Reinforcement learning for contact-rich tasks in automation systems
Systems and methods for controlling robots including industrial robots. A method includes executing ( 402 ) a program ( 550 ) to control a robot ( 102 ) by the robot control system ( 120, 500 ). The method includes receiving ( 404 ) robot state information ( 554 ). The method includes receiving ( 406 ) force torque feedback ( 556 ) inputs from a sensor ( 554 ) on the robot ( 102 ). The method includes producing ( 410 ) a robot control command for the robot ( 102 ) based on the robot state information ( 554 ) and the force torque feedback ( 556 ) inputs. The method includes controlling ( 412 ) the robot ( 102 ) using the robot control command.
1. A process performed by a robot control system, comprising:
executing a program to control a robot by the robot control system;
receiving robot state information by the robot control system;
receiving force torque feedback inputs from a sensor on the robot by the robot control system;
producing a robot control command for the robot using a guided policy search process, by the robot control system, based on the robot state information and the force torque feedback inputs and on a reference signal of a neural network, wherein the reference signal is based on the robot state information and the force torque feedback inputs, the force torque feedback inputs being added to a next-to-last layer of the neural network at force torque information nodes; and
controlling the robot using the robot control command, by the robot control system.
2. The process of claim 1 , further comprising filtering the force-torque feedback inputs using a low-pass filter.
3. The process of claim 1 , wherein an admittance controller is used to process the robot state information and the force torque feedback inputs.
4. The process of claim 1 , wherein producing the robot control command is performed using a mirror descent guided policy search process.
5. The process of claim 1 , wherein the sensor is a force/torque sensor mounted on a wrist of an arm of the robot.
6. The process of claim 1 , wherein the robot control system is a reinforcement learning control system.
7. A robot control system comprising:
a memory; and
a processor in communication with the memory, wherein processor is configured to perform a process as in claim 1 .
8. A non-transitory computer-readable medium storing executable instructions that, when executed, cause robot control system to perform a process as claim 1 .