📊 In our preliminary experiment results, we benchmarked 7 state-of-the-art imitation learning and demo-driven reinforcement learning algorithms on all tasks, and show that existing algorithms are not yet capable of solving the challenging BiGym tasks.

🧵 We hereby introduce BiGym, a large-scale demo-driven benchmark for mobile bi-manual manipulation. It covers tasks from simple target reaching to complex kitchen cleaning. Each task includes 30+ demos, reflecting algorithm performance with real-world multi-modal data.

🧵 In robotics research, comprehensive benchmarks have greatly reduced the cost of iterating and developing novel algorithms. However, prior benchmarks either focus on pure RL with dense reward functions, or are restricted to tasks with planner generated unnatural trajectories.

🚀 Looking for a benchmark for bi-manual mobile manipulation with nicely collected demonstrations? We are excited to release BiGym, a new benchmark with human-collected demos! 🌐 Website: chernyadev.github.io/bigym/ 📄 Paper: arxiv.org/abs/2407.07788 💻 Code: github.com/chernyadev/bigym