MCCoder: Streamlining Motion Control with LLM-Assisted Code Generation and Rigorous Verification

Introduction

MCCoder is an LLM-powered system designed to generate motion control code efficiently and safely. By integrating multitask decomposition, hybrid retrieval-augmented generation (RAG), and iterative self-correction, MCCoder enhances code accuracy using a structured workflow and a well-established motion library. It also features a 3D simulator for motion validation and logs full motion trajectories for verification.

MCCoder is publicly available at GitHub.

Folder

• MCEval/: Contains evaluation data programs and results.
• docs/: Includes soft-motion documentation, sample codes, and the MCEVAL dataset (WMX3API_MCEval_Evaluation_Dataset).

Installation

To install dependencies, run:

pip install -r requirements.txt

Usage

To launch the main program, run:

chainlit run app_MCEval.py -w

Running MCCoder requires Soft-Motion's coordination and a valid license. Please contact the author for further details.