Robot and control system

• US 8,849,452 B2
• Filed: 05/25/2010
• Issued: 09/30/2014
• Est. Priority Date: 08/03/2009
• Status: Active Grant

• Current Assignee: Honda Motor Co., Ltd. (Honda Motor Company)
• Original Assignee: Honda Motor Co., Ltd. (Honda Motor Company)
• Inventors: Takeda, Masanori, Kuroda, Mitsuhide, Yoshiike, Takahide, Kamioka, Takumi, Kanzaki, Shigeru
• Primary Examiner(s): Holloway, Jason
• Assistant Examiner(s): Sample, Jonathan L
• Application Number:
  US12/999,791
• Publication Number:
  US 20110231017A1
• Time in Patent Office: 1,589 Days
• Field of Search: None
• US Class Current: 700/250
• CPC Class Codes: B25J 9/1679 characterised by the tasks executed
• CPC Class Codes: B62D 57/032 with alternately or sequentially lifted supporting base and legs; with alternately or sequentially lifted feet or skid B62D57/024 takes precedence
• CPC Class Codes: G05B 2219/40496 Hierarchical, learning, recognition level controls adaptation, servo level

Abstract

A control system or the like capable of causing a controlled object to act in an appropriate form in view of an action purpose of the controlled object to a disturbance in an arbitrary form.

Each of a plurality of modules modi, which are hierarchically organized according to the level of a frequency band, searches for action candidates which are candidates for an action form of a robot R matching with a main purpose and a sub-purpose while giving priority to a main purpose mainly under the charge of the module over a sub-purpose mainly under the charge of any other module. The actions of the robot R is controlled in a form in which the action candidates of the robot R searched for by a j-th module of a high frequency are reflected in preference to the action candidates of the robot R searched for by a (j+1)th module of a low frequency.

Claims

1. A control system comprising at least one processor, the at least one processor comprising a plurality of modules hierarchically organized according to the level of a frequency band, wherein:

each of the plurality of modules has a main purpose, and is configured to search for a plurality of action candidates which are candidates for an action form of a controlled object, and to evaluate total fitness in view of both the main purpose and a sub-purpose, which is the main purpose of a lower-order module, in such a way that the contribution of a main fitness to the main purpose is higher than the contribution of a sub-fitness to the sub-purpose, with respect to each of the action candidates; and

the control system is configured to control an operation of the controlled object in a form in which an evaluation result of the total fitness obtained by a high-order module of a high frequency is reflected in preference to an evaluation result of the total fitness obtained by a low-order module of a low frequency, and the main purposes of each the plurality of modules are reflected in the control of the operation of the controlled object by the control system.

2. The control system according to claim 1, wherein the plurality of modules comprise a highest-order module of highest frequency, a lowest-order module of lowest frequency, and an intermediate module configured to evaluate the total fitness in such a way that high-order fitness in view of a high-order purpose is higher in contribution than low-order fitness in view of a low-order purpose with respect to each of the action candidates with the high-order purpose of the high-order module of a frequency higher than the low-order module, and the low-order purpose of the low-order module of a frequency lower than the high-order module as the sub-purposes.

3. The control system according to claim 2, wherein the low-order module is configured to evaluate estimated low-order fitness by estimating the low-order fitness with respect to each of the action candidates and evaluate the total fitness with the estimated low-order fitness as the low-order fitness.

4. The control system according to claim 3, wherein the low-order module is configured to receive a low-order action candidate whose low-order total fitness, as the total fitness evaluated by the low-order module, among low-order action candidates as the action candidates searched for by the low-order module, is a maximum or a local maximum as a low-order action guideline from the low-order module and evaluate a degree of approximation to the low-order action guideline as the estimated low-order fitness with respect to each of the action candidates searched for by the low-order module.

