the sensor group 13 is converted by the state input unit 14 into input information of polar coordinate values representing the position and speed of the obstacle as seen from the robot coordinate system. This input information is sent to the judgment mechanism 12, and in the judgment mechanism 12, a fuzzy recognition section 16 calculates the degree of danger for the current obstacle condition using the membership function stored in the database section 15. The action determining unit 17 uses the decision table in the database unit 15 to determine the actual avoidance direction.

Furthermore, the movement data generation unit 18 converts the obtained avoidance direction into movement data of a specific movement amount, and provides it to the servo command unit 19 to drive the actuator group 20 in accordance with the servo command.

Next, the operation of the determining mechanism 12 will be explained in more detail with reference to FIG. As the status input information from the sensor group 13,
Mobile robot 1. When the current positions of obstacles and target points are given, the state input unit 14 first converts them into relative positions (S-,
St) and relative velocity (uJ, , IJt), and finally converted to polar coordinates as seen from the robot coordinate system. Next, the fuzzy recognition unit 16 determines the static risk α using the polar coordinate representation of the position, and determines the dynamic danger β using the polar coordinate representation of the speed. In other words, the static danger level α is a parameter that indicates that objects in front of the direction in which a person is moving poses a danger, while the dynamic danger level β indicates that an object that is moving toward a person at a high speed poses a danger. It can be thought of as a parameter that indicates what you feel. Note that the fuzzy recognition unit 16 uses the membership function of the database unit 15 when determining the static risk level α and the dynamic risk level β.

Next, based on the static risk level α and dynamic risk level β, the action determining unit 17 determines the avoidance vector 0 from the decision table in the database unit 15. Furthermore, if the target reaching vector D for the target point is obtained, the action determining unit 17 determines the steering vector M by combining the avoidance vector 0 and the target reaching vector D. Note that the target arrival vector D is given by a constant vector that constantly points toward the target point. Based on the obtained steering vector M, a command steering angle is determined in the movement data generating section IB and outputted as movement data.

Here, FIG. 3 shows a state model in which the mobile robot 1 moves toward a target point while avoiding moving obstacles. Here, the vectors in the figure indicate the following.

1'r,P,,PL,: Mobile robot 1. Absolute position vectors of obstacles and target points, V,,V. , Vt: Mobile robot 1. Obstacle, absolute velocity vector of target point $,, $t: Moving robot) Obstacle seen from one coordinate system,
Relative position vector a,,, Ut of target point: Obstacle seen from mobile robot 1 coordinate system, relative velocity vector of target point 0; Obstacle avoidance vector ID of mobile robot 1 = Goal arrival vector M of mobile robot 1 = Steering Vector of Mobile Robot 1 Next, more detailed processing contents in the determination mechanism 12 will be further explained below. First, based on fuzzy theory,
The expression of danger arising from the static positional relationship between the mobile robot 1 and obstacles will be explained.

Orientation (progressing direction) of mobile robot 1 in the absolute coordinate system
Let φ be the absolute velocity vector vr-[V,8. V.P.
]7 ([]” indicates transposition), φ is expressed by the following formula.

φ=K/2 tan-' (Vry/Vry
) Also, the relative position vector $ of the obstacle seen from the robot Cartesian coordinate system. is determined by the following formula.

$,=R・(P,-P,)=[S,x,S,,]'
(2) Therefore, the robot polar coordinate system (the angle is 0 with respect to the direction of movement)
degree, the right side is positive and the left side is negative), and the direction θ1 of the obstacle is L,=S,, ``+S,,'' (3
) θm= π/ 2 jan-' (soy/ 5o
ll) (4) becomes. Then, the membership functions shown in FIGS. 4(a) and 4(b) are constructed from θ8.

The polar coordinate direction θ8 is considered only when an object exists in the range from 190'' to 90' (the range from directly beside to in front of the moving direction of the mobile robot 1), and has four levels of membership as shown in Fig. 4(a). Set the function.

Here, L L (Left L
arge) is a large angle to the left, L S (Left
Small) is a small angle to the left, RS (Right
Small) is a small angle on the right side, RL (Right
t Large) indicates when an object exists at a large angle on the right side. Note that a triangular function was used as the membership function in this case.

Next, we considered the case where an object exists within 150 cm from the polar coordinate distance 0, and provided a three-stage membership function as shown in FIG. 4(b). That is, N (Near
) is close, M (Medium) is in the middle range, F (
Par) indicates a far time.

