dongzhangim / t01-tool-path-planning-on-3-axis-milling-machine-using-drl

Tool-path-planning-on-3-axis-milling-machine-using-DRL

Objective

This project attempts to use a classic DRL algorithm, i.e. DQN to finish tool path plnning task on 3-axis milling machine automatically.

Implementation

• To realize quantitative analysis, the 3d models need to be voxelized and read as 3d-arrays, in which binvox tools are applied.
• The 3d-arrays which take geometric information of original models are then used to built up the simulation scenario.
• DQN algorithm is then used to train the agent (cutter) to interact with the environment (workspace on the machine tool) and finish the task (processing of the material).
• The trained strategies are finally tested and evaluated.

Usage

• To train or test the neural networks in scenarios in different size, some parameters need to be modified in the scripts.
• A simple GUI system may be added in future versions.

Demonstration

According to the demos, the well trained processing strategies will have good performance and generalization ability.