Real-time Obstacle Avoidance for Unmanned Aerial Vehicles (UAVs)

Real-time obstacle avoidance is a vital component for unmanned aerial vehicles (UAVs) when autonomously following mobile ground vehicles (MGVs) in unstructured and dynamic environments. The Artificial Potential Field (APF) technique is a widely used method for UAV path planning due to its simplicity, ease of use and its inherent efficiency in obstacle avoidance. However, this technique has many shortcomings with obstacles such as falling in local minima, tendency to produce longer planned paths when avoiding obstacles, high probability of colliding with symmetrical obstacles, and increase in oscillatory movements near obstacles. To overcome these drawbacks, this paper presents a novel two-dimensional path planning technique for obstacle avoidance based on the APF method. The proposed technique produces velocity waypoints for the UAV's planned path based on GPS data and use of basic distance sensors for obstacle avoidance; therefore, it can be deployed on most types of UAVs utilizing flight controllers with autopilots such as PX4 and Ardupilot. The performance of the proposed technique is validated for different simulation scenarios in ROS and Gazebo supported PX4-SITL. The results show the suitability of the proposed technique for real-time obstacle avoidance for UAVs autonomously following MGVs in dynamic environments with different types of obstacles.