ACACT: Adaptive Collision Avoidance Algorithm Based on Estimated Collision Time for Swarm UAVs

• Published in: IEEE access 2023, Vol.11, p.120179-120191
• Main Authors: Min, Sewoong, Nam, Haewoon
• Format: Article
• Language: English
• Subjects: Adaptive algorithms; Algorithms; Amplification; Autonomous aerial vehicles; Barriers; Collision avoidance; collision avoidance of swarm UAVs; Collision dynamics; Contingency; Drones; Heuristic algorithms; Obstacle avoidance; Optimization; Path planning; path planning of swarm robots; Planning; Potential fields; Prediction algorithms; swarm robotics; Swarm UAVs; Travel time; Unmanned aerial vehicles; Vehicle dynamics
• ISSN: 2169-3536; 2169-3536
• DOI: 10.1109/ACCESS.2023.3327928

The Adaptive Collision Avoidance Algorithm Based on the Estimated Collision Time (ACACT) is proposed in this paper, representing a novel approach designed for effective and efficient collision avoidance and path planning in highly dynamic environments, notably those with swarm Unmanned Aerial Vehicles (UAVs). The fundamental challenge in swarm UAV operations revolves around dynamic collision avoidance and nimble path planning. Addressing this, the ACACT algorithm exhibits the capability of predicting imminent collisions by estimating their likely occurrence times and then adeptly adjusting the UAVs' trajectories in real-time. A significant facet of the algorithm is the employment of adaptive target velocity, updated in accordance with the predicted collision timelines. This ensures not only that UAVs can sidestep potential collisions but also that they can pursue more direct and efficient routes in comparison to conventional methodologies. Highlighting its superiority over existing techniques, the ACACT algorithm successfully resolves some long-standing issues linked with the Artificial Potential Field (APF) method, especially concerning unreachability and oscillation. This is accomplished by integrating a strategic contingency plan coupled with enhanced obstacle navigation, particularly in proximity to target locations. For a comprehensive evaluation of the algorithm's prowess in collision avoidance and path planning, a novel metric named the Path Traveling Time Ratio (PTTR) is introduced. PTTR assesses both the traveling time taken for a vehicle to reach its target position and the duration it spends within collision-prone zones. This metric offers a more advanced evaluation method than merely comparing path lengths, collision counts, or traveling times. Through rigorous experimentation, it is observed that the ACACT algorithm enhances collision avoidance and path planning by an impressive margin of up to 20% compared to its traditional counterparts. Furthermore, a distinct advantage of the ACACT is its ability to uniformly tackle obstacles, irrespective of their speeds and independent of PID gain variations. It not only boosts the safety parameters but also amplifies operational efficiency, setting new benchmarks for UAVs to reach their target points with swiftness and security.