5. The control system according to claim 3, further comprising first to N-th modules as the plurality of modules, wherein:

a first module as the highest-order module is configured to evaluate first fitness e₁ to a first purpose as the main purpose with respect to each of first action candidates as the action candidates searched for by the first module, evaluate second estimated fitness e₂^ by estimating fitness to a second purpose of a second module as a low-order module, and then evaluate first total fitness f₁ according to an evaluation equation f₁=e₁(e₂^+1);

an i-th module (i=2 to N−

1) as the intermediate module is configured to evaluate i-th fitness e_i in view of an i-th purpose as the main purpose with respect to each of i-th action candidates as the action candidates searched for by the i-th module, evaluate (i−

1)th fitness e<sub>−

1</sub> in view of the (i−

1)th purpose as the high-order purpose, evaluate (i+1)th estimated fitness by estimating fitness in view of an (i+1)th purpose as the low-order purpose, and then evaluate the i-th total fitness f_i according to an evaluation equation f_i=e_i(e<sub>i−

1</sub>(e_i+1^+1)+1); and

an N-th module as the lowest-order module is configured to evaluate N-th fitness e_N to an N-th purpose as the main purpose with respect to each of N-th action candidates as the action candidates searched for by the N-th module, evaluate the (N−

1)th fitness e<sub>N−

1</sub> of an (N−

1)th module as the high-order module, and then evaluate the i-th total fitness f_i according to an evaluation equation f_N=e_N(e<sub>N−

1</sub>1).

6. The control system according to claim 5, wherein the j-th module (j=1, 2, - - - , N) is configured to evaluate j-th fitness e_j to a j-th purpose as the main purpose at a fixed value, with respect to j-th action candidates included in j-th specified action candidate group among all j-th action candidates as the action candidates searched for by the j-th module, while evaluating the j-th fitness e_j at a value lower than the fixed value with respect to the j-th action candidates not included in the j-th specified action candidate group among all j-th action candidates.

7. The control system according to claim 6, wherein the j-th module is configured to evaluate the j-th fitness e_j at the fixed value, with the j-th action candidates whose divergence from a reference point defined in a state space adequate for the j-th purpose among all of the j-th action candidates is equal to or less than a j-th threshold as the j-th action candidates included in the j-th specified action candidate group, while evaluating the j-th fitness e_j according to a decreasing function in which the divergence is continuous or progressive on the basis of the divergence with respect to the j-th action candidates not included in the j-th specified action candidate group among all of the j-th action candidates.

8. The control system according to claim 1, wherein each of the modules is configured to search for action candidates which are candidates for the action form of the controlled object matching with the main purpose and the sub-purpose while giving priority to the main purpose over the sub-purpose.

9. The control system according to claim 8, wherein each of the modules is configured to perform the current time'"'"'s search for action candidates of the controlled object according to a search strategy based on a previous time'"'"'s search result obtained by the module and a previous time'"'"'s search result of the action candidates of the controlled object obtained by the any other module while giving priority to the previous time'"'"'s search result of the action candidates of the controlled object by the module over the previous time'"'"'s search result of the action candidates of the controlled object by the any other module.

10. The control system according to claim 9, wherein each of the modules is configured to select a part of the action candidates searched for at the previous time with higher probability as the total fitness evaluated at the previous time is higher and perform the current time'"'"'s search for the action candidates in a peripheral region of the selected action candidates in the state space defined by the action candidates.

11. The control system according to claim 10, wherein each of the modules is configured to perform the current time'"'"'s search for the action candidates in a narrower peripheral region as the total fitness evaluated at the previous time is higher with reference to the action candidates selected among the action candidates searched for at the previous time in the state space.

12. The control system according to claim 10, wherein each of the modules is configured to perform the current time'"'"'s search for the action candidates in a peripheral region of irregularly selected points in the state space in addition to the peripheral region of the action candidates selected among the action candidates searched for at the previous time.

13. The control system according to claim 12, wherein each of the modules is configured to search for the action candidates of the controlled object, with a high-order module as the any other module in the case where there is a module one order higher than the module and with a low-order module as the any other module in the case where there is a module one order lower than the module.

14. The control system according to claim 8, wherein:

each of the modules is configured to search for the action candidates which determine the position or the position and posture of a moving apparatus as the controlled object; and

each of the modules is configured to search for the action candidates of the moving apparatus for a shorter period in comparison with the low-order module of a low frequency as the high-order module has a higher frequency.

