Event-Based Optical Flow for UAV Obstacle Avoidance: A Direction Selective Filter Approach

This paper presents an event-based obstacle detection method for unmanned aerial vehicles (UAVs) operating in high dynamic range (HDR) lighting warehouse conditions. The proposed approach integrates the concepts of Focus of Expansion (FOE) and Time-to-Contact (TTC) through a novel event-based optical flow estimation using direction selective filters. A new FOE method, termed Adaptive Spatial Clustering for FOE Estimation (ASC-FE), combines spatial clustering with adaptive filtering to precisely pinpoint the FOE, even in intricate environments. By clustering regions with similar flow characteristics and then weighting each potential FOE by cluster properties, ASC-FE offers a more accurate and noise-resistant FOE determination. In this proposed approach, the FOE provides insights into the UAV's direction of travel, thereby indicating the potential points of collision on its trajectory, and the TTC provides information about the relative velocity of points in the image with respect to the UAV, thus allowing the prediction of the time when an obstacle will be encountered if the current trajectory is maintained. Furthermore, in conjunction with a 3D Artificial Potential Field (APF) for obstacle avoidance, the proposed approach generates smooth, collision-free paths, addressing local minima issues through a rotational force component to ensure efficient UAV navigation. The effectiveness of this novel event-based obstacle detection and avoidance method is substantiated through simulations, demonstrating its potential for enhancing UAV navigation safety and efficiency in warehouse environments.