Collision avoidance algorithm design for UAV base on parametric theorem and circle overlapping method / Nur Fadzilah Mohamad Radzi

• Main Author: Nur Fadzilah, Mohamad Radzi
• Format: Thesis
• Published: 2013
• Subjects: T Technology (General); TA Engineering (General). Civil engineering (General)
• Online Access: http://studentsrepo.um.edu.my/8648/4/Thesis_%2D_KGA_090073.pdf; http://studentsrepo.um.edu.my/8648/

This research presents a collision avoidance algorithm for resolving the conflict between two cooperative UAVs (Unmanned Aerial Vehicle) that fly in a straight trajectory with fixed altitude (2-horizontal dimension). However, it is not applicable to be used for obstacle conflicts. This proposed algorithm will modify the flight plan of hosting UAV (called as UAV in this thesis) to avoid the other UAV (called as target in this thesis) once the UAV’s collision avoidance system detects the collision in near future. Firstly, a mission waypoint path of UAV is given and assumes that all target information is collected from sensors (such as position sensors, angle heading sensor, and velocity sensors). Then, a collision detection algorithm is developed to calculate the potential of collision in future. The algorithm is based on PTCOT (Parametric Theorem and Circle Overlapping Test) between two UAVs. The collision potential will determine whether the avoidance mode needs to be activated or not and collision data (such as collision point, overlap time range, and collision time) will be used in conflict resolution. Secondly, a collision avoidance algorithm is designed to propose a new flight path in order to perform an avoidance maneuver for hosting UAV. In this research, two collision avoidance commands are proposed. First command is change position command, where the UAV will turn left or right based on relative collision angle between the conflicting agents. The second command is change speed command, where the speed of UAV will increase or decrease depends on the conflict situation. Finally, the collision avoidance algorithm is verified through MATLAB software. Various cases are tested to demonstrate the robustness of both collision detection and avoidance algorithms.