Cross-Modal Reinforcement Learning for Navigation with Degraded Depth Measurements
This addresses navigation robustness for robots in adverse conditions, but it is incremental as it builds on existing cross-modal and reinforcement learning methods.
The paper tackles the problem of robust navigation in unstructured environments when depth sensors degrade, by developing a cross-modal learning framework that uses depth and grayscale images to infer depth-relevant features, resulting in maintained performance under significant degradation and successful real-world transfer.
This paper presents a cross-modal learning framework that exploits complementary information from depth and grayscale images for robust navigation. We introduce a Cross-Modal Wasserstein Autoencoder that learns shared latent representations by enforcing cross-modal consistency, enabling the system to infer depth-relevant features from grayscale observations when depth measurements are corrupted. The learned representations are integrated with a Reinforcement Learning-based policy for collision-free navigation in unstructured environments when depth sensors experience degradation due to adverse conditions such as poor lighting or reflective surfaces. Simulation and real-world experiments demonstrate that our approach maintains robust performance under significant depth degradation and successfully transfers to real environments.