Unmanned Aerial Vehicles (UAVs), commonly known as drones, have many uses in aerial photography, precision agriculture, disaster response, and of course military applications. Indeed, drones are taking on tasks that demand even greater agility, precision, and autonomy than before. However, a persistent problem that limits their broader deployment is how to navigate safely through environments crowded with buildings and trees, for instance, and moving hazards such as vehicles, people, or other drones. Current systems often require constant human oversight, reducing efficiency and increasing operational risk.
Research in the International Journal of Wireless and Mobile Computing could offer a solution. The approach provides a significant upgrade to a widely used navigation method called the Artificial Potential Field (APF) algorithm. APF imagines a drone as being pulled towards its target while being pushed away from obstacles. Its appeal lies in its simplicity and its real-time responsiveness. Unfortunately, there are some inherent flaws in conventional APF. Drones following the APF can become trapped in local minima, points where opposing forces cancel each other out so that the drone never reaches its target or fails to navigate around moving objects. This obviously limits operational reliability.
The new research addresses these limitations with four complementary innovations. First, an adaptive repulsive potential function adjusts the balance between attraction and obstacle avoidance in real time, ensuring the drone continues toward its destination. Secondly, randomized directional perturbations help the drone escape local minima by introducing brief, controlled deviations. Thirdly, real-time collision risk prediction allows the drone to manoeuvre proactively, slowing, steering, or recovering after avoiding moving obstacles. Finally, fuzzy logic rules are used to optimize safety distances and avoidance speeds without requiring heavy computation.
The researchers carried out simulations and achieved a 96.3% success rate in environments with static and dynamic obstacles. Their approach generated smoother, shorter, and faster-computing trajectories than conventional APF methods. This could allow drones to be deployed in disaster zones, around collapsed buildings or in dense forests more reliably than before, improving search-and-rescue operations. Agricultural drones could fly efficiently between crop rows, avoiding obstacles without human intervention. The same technology could open up new possibilities for delivery drones that could move safely among traffic and pedestrians.
Li, H. and Duan, X. (2025) ‘Research on adaptive artificial potential field obstacle avoidance technology for unmanned aerial vehicles in complex environments’, Int. J. Wireless and Mobile Computing, Vol. 29, No. 5, pp.1–12.
No comments:
Post a Comment