Interpretable UAV Collision Avoidance using Deep Reinforcement Learning

TL;DR

This paper introduces an interpretable deep reinforcement learning algorithm with self-attention for UAV collision avoidance, demonstrating robustness across diverse weather and environments.

Contribution

It combines self-attention mechanisms with deep reinforcement learning to enhance interpretability and robustness in UAV collision avoidance tasks.

Findings

Robust performance in varied weather conditions

Enhanced interpretability of the collision avoidance algorithm

Outperforms conventional deep reinforcement learning methods

Abstract

The significant components of any successful autonomous flight system are task completion and collision avoidance. Most deep learning algorithms successfully execute these aspects under the environment and conditions they are trained. However, they fail when subjected to novel environments. This paper presents an autonomous multi-rotor flight algorithm, using Deep Reinforcement Learning augmented with Self-Attention Models, that can effectively reason when subjected to varying inputs. In addition to their reasoning ability, they are also interpretable, enabling it to be used under real-world conditions. We have tested our algorithm under different weather conditions and environments and found it robust compared to conventional Deep Reinforcement Learning algorithms.