Event
MAPLE: A Training-Free Prediction Method for Indoor Exploration
1 reports · 1 sources
Kelvin Aladum et al. propose MAPLE, a training-free prediction method for indoor exploration that adapts the Mondrian process into an online generative prior based on geometric cues from the current partial map, used to predict occupancy in unknown space. On KTH floor plans and in real SLAM environments under various sensor ranges, MAPLE consistently outperforms non-predictive exploration methods and performs comparably to learning-based prediction methods. Experiments also show that more accurate map prediction does not necessarily lead to better exploration outcomes; how predictions are represented and used for exploration decisions is equally critical.
Report timeline
- MAPLE: A Training-Free Indoor Exploration Method Based on the Mondrian Process
An arXiv paper proposes MAPLE (Mondrian-Anchored Prior for Learning-free Indoor Exploration), a training-free predictive exploration method that adapts the Mondrian process into an online generative prior conditioned on geometric cues from the current partial map.
arXiv cs.RO ·