15. The control system according to claim 14, comprising a first module, a second module, and a third module, as the plurality of modules, configured to search for the position trajectory or the position trajectory and posture trajectory of a robot as the moving apparatus having a base and a plurality of legs extended from the base, as the action candidates, wherein:

the first module is configured to search for a gait over a first specified number of steps of the robot for causing the robot to stabilize the posture as a first action candidate;

the second module is configured to search for a local route, as a second action candidate, which determines a gait over a second specified number of steps, which is greater than the first specified number of steps, for causing the robot to prevent a contact with an object; and

the third module is configured to search for a general route for causing the robot to reach a target position as a third action candidate.

16. The control system according to claim 1, wherein each of the modules is configured to predict a plurality of future states of the controlled object according to each of the action candidates searched for by the module and evaluate the total fitness on the basis of each of the future states with respect to each of the action candidates.

17. The control system according to claim 16, wherein:

the module having a high-order module of a higher frequency than the module is configured to predict further future states of the controlled object subsequent to the future states of the controlled object predicted by the high-order module; and

the operation of the controlled object is configured to be controlled in a form in which the evaluation result of the total fitness obtained by the high-order module of a high frequency is reflected in precedence to the evaluation result of the total fitness by the low-order module of a low frequency.

18. The control system according to claim 17, wherein each of the modules is configured to search for the action candidates of a moving apparatus for a shorter period in comparison with the low-order module of a low frequency as the high-order module has a higher frequency.

19. The control system according to claim 18, wherein at least one module among the plurality of modules is configured to predict the state of the controlled object at a future time point after an elapse of a shorter period of an action form than the action candidates searched for by a highest frequency module according to the current state of the controlled object, as a nearest future state;

and

each of the plurality of modules is configured to predict the future state of the controlled object according to the action candidates searched for by the module, subsequent to the nearest future state.

20. The control system according to claim 18, wherein each of the modules is configured to search for the action candidates which determine the position or the position and posture of the moving apparatus as the controlled object and predict the position or the position and posture of the moving apparatus as the future state of the controlled object.

21. The control system according to claim 20, comprising a first module, a second module, and a third module, as the plurality of modules, configured to search for the position trajectory or the position trajectory and posture trajectory of a robot as the moving apparatus having a base and a plurality of legs extended from the base as the action candidates, wherein:

the first module is configured to search for a gait over a first specified number of steps of the robot for causing the robot to stabilize the posture as a first action candidate and predict the position or the position and posture in the future of the robot according to each of the first action candidates as a plurality of first future states;

the second module is configured to search for a local route, as a second action candidate, which determines a gait over a second specified number of steps, which is greater than the first specified number of steps, for causing the robot to prevent a contact with an object and predict the position or the position and posture in the future of the robot according to each of the second action candidates as a plurality of second future states; and

the third module is configured to search for a general route to cause the robot to reach a target position or to cause the robot to reach the target position with a target posture as a third action candidate, and predict the position or the position and posture in the future of the robot according to each of the third action candidates as a plurality of third future states.

22. The control system according to claim 21, wherein:

at least one module among the first module, the second module, and the third module is configured to predict the position or the position and posture of the robot at a future time point after an elapse of a gait over a reference number of steps less than the first specified number of steps as the nearest future state according to the current state of the robot; and

the first module is configured to predict the position or the position and posture of the robot at the future time point after an elapse of a gait over the first specified number of steps as the first future state with the nearest future state as the origin;

the second module is configured to predict the position trajectory or the position trajectory and posture trajectory of the robot moving along the local route as the second future state with the nearest future state as the origin; and

the third module is configured to predict the position trajectory or the position trajectory and posture trajectory of the robot moving along the general route as the third future state with the nearest future state as the origin.

