Systems and methods for virtual environment for reinforcement learning in manufacturing

1. A method comprising:

receiving, by an interpreter logic of a computing device having a processor and addressable memory, a first state of a tool of a computer numerical control (CNC) machine;

determining, by the interpreter logic, a reward and a value of the reward based on the received first state, wherein the reward is at least one of: positive and negative;

transmitting, by the interpreter logic to an agent logic of the computing device, a set of information comprising the determined reward and the value of the reward and the received first state;

performing, by the agent logic, at least one action as part of generating a tool path, wherein the at least one action is limited by a policy comprising data for a set of possible actions available to the CNC machine, wherein the policy is at least one of: avoiding a collision and avoiding a part violation in manufacturing of the part;

creating, by the agent logic, a subsequent state to proceed to based on the performed at least one action, the subsequent state being a second state, and wherein the second state is combined with the first state;

determining, by the agent logic, a series of actions for the tool path until a final state is reached, based on the determined reward and value associated with the performed at least one action at each subsequent state;

training manufacturing machinery within a virtual environment based on reinforcement learning to determine the series of actions for the tool path; and

wherein each action of the series of actions is determined based on at least one of: changing the tool, moving the tool, turning a spindle on, and turning the spindle off.

2. The method of claim 1 , wherein the first state comprises information regarding at least one of: state of the cutting tool, state of the machine, and both state of the cutting tool and state of the machine making up an overall system.

3. The method of claim 2 , wherein the overall system is represented by at least one of: a cutting insert, a discrete part, a cutting tool, and a stock material.

4. The method of claim 1 , wherein the at least one action comprises at least one of: turning operations, milling operations, additive Directed Energy Deposition (DED), wire Electrical discharge machining (EDM), and an additive powder bed.

5. A computing device having a processor and memory, the computing device configured to:

receive a first state of a tool of a computer numerical control (CNC) machine;

determine a reward and a value of the reward based on the received first state, wherein the reward is at least one of: positive and negative;

perform at least one action as part of generating a tool path, wherein the at least one action is limited by a policy comprising data for a set of possible actions available to the CNC machine, wherein the policy is at least one of: avoiding a collision and avoiding a part violation in manufacturing of the part;

proceed to a subsequent state based on the performed at least one action, the subsequent state being a second state, wherein the second state is combined with the first state;

determine, a series of actions for the tool path until a final state is reached, based on the determined reward and value associated with the performed at least one action at each subsequent state;

train manufacturing machinery within a virtual environment based on reinforcement learning to determine the series of actions for the tool path; and

wherein each action of the series of actions is determined based on at least one of: changing the tool, moving the tool, turning a spindle on, and turning the spindle off.

6. The computing device of claim 5 , wherein the first state comprises information regarding at least one of: state of the cutting tool, state of the machine, and both state of the cutting tool and state of the machine making up an overall system.

7. The computing device of claim 6 , wherein the overall system is represented by at least one of: a cutting insert, a discrete part, a cutting tool, and a stock material.

8. The computing device of claim 5 , wherein the at least one action comprises at least one of: turning operations, milling operations, additive Directed Energy Deposition (DED), wire Electrical discharge machining (EDM), and an additive powder bed.

9. A computer-aided manufacturing (CAM) system of a computer numerical control (CNC) machine comprising a computing device, wherein the computing device comprising an interpreter logic, an agent logic, and a policy logic;

wherein the interpreter logic of the computing device is configured to:

receive a first state of a tool of the CNC machine;

determine a reward and a value of the reward based on the received first state, wherein the reward is at least one of: positive and negative;

transmit a set of information comprising the determined reward and the value of the reward to the agent logic of the computing device; and

wherein the agent logic of the computing device is configured to:

perform at least one action to generate a tool path, receive a policy from the policy logic, wherein the at least one action is limited by the policy and the policy comprises data for a set of possible actions available to the CNC machine, wherein the policy is at least one of: avoiding a collision and avoiding a part violation in manufacturing of the part;

proceed to a subsequent state based on the performed at least one action, the subsequent state being a second state, wherein the second state is combined with the first state;

determine a series of actions for the tool path until a final state is reached, based on the determined reward and value associated with the performed at least one action at each subsequent state;

train manufacturing machinery within a virtual environment based on reinforcement learning to determine the series of actions for the tool path; and

wherein each action of the series of actions is determined based on at least one of: changing the tool, moving the tool, turning a spindle on, and turning the spindle off.

10. The system of claim 9 , wherein a state of a tool of a computer numerical control (CNC) machine is a representation of a virtual environment.

11. The system of claim 10 , wherein the virtual environment comprises a target, a stock, and a cutting tool.

12. The system of claim 11 , wherein the final state is based on at least one of: no stock is left to be machined, no target is empty and thus to be added, and number of trials or iterations that have been performed is greater than a previously determined threshold.

13. The system of claim 9 , wherein the policy limits the at least one action based on a set of fixed rules.

14. The system of claim 9 , wherein the agent logic of the computing device if further configured to choose a subsequent action as a function of a history of a previously performed set of actions.

15. The system of claim 9 , wherein the virtual environment for reinforcement learning in manufacturing utilizes a continuous, non-discretized, contour-based environment.