For these fuzzy expressions of direction and distance, static risk α is expressed using a membership function as shown in FIG. 4(C). Here, the left side is LDL (Left
Danger Large), L D S (L
eft Danger3s+aJI), L S S
(Left 5afe Sn+all), L
In the order of SL (Left Small Large), the right side is RDL (Right Danger L).
arge), RD S (Right D
(angerSmall), RS S (Right
5afe Small), RS L (Right
Safe, Large) indicates the highest degree of risk.

In the membership function, the negative side of the horizontal axis indicates the degree of danger on the left side of the robot, and the positive side indicates the degree of danger on the right side of the robot, with the vicinity of O indicating safety and the vicinity of -4 and 1 indicating danger.

Fuzzy inference is performed on the above membership functions according to FIG. In the figure, LL, LS, R3゜RL, N, M
, F are fuzzy representations of direction and distance, and the static risk is determined for these as shown in the figure. For example, slightly to the right (R
In 3), when there is an obstacle nearby (N), the right side is considered very dangerous (RDL), and when it is thick (left (LL)) and in the middle range (M), the left side is considered slightly safer (LSS). In this way, given the polar coordinate values of direction and distance,
The static risk α is determined as the center of gravity of the composite membership function by fuzzy inference.

Next, we will consider expressing the danger caused by the direction and magnitude of the relative velocity between the mobile robot 1 and an obstacle using fuzzy theory.

Using the coordinate system transformation matrix R of the mobile robot 1 described above,
The relative velocity vector U0 of the obstacle seen from the robot orthogonal coordinate system is as follows.

TJo=R・(Vo−Vr)=[Uax+Uoy]”
(5) Furthermore, this vector is converted into relative velocity vectors B and r as seen from the robot orthogonal coordinate system with the direction of the obstacle as the y-axis direction.

U,'=R,・U,=[U,,',U,,']"
(6) Therefore, the speed Wu and speed direction φ1 of the obstacle as seen from the polar coordinate system of the robot facing the direction of the obstacle (the angle is 0° in the direction of the obstacle, positive on the right, negative on the left) are as follows: become.

we4j胛子〒+-U、,' ” (7)φ
8-π/2 jan-'(Uoy'/Uox
') (8) Figure 4 (b) shows the polar coordinate velocity direction φ. The membership function of The velocity direction is considered from -180" to 180° (considering the velocity in all directions centering on the direction of the obstacle to the mobile robot 1), and a six-step membership function is set. In the figure, VLL (Veloci
ty Left Large) + V L M (Ve
location LeftMedium+n) + V L
S (Velocity Left Small) is on the left side, VRS (Velocity Right Sm
all ), V RM (Velocity Rig
ht Medium), V RL (Velocit
yRight Large) indicates the right velocity direction.

What should be noted here is that the speed direction uses the relative speed obtained by subtracting the robot speed from the obstacle speed, so when it is around 0°, it means that the obstacle is moving away, and when it is around 180° or 180', it means that the obstacle is moving away. It indicates that something is approaching.

Next, the membership function of the polar coordinate velocity Wu is shown in Figure 4 (e
). Considering the speed from 0 to 100 CI/S, a two-stage membership function as shown in the figure is provided. That is, V S (Velocity Slo
w) is slow, V F (Velocity Fas
t) indicates fast speed.

Next, the dynamic risk level β uses the same membership function shown in FIG. 4 (b) as the static risk level α. Fuzzy inference is performed using these membership functions according to FIG. In the figure, ■S and ■F are the magnitude of the velocity, VLL, and VLM for mobile robot 1.

VLS, VR5, VRM, and VRL are velocity direction (7)7agii expressions. For example, the speed of an obstacle approaching with a large (VF) rightward speed (VRL) is very dangerous to the right (RDL), and the speed of a small (VS) obstacle that is moving away with a leftward speed (VLS) is moving leftward. It is judged as fairly safe (LSL). Once the speed and speed direction are obtained as described above, the dynamic risk degree β can be determined by fuzzy inference.

Furthermore, the already obtained static risk α is equal to the dynamic risk β
Consider determining the avoidance vector for an obstacle using a decision table. FIG. 6 shows a table in which the avoidance directions are determined, with α being the horizontal axis and β being the vertical axis. This decision table can define various avoidance strategies, but in this method, the strategies were set to make decisions relatively similar to humans. The numbers in the table represent the avoidance direction, and this avoidance direction is shown in FIG.

