OMNI-DRL: Learning to Fly in Forests with Omnidirectional Images

Perception is crucial for drone obstacle avoidance in complex, static, and unstructured outdoor environments. However, most navigation solutions based on Deep Reinforcement Learning (DRL) use limited Field-Of-View (FOV) images as input. In this paper, we demonstrate that omnidirectional improve these methods. Thus, provide a comparative benchmark of several visual modalities navigation: ground truth depth, semantic segmentation, RGB images. These exhaustive comparisons reveal it superior to an camera navigate with classical DRL Finally, show two different virtual forest environments adapting the convolution take into account spherical distortions improves results even more.