High fidelity Mars Rover simulation platform based on digital twin and machine learning integration: Innovation and application of ISMRS 

Abstract The Isaac Sim Martian Robotic Simulator (ISMRS) is a cutting-edge digital twin-based platform for Mars rover development, leveraging Nvidia's robotics simulator. It rapidly generates realistic Mars terrains, supports diverse rover capabilities, and integrates machine learning for environmental data synthesis. ISMRS is ROS2 compatible, enabling autonomous navigation and obstacle avoidance. The simulator's image recognition prowess is validated through real rock instance segmentation. A YOLOv8 model trained on synthetic data showed a 6% performance deficit compared to real data, which narrowed to a 15% accuracy improvement upon real data fine-tuning. This underscores ISMRS's high fidelity in simulating Mars environments for rover training and machine learning model development..