23. A robot comprising:

a base and a plurality of legs extended from the base, the robot being a moving apparatus which moves with the movements of the plurality of legs; and

a control system comprising at least one processor, the at least one processor comprising a plurality of modules hierarchically organized according to the level of a frequency band, wherein;

each of the plurality of modules has a main purpose, and is configured to search for a plurality of action candidates which are candidates for an action form of a controlled object, and to evaluate total fitness in view of both the main purpose and a sub-purpose, which is the main purpose of a lower-order module, in such a way that the contribution of a main fitness to the main purpose is higher than the contribution of a sub-fitness to the sub-purpose, with respect to each of the action candidates;

the control system is configured to control an operation of the controlled object in a form in which an evaluation result of the total fitness obtained by a high-order module of a high frequency is reflected in preference to an evaluation result of the total fitness obtained by a low-order module of a low frequency, and the main purposes of each the plurality of modules are reflected in the control of the operation of the controlled object by the control system;

each of the modules is configured to search for action candidates which are candidates for the action form of the controlled object matching with the main purpose and the sub-purpose while giving priority to the main purpose over the sub-purpose;

each of the modules is configured to search for the action candidates which determine the position or the position and posture of a moving apparatus as the controlled object;

each of the modules is configured to search for the action candidates of the moving apparatus for a shorter period in comparison with the low-order module of a low frequency as the high-order module has a higher frequency; and

the control system further comprises a first module, a second module, and a third module, as the plurality of modules, configured to search for the position trajectory or the position trajectory and posture trajectory of a robot as the moving apparatus having a base and a plurality of legs extended from the base, as the action candidates, wherein;

the first module is configured to search for a gait over a first specified number of steps of the robot for causing the robot to stabilize the posture as a first action candidate;

the second module is configured to search for a local route, as a second action candidate, which determines a gait over a second specified number of steps, which is greater than the first specified number of steps, for causing the robot to prevent a contact with an object; and

the third module is configured to search for a general route for causing the robot to reach a target position as a third action candidate.

24. A robot comprising:

a base and a plurality of legs extended from the base, the robot being a moving apparatus which moves with the movements of the plurality of legs; and

a control system comprising at least one processor, the at least one processor comprising a plurality of modules hierarchically organized according to the level of a frequency band, wherein;

each of the plurality of modules has a main purpose, and is configured to search for a plurality of action candidates which are candidates for an action form of a controlled object, and to evaluate total fitness in view of both the main purpose and a sub-purpose, which is the main purpose of a lower-order module, in such a way that the contribution of a main fitness to the main purpose is higher than the contribution of a sub-fitness to the sub-purpose, with respect to each of the action candidates;

the control system is configured to control an operation of the controlled object in a form in which an evaluation result of the total fitness obtained by a high-order module of a high frequency is reflected in preference to an evaluation result of the total fitness obtained by a low-order module of a low frequency, and the main purposes of each the plurality of modules are reflected in the control of the operation of the controlled object by the control system;

each of the modules is configured to predict a plurality of future states of the controlled object according to each of the action candidates searched for by the module and evaluate the total fitness on the basis of each of the future states with respect to each of the action candidates;

the module having a high-order module of a higher frequency than the module is configured to predict further future states of the controlled object subsequent to the future states of the controlled object predicted by the high-order module;

the operation of the controlled object is configured to be controlled in a form in which the evaluation result of the total fitness obtained by the high-order module of a high frequency is reflected in precedence to the evaluation result of the total fitness by the low-order module of a low frequency;

each of the modules is configured to search for the action candidates of a moving apparatus for a shorter period in comparison with the low-order module of a low frequency as the high-order module has a higher frequency;

each of the modules is configured to search for the action candidates which determine the position or the position and posture of the moving apparatus as the controlled object and predict the position or the position and posture of the moving apparatus as the future state of the controlled object; and

the control system further comprises a first module, a second module, and a third module, as the plurality of modules, configured to search for the position trajectory or the position trajectory and posture trajectory of a robot as the moving apparatus having a base and a plurality of legs extended from the base as the action candidates, wherein;

the first module is configured to search for a gait over a first specified number of steps of the robot for causing the robot to stabilize the posture as a first action candidate and predict the position or the position and posture in the future of the robot according to each of the first action candidates as a plurality of first future states;

the second module is configured to search for a local route, as a second action candidate, which determines a gait over a second specified number of steps, which is greater than the first specified number of steps, for causing the robot to prevent a contact with an object and predict the position or the position and posture in the future of the robot according to each of the second action candidates as a plurality of second future states; and

the third module is configured to search for a general route to cause the robot to reach a target position or to cause the robot to reach the target position with a target posture as a third action candidate, and predict the position or the position and posture in the future of the robot according to each of the third action candidates as a plurality of third future states.