Learning Long-Horizon Robot Exploration Strategies for Multi-object Search in Continuous Action Spaces

Recent advances in vision-based navigation and exploration have shown impressive capabilities photorealistic indoor environments. However, these methods still struggle with long-horizon tasks require large amounts of data to generalize unseen In this work, we present a novel reinforcement learning approach for multi-object search that combines short-term long-term reasoning single model while avoiding the complexities arising from hierarchical structures. contrast existing act granular discrete action spaces, our achieves exceptional performance continuous spaces. We perform extensive experiments show it generalizes apartment environments limited data. Furthermore, demonstrate zero-shot transfer learned policies an office environment real world experiments.