Automated personalized feedback for interactive learning applications
A robot-training system permits a user touch, click on or otherwise select items from a display projected in the actual workspace in order to define task goals and constraints for the robot. A planning procedure responds to task definitions and constraints, and creates a sequence of robot instructions implementing the defined tasks.
1. A method of training a robot situated in a physical workspace, the method comprising the steps of:
(a) computationally representing, by a controller, the robot in a 3D coordinate space corresponding to at least a portion of the workspace;
(b) electronically detecting, by the controller, a first user gesture associated with the workspace and corresponding to a task goal, wherein the task goal includes the robot manipulating a physical object;
(c) electronically detecting, by the controller, a second user gesture associated with the workspace and corresponding to a task constraint;
(d) computing, by the controller, a motion plan for the robot based on the task goal and the task constraint; and
(e) causing, by the controller, the robot to execute control functions corresponding to the task goal without violating the task constraint in accordance with the computed motion plan.
2. The method of claim 1 , further comprising the step of analyzing a recorded image of the workspace to computationally define at least one implicit constraint associated with the task goal.
3. The method of claim 1 , further comprising the step of detecting user selection of an item within the workspace upon which the robot control function is to operate, the robot executing the control function on the selected item.
4. The method of claim 1 , wherein the task goal is a manipulation goal.
5. The method of claim 4 , wherein the manipulation goal comprises an object pick goal and an object place goal including a placement location.
6. The method of claim 5 , wherein the task constraint comprises an obstacle between an initial object location and the object placement location.
7. The method of claim 1 , wherein the task goal is an end-effector goal.
8. The method of claim 1 , wherein the task goal is an object-relative goal.
9. The method of claim 1 , wherein the task goal is a volume/area goal.
10. For use with a robot in a physical workspace, a robot controller comprising:
a processor;
a memory for storing instructions executable by the processor to provide:
perception and interaction modules configured for (i) computationally representing the robot in a 3D coordinate space corresponding to at least a portion of the workspace, (ii) detecting user gestures associated with the workspace and indicating a task goal and a task constraint, wherein the task goal includes the robot manipulating a physical object; and
a planning module configured for (i) computing a motion plan for the robot based on the task goal and the task constraint and (ii) causing the robot to execute control functions corresponding to the task goal without violating the task constraint in accordance with the computed motion plan.
11. The robot controller of claim 10 , wherein the robot controller is within the robot.
12. The robot controller of claim 10 , wherein the controller is external to the robot and communicates therewith via a network.
13. The robot controller of claim 10 , wherein the perception and interaction modules are further configured to analyze a recorded image of the workspace to computationally define at least one implicit constraint associated with the task goal.
14. The robot controller of claim 10 , wherein the perception and interaction modules are further configured to detect user selection of an item within the workspace upon which the robot control function is to operate, the robot controller being configured to responsively cause the robot to execute the control function on the selected item.
15. The robot controller of claim 10 , wherein the task goal is a manipulation goal.
16. The robot controller of claim 10 , wherein the manipulation goal comprises an object pick goal and an object place goal including a placement location.
17. The robot controller of claim 16 , wherein the task constraint comprises an obstacle between an initial object location and the object placement location.
18. The robot controller of claim 10 , wherein the task goal is an end-effector goal.
19. The robot controller of claim 10 , wherein the task goal is an object-relative goal.
20. The robot controller of claim 10 , wherein the task goal is a volume/area goal.