1+ to 5+ represent rightward steering, and 1- to 5- represent leftward steering, each increasing by 15'' from 1, with 5 indicating 75'' steering.

Note that when the vehicle is set to 0, the vehicle continues to drive without turning the steering wheel. It is possible to incorporate characteristic avoidance strategies that correspond to human behavioral know-how here.

The avoidance direction determined as described above is given as a unit vector in the robot orthogonal coordinate system, and this becomes the obstacle avoidance vector 0.

The above-mentioned problem was related to determining the response to obstacles, but we will also discuss the method for reaching the target point. For obstacles, we used fuzzy and decision tables to simulate rough risk judgments and complex avoidance strategies, but for the target point, we simply made the robot face the target force direction. Let be given the following unit vector.

D=st/1stl = (rt-r,)/1rt-r,l (9) The second goal arrival vector D is the same as the obstacle avoidance vector 0,
It is expressed as a unit vector in the robot Cartesian coordinate system.

The avoidance vector for the obstacle is given by the unit vector in the avoidance direction determined in FIG. 6, and the arrival vector for the target point is given by the unit vector determined by equation (9).

From these vectors, the steering vector is determined using the following formula.

M=(D+0)/I I)+01 01 this 0
(The steering of the mobile robot l is set in the direction of the steering vector M obtained by equation D, and the magnitude obtained by multiplying this unit vector by the speed given in advance is output to the movement data generation unit 18 as the next command value. Ru.

In the above embodiment, an example of obstacle avoidance for the mobile robot 1 was shown, but assuming a manipulator as the moving object, the present invention can also be applied to obstacle avoidance at the tip of the hand or each joint of the manipulator.

Further, in the above embodiment, an example was shown in which the avoidance direction was ultimately determined as the avoidance method, but a method of determining the avoidance speed or the like may also be used.

Further, in the above embodiments, an ultrasonic sensor is used as an external sensor, but any sensor may be used as long as it can roughly measure distance and speed.

〔Effect of the invention〕

As described above, according to the present invention, there is a database unit that stores data etc. for determining the vague degree of risk using fuzzy theory, a fuzzy recognition unit that determines the degree of risk, and an avoidance direction of the moving object based on the determined degree of risk. Since we configured an obstacle avoidance control device for a moving object using the desired action determining unit, we are able to perform flexible avoidance movements similar to those of humans without deadlocking against dynamic obstacles in any state. There is a certain effect.

Furthermore, since the obstacle detection information may be ambiguous information obtained by a low-precision external sensor such as an ultrasonic sensor, the sensor system can be configured at a low cost.

[Brief explanation of the drawing]

FIG. 1 is a configuration diagram of an obstacle avoidance control device for a moving object according to an embodiment of the present invention, FIG. 2 is an explanatory diagram showing a judgment flowchart of the judgment mechanism 12, and FIG. 3 is an explanation of setting conditions for obstacle avoidance. Figure 4 is an explanatory diagram of the membership function in the database part δ, and Figure 5 is a static risk level α and a dynamic risk level β.
FIG. 6 is an explanatory diagram of the decision table in the database unit 15, FIG. 7 is an explanatory diagram of the avoidance direction, and FIG. 8 is an explanatory diagram of the conventional obstacle avoidance control device for a moving object. It is a diagram. 1 is a mobile robot (mobile body), 13 is a sensor group (external sensor), 15 is a database section, 16 is a fuzzy recognition section, 17 is an action determination section, and 18 is a movement data generation section. In addition, in the figures, the same reference numerals indicate the same or equivalent parts.

Claims (1)

[Claims] an external sensor that detects an obstacle on a moving body; data for determining a degree of danger consisting of a static degree of danger and a dynamic degree of danger of the mobile body with respect to the obstacle; and an avoidance direction of the mobile body based on the degree of danger; a database unit that stores data for determining the above, and a fuzzy recognition unit that calculates the static risk level and the dynamic risk level based on the data in the database unit based on the input information from the external sensor using fuzzy theory; an action determining unit that determines the avoidance direction of the moving object based on the static risk level and dynamic risk level determined by the fuzzy recognition unit and data in the database unit; An obstacle avoidance control device for a moving body, comprising: a movement data generation unit that obtains movement data of a body and causes the moving body to move in the avoidance direction.