Abstract This paper presents a novel virtual-target-based local reactive obstacle avoidance method for autonomous unmanned aerial vehicles (UAVs) that operate in unstructured and unknown environments. Unlike ground robots, UAVs face far more significant challenges during lowflight operations, mainly due to their higher speed, which necessitates faster reaction time and a more extraordinary ability to encounter rapid environmental changes. Furthermore, small UAVs like multi-copters often have performance constraints like limited take-off weight, computational resources, flight duration and on-board perception sensors. A virtual-target-based method is proposed in this research to enhance the autonomous navigation capabilities of small multi-copter UAVs while adhering to their operational constraints. The proposed method leverages high-resolution 3600 scanning LiDAR and GPS sensors to avoid unknown non-cooperative obstacles encountered during outdoor flight operations. The method guarantees the collisionless navigation of the low-flying UAV along the shortest possible path, avoiding obstacles that are hard to detect, like small tree trunks and narrow electric poles. The proposed reactive approach fulfils the constraint of fewer computational resources since the UAVs do not require prior knowledge of the exact map of the unstructured environment. Software-in-the-loop (SITL) simulation and outdoor flight tests conducted at a low altitude of 1.5 meters are used to validate the method, which demonstrates collisionless navigation along the shortest path towards the target.
Butt et al. (Mon,) studied this question.