A Flexible Genetic Algorithm System for Multi-UAV Surveillance: Algorithm and Flight Testing

Abstract

This paper discusses the development and testing of a flexible genetic algorithm (GA)-based system used for tasking a team of unmanned aerial vehicles (UAVs) to complete a coordinated surveillance mission. The GA development, laboratory testing of the GA to ensure convergence to a \good" solution, integration testing with two ground stations, and the field testing of the algorithms are explained. The algorithm was found to be robust and flexible enough to work in various settings with different UAV types and ground stations.

Heterogeneous Aerial Platform Adaptive Mission Planning Using Genetic Algorithms

“…The complexity of the models used in agent-based approaches offer no assurance of an optimal final result and thus are often avoided [17]. GA are well known optimizer that have been shown to successfully plan balanced missions for simultaneous mutli-UAV sets [18] surveying multiple POIs [2] and are particularly well-suited in UAV tasking and path planning [18][19][20][21][22]. Several GA approaches for coordinated tasking in surveillance missions under number of targets, task load requirements and reconnaissance time constraints exist in literature [23][24][25][26].…”

“…A flight tested GA method using homogeneous fixed wing UAVs was investigated et al [2] where the fitness function balanced UAV flight times, example shown in Fig. 1, was able to converge on a desktop PC in less than 5 min.…”

“…Existing GA systems operated assuming a homogeneous set of UAVs, meaning that the fitness function will need to be redesigned to address different UAVs. The existing GA mission planning effort by Darrah et al [2] was chosen as a basis for modifications that would include heterogeneous aircraft that undertake a search and deploy style mission. In order to build upon existing knowledge of GA mission planners, a basic operational concept needs to be developed that will highlight capabilities of fixed wing and multi-rotor UAVs.…”

“…Assignment of the UAVs to specific tasks will require an optimizer that is able to scale up beyond traditional deterministic methods [1]. Nondeterministic algorithms, such as the Genetic Algorithm (GA), has been shown to handle asset tasking optimization problems that are NP-hard and scale up to allow multiple vehicles [2]. The final deployment of the algorithm will allow an operator to choose POI locations and number of fixed wing aircraft used.…”

Collision Avoidance for Cooperative UAVs with Rolling Optimization Algorithm Based on Predictive State Space

“…Recently, with the advances of dynamics [1,2], navigation [3,4], and sensors [5,6], more civilian applications have been realized for UAVs, such as surveillance [7,8], aerial photography [9,10], search and rescue [11,12], and others. For all of these applications with multiple UAVs, one critical technical problem is that the UAVs may collide with each other in the increasingly dense airspace.…”