Browsing by Author "Mphogo, Dinorego"

Now showing 1 - 1 of 1
  • Thumbnail Image
    Item
    Cooperative collision avoidance for unmanned aerial vehicles
    (Stellenbosch : Stellenbosch University, 2020-03) Mphogo, Dinorego; Engelbrecht, J. A. A.; Stellenbosch University. Faculty of Engineering. Dept. of Electrical and Electronic Engineering.
    ENGLISH ABSTRACT: This thesis presents the design, implementation, and verification of a cooperative collision avoidance algorithms for unmanned aerial vehicles (UAVs) in multi-aircraft conflict scenarios. Two types of collision avoidance algorithms are developed and verified in simulation: a rules-based algorithm and a cooperative path planning based algorithm. The rules-based collision avoidance algorithm is modelled after the tactical Traffic Collision Avoidance System (TCAS) that is used on commercial passenger airliners. To enable multi-aircraft collision avoidance, two methods for combining the pairwise rulesbased collision avoidance actions are proposed, namely Resolution Action Superposition (RAS) and pairwise Closest-Intruder-First (CIF). The path planning based collision avoidance algorithm grows a search tree of admissible conflict resolution paths, and searches the tree to find the conflict-free path with the lowest cost. To enable cooperative collision avoidance, all aircraft communicate their current positions and intended flight paths to all other aircraft. A token allocation strategy is used so that the individual aircraft plan their new collision avoidance paths sequentially according to a predetermined priority order. The rules-based and path planning based collision avoidance algorithms were implemented and verified in simulation. A simulation environment was created to test both the rules-based and path planning based collision avoidance algorithms. Set-piece conflict avoidance scenarios were performed to produce illustrative results. The simulations illustrated that both rules-based and path planning based collision avoidance can resolve both pairwise and multi-aircraft conflicts. Furthermore, Monte Carlo simulations were performed to produce statistical results and evaluate the performance of both algorithms in random conflict scenarios. The simulation results show that both the rules-based and path planning based solutions are able to successfully resolve collision scenarios involving multiple unmanned aerial vehicles. The rules-bases solution requires less computational effort but does not optimise the collision avoidance plans. The path planning based solution requires much more computational effort, but provides optimal solutions that minimises the deviation from the original flights, and minimises the control effort of the avoidance actions.