Integrated task assignment and path optimization for cooperating uninhabited aerial vehicles using genetic algorithms

• Published in: Computers & operations research Vol. 38; no. 1; pp. 340 - 356
• Main Authors: Edison, Eugene, Shima, Tal
• Format: Journal Article
• Language: English
• Published: Kidlington Elsevier Ltd 2011; Elsevier; Pergamon Press Inc
• Subjects: Applied sciences; Automobiles; Automotive engineering; Combinatorial analysis; Combinatorial optimization; Computer simulation; Dubins’ car; Exact sciences and technology; Flows in networks. Combinatorial problems; Genetic algorithms; Mathematical models; Monte Carlo simulation; Operational research and scientific management; Operational research. Management science; Optimization; Route optimization; Stochastic models; Studies; Task assignment; Tasks; Unmanned aerial vehicles; Vehicles; Dubins’ car; Unmanned aerial vehicles; Combinatorial optimization; Task assignment; Genetic algorithms; Autonomous system; Dubins' car; Monte Carlo method; Motor car; Navigation; Assignment problem; Multiple target; Path planning; Computational complexity; Modeling; Stochastic programming; Discretization; Genetic algorithm; Refining; Optimal trajectory; Kinematics; Motion control
• ISSN: 0305-0548; 1873-765X; 0305-0548
• DOI: 10.1016/j.cor.2010.06.001

The problem of integrating task assignment and planning paths for a group of cooperating uninhabited aerial vehicles, servicing multiple targets, is addressed. In the problem of interest the uninhabited aerial vehicles need to perform multiple consecutive tasks cooperatively on each ground target. A Dubins car model is used for motion planning, taking into account each vehicle's specific constraint of minimum turn radius. By using a finite set to define the visitation angle of a vehicle over a target we pose the integrated problem of task assignment and path optimization in the form of a graph. This new approach results in suboptimal trajectory assignments. Refining the visitation angle discretization allows for an improved solution. Due to the computational complexity of the resulting combinatorial optimization problem, we propose genetic algorithms for the stochastic search of the space of solutions. We distinguish between two cases of vehicle group composition: homogeneous, where all vehicles are identical; and heterogeneous, where the vehicles may have different operational capabilities and kinematic constraints. The performance of the genetic algorithms is demonstrated through sample runs and a Monte Carlo simulation study. Results show that the algorithms quickly provide good feasible solutions, and find the optimal solution for small sized problems.