Benchmark RadarMineXplore
MineXplore is an open-source MuJoCo-based benchmark for robot navigation in GNSS-denied underground mine environments, derived from the Leung et al. 2017 Chilean copper mine dataset. It reconstructs a 104,423 sq.m tunnel network with octagonal wall cross-sections, LiDAR-sourced jagged wall geometry, three terrain friction zones, a global 5 degree incline, and periodic spot lighting. The benchmark includes an evaluation protocol based on coverage percentage, with a 90% target for single-agent PPO policies.
- Released
- 2026-06-03
- Readiness
- Paper only
- Primary field
- Robotics & Autonomous Systems
Why it matters
MineXplore addresses the lack of realistic simulation benchmarks for underground mine navigation in the open-source ecosystem, providing a GPU-compatible environment grounded in real production-mine geometry. It enables reproducible evaluation of reinforcement learning policies under degraded sensing and complex topology, supporting progress in autonomous navigation for safety-critical applications.
Motivation
Underground mines present extreme conditions for autonomous robot navigation: GPS is denied, lighting is degraded, and tunnel topology is loop-rich and non-convex.
Primary resources
Benchmark Radar records only publicly supported details and links back to primary sources for verification.