SIMPLEΒΆ

SIMPLE stands for SIMulation-based Policy Learning and Evaluation

Our goal is to build a simulation platform for policy learning and evalutions, featuring

  • Diverse Built-in Tasks

    • Provide a wide range of environments for data collection, imitation learning, and evaluation.

    • Cover tasks from low-level motor control to high-level decision making, enabling cross-domain generalization.

  • Evaluation-Oriented Architecture

    • Modular design for quickly creating real-to-sim evaluation environments.

    • Support for diverse robotic embodiments (manipulation, locomotion, multi-agent, etc.).

    • Focus on evaluation first to ensure fairness and replicability before optimization.

  • Exhaustive Benchmarking Metrics

    • Standardize evaluation protocols with clear, reproducible metrics.

    • Support comparisons across state-of-the-art methods.

    • Enable community-driven leaderboards for transparent progress tracking.

    • Emphasis policy performance alignment between Real and Sim.