AI‐Enhanced Simulation Platform for Robotics Education: From Kinematics to Reinforcement Learning
Haihui Dong, Hongguang Wang, Yanfang Song, Ganxi Luo, Bolin Li, Adham Manyara, Mucong Chi, Junda Huang, Lianhui LiABSTRACT
Undergraduate robotics education faces a widening gap between classical kinematic/dynamic theory and contemporary AI‐driven industrial practice. While simulation tools bridge theory and practice, most lack native support for data‐driven algorithm development. This paper presents an AI‐enhanced virtual‐real integrated simulation platform built on MATLAB/Simulink and deployed via VxWorks real‐time hardware‐in‐the‐loop. The platform integrates three AI modules—genetic‐algorithm/reinforcement‐learning trajectory optimization, neural‐network inverse kinematics, and an LLM coding advisor—within a coherent pedagogical workflow. A multi‐year deployment across 600+ students shows a 12.1% increase in complex‐task performance, reduced score variance, and significant gains in AI literacy. This work offers a replicable model for embedding AI competencies into foundational engineering courses.