A processing machine is manipulated by a sequence of tool paths. The sequence includes individual processing paths that form material and detour paths that connect processing paths. Detour path generation is a problematic task similar to individual processing path generation when processing machines are applied for high-mix low-volume production.
This paper reports a detour path generation technique using deep reinforcement learning. A detour path is automatically derived from a pair of processing paths using the proposed technique. A test using an actual machine showed that the machine was appropriately manipulated by the derived detour paths.
