Granted Patent B2

US 12,544,927 · App. 18/343,748 · Granted Feb 10, 2026

Inventors: Dirk Ruiken (Offenbach, DE); Simon Manschitz (Offenbach, DE)

Assignee: Honda Motor Co., Ltd.

B25J9/1666B25J9/1689B25J13/089

Abstract

Claims (66)

1 . A method for controlling motion of an object towards a target pose, wherein the method comprises:

determining a current state of the object;

randomizing multiple configurations of the object with regard to the current state of the object;

predicting for each configuration of the multiple configurations of the object a constraint cost with regard to each constraint of one or more constraints;

determining for each constraint of the one or more constraints, among the predicted constraints costs that have been predicted for the multiple configurations of the object, a predicted constraint cost that meets the constraint the worst;

predicting a target configuration of the object using the multiple configurations of the object, the one or more determined constraint costs that have been determined for the one or more constraints, and the target pose; and

controlling the object to change its current state towards the predicted target configuration.

2 . The method according to claim 1 , wherein predicting for each configuration of the multiple configurations of the object the constraint cost with regard to each constraint of the one or more constraints comprises:

using a current environmental state around the object for predicting for each configuration of the multiple configurations of the object the constraint cost.

3 . The method according to claim 1 , wherein

the method is performed with regard to multiple constraints.

4 . The method according to claim 1 , wherein predicting for each configuration of the multiple configurations of the object the constraint cost with regard to each constraint of the one or more constraints comprises:

generating a value for the constraint cost that is smaller than or equal to zero in case the configuration of the object meets the constraint, wherein the smaller the value the more the configuration of the object meets the constraint; and

generating a positive value for the constraint cost in case the configuration of the object does not meet the constraint, wherein the greater the positive value the less the configuration of the object meets the constraint.

5 . The method according to claim 4 , wherein determining for each constraint of the one or more constraints, among the predicted constraints costs that have been predicted for the multiple configurations of the object, the predicted constraint cost that meets the constraint the worst comprises:

determining for each constraint of the one or more constraints, among the predicted constraints costs that have been predicted for the multiple configurations of the object, the predicted constraint cost with the greatest value.

6 . The method according to claim 1 , wherein the method comprises:

setting the one or more constraints according to a task to be performed by the object.

7 . The method according to claim 1 , wherein the method comprises:

selecting, according to a task to be performed by the object, the one or more constraints from a plurality of constraints.

8 . The method according to claim 1 , wherein predicting for each configuration of the multiple configurations of the object the constraint cost with regard to each constraint of the one or more constraints comprises:

using a neural network for each constraint of the one or more constraints that predicts for each configuration of the multiple configurations of the object the constraint cost with regard to the constraint.

9 . The method according to claim 1 , wherein predicting the target configuration of the object using the multiple configurations of the object, the one or more determined constraint costs that have been determined for the one or more constraints, and the target pose comprises:

using a neural network that predicts the target configuration based on the multiple configurations of the object, the one or more determined constraint costs that have been determined for the one or more constraints, and the target pose.

10 . The method according to claim 1 , wherein

the target pose is controlled by an operator via teleoperation.

11 . The method according to claim 1 , wherein

the one or more constraints comprise at least one of one or more constraints due to a hardware of the object, one or more constraints due to a task to be performed by the object, and one or more constraints due to an environment around the object.

12 . The method according to claim 11 , wherein

the one or more constraints due to the hardware of the object comprise at least one of considering one or more joint limits of the object, considering one or more speed limits of the object, considering one or more energy consumption limits of the object, and considering reachability of the object.

13 . The method according to claim 11 , wherein

the one or more constraints due to the environment around the object comprise at least one of avoiding collision with one or more obstacles, considering one or more characteristics of a surface on which the object moves, and meeting one or more safety rules of an area within the object moves.

14 . The method according to claim 1 , wherein

the one or more constraints comprise at least one of one or more constant constraints and one or more dynamic constraints.

15 . The method according to claim 1 , wherein

the steps of the method are performed at repeating time steps.

16 . The method according to claim 1 , wherein

the object is a robot or a virtual avatar.

17 . A control unit for controlling motion of an object towards a target pose, wherein the control unit is configured to

determine a current state of the object;

randomize multiple configurations of the object with regard to the current state of the object;

predict for each configuration of the multiple configurations of the object a constraint cost with regard to each constraint of one or more constraints;

determine for each constraint of the one or more constraints, among the predicted constraints costs that have been predicted for the multiple configurations of the object, a predicted constraint cost that meets the constraint the worst;

predict a target configuration of the object using the multiple configurations of the object, the one or more determined constraint costs that have been determined for the one or more constraints, and the target pose; and

control the object to change its current state towards the predicted target configuration.

18 . A robot comprising a control unit, wherein the control unit is configured to control motion of the robot towards a target pose by

determining a current state of the robot;

randomizing multiple configurations of the robot with regard to the current state of the robot;

predicting for each configuration of the multiple configurations of the robot a constraint cost with regard to each constraint of one or more constraints;

determining for each constraint of the one or more constraints, among the predicted constraints costs that have been predicted for the multiple configurations of the robot, a predicted constraint cost that meets the constraint the worst;

predicting a target configuration of the robot using the multiple configurations of the robot, the one or more determined constraint costs that have been determined for the one or more constraints, and the target pose; and

controlling the robot to change its current state towards the predicted target configuration.

19 . A program comprising program-code means for executing a method for controlling motion of an object towards a target pose, when the program is executed on a computer or digital signal processor, wherein the method comprises:

determining a current state of the object;

randomizing multiple configurations of the object with regard to the current state of the object;

predicting for each configuration of the multiple configurations of the object a constraint cost with regard to each constraint of one or more constraints;

determining for each constraint of the one or more constraints, among the predicted constraints costs that have been predicted for the multiple configurations of the object, a predicted constraint cost that meets the constraint the worst;

predicting a target configuration of the object using the multiple configurations of the object, the one or more determined constraint costs that have been determined for the one or more constraints, and the target pose; and

controlling the object to change its current state towards the predicted target configuration.

20 . A non-transitory computer-readable storage medium embodying a program of machine-readable instructions executable by a digital processing apparatus, which cause the digital processing apparatus to perform a method for controlling motion of an object towards a target pose, wherein the method comprises

determining a current state of the object;

randomizing multiple configurations of the object with regard to the current state of the object;

predicting for each configuration of the multiple configurations of the object a constraint cost with regard to each constraint of one or more constraints;

determining for each constraint of the one or more constraints, among the predicted constraints costs that have been predicted for the multiple configurations of the object, a predicted constraint cost that meets the constraint the worst;

predicting a target configuration of the object using the multiple configurations of the object, the one or more determined constraint costs that have been determined for the one or more constraints, and the target pose; and

controlling the object to change its current state towards the predicted target configuration.

Assignments (2)

Continuity (1)

References Cited (14)