Autonomous ExplorationExpert
GPU-Accelerated Path-Dependent Marginal Information Gain for Autonomous Exploration
This preprint introduces a GPU-accelerated method for path-dependent marginal information gain in autonomous exploration, using per-node depth buffers instead of voxel hash maps. It stays within 5–10% of exact marginal gain while accelerating evaluation up to 118× on a desktop GPU and 28× on a Jetson Orin NX, and in real UAV tests it cut time to 95% coverage by 30%.