Ant Colony Optimization Theory Applied To UAV Tactical Control Problem

Since unmanned aerial vehicles (UAVs) are widely used in modern wars and the concept of UAVs joint combat develops which is suitable to the new revolution in military affairs, UAV tactical control (UAVTC) becomes a key problem in performing the swarm predominance of UAVs. UAVTC will shorten the decision making chain from payloads on the vehicle to tactical users, and the cross-organic, flat and network-based control of UAVs will be realized. UAVTC relates to not only public information standards, open architecture and the configurable functional components of mission planning and C2 (Command and Control), but also the cooperation and control of UAVs conducting tactical tasks in complicated battleground, which challenges the existing optimization theories and control methods.The dissertation researches some key problems of UAVTC based on ant colony optimization theory, including multiple UAVs cooperative mission planning, network routing in UAV intelligence dissemination and self-organization of UAVs. We improve the classic ant colony algorithm (ACA), extend ACA optimization mechanism to explore new methodologies to resolve the UAVTC, which is a problem of multi-objective optimization and multi-agent cooperative control. The main work and contributions of this dissertation are as follow:(1) A reinforcement learning ant colony algorithm (RLACA) is proposed, a pre-numbered strategy and a team-matching strategy are presented to simplify the complexity of multi-ant-colony optimization (MACO) in the dissertation. A reinforcement learning mechanism (RLM) for pheromone updating is presented to solve the problem of ant colony information utilized insufficiently in some other improved ACAs. Simulation results show that the RLACA, in which RLM is introduced, converges faster to a better resolution than some other algorithms. MACO is also discussed, and a framework of multi-difference-ant-colony algorithm (MDACA) is presented, the complexity is analyzed. The Pre-numbered strategy and the team-matching strategy are proposed to solve the "combination of sub-problems" in MDACA, so that the exponential complexity of computation declines to determined scale.(2) The application of ACA in multi-UAV cooperative mission planning is researched. In order to decrease the complexity of multi-UAV cooperative mission planning, the hierarchical and iterative strategy is applied. And then the problem is decoupled and decomposed into two coherent sub-problems, i.e. task allocation and vehicle route planning. Afterwards an extended mathematical model of task allocation based on the CMTAP is established, and an ACA on the basis of job-division mechanism is designed according to MDACA. A pre-numbered strategy is introduced into the algorithm to decrease the computational complexity. The simulation results show that the multi-tasks allocation problem under complex constraints can be solved effectively using the algorithm. Probabilistic road map (PRM) is introduced to describe the topology of battlefield, which can decrease the complexity of battlefield in vehicle route planning. Then another ACA, which introduce the reinforcement learning mechanism to enhance the search efficiency, is designed to plan the routes based on the topology description. The simulation results demonstrate that the algorithm can converge to a good feasible solution rapidly.(3) The application of ACA in UAV intelligence dissemination routing is researched. To satisfy tactical users (TUs) mostly and lighten the burden of the net, a unicast routing mathematical model, which maximizes the degree of TUs' satisfaction, and a multicast routing mathematical model, which minimizes the communicational cost, is established. Then a new ACA for intelligence dissemination unicast routing is designed based on RLACA. The state transition rule is improved and a new mechanism for local pheromone updating is designed in the algorithm to get a good fesaible route rapidly. The simulation results show that the algorithm can converge to a nearly optimal route quickly. A team-matching ant colony algorithm (TMACA) is designed based on MDACA, which introduce the team-matching strategy to overcome the prematurity of colony. TMACA adopts a wide-range pheromone updating mechanism based on the structure of multicast tree. Thus the algorithm can quickly converge to a feasible solution and keep the global search capacity. The simulation results show that TMACA can escape from the local optimum and then converge to the global best solution rapidly, and its performance will not be tempestuously influenced as the network size increases.(4) An ant colony algorithm for UAVs self-organization in the uncertain environment is designed. A distributed raid-pattern ant colony algorithm (DRPACA) on the basis of the raid behavior of ant-soldiers is designed according to the specialties of UAV. DRPACA will enhance UAVs' adaptive ability to the uncertain environment. To keep the stability and viability of the system, DRPACA takes a distributed, decentralized architecture. And there is indirect communication based on pheromone and direct communication via data link between ants, so that ants can effectively sense the environment and the task progress. A new mechanism of pheromone updating is presented, according to the dynamic environment and value-restricted targets. A state prediction model based on the distribution of pheromone is proposed to decrease the influence of delay and packet losses in the direct communication. With the mechanism, DRPACA can effectively perform different tasks. The simulation results demonstrate that the method can effectively coordinate the UAVs conducting reconnaissance and attack tasks over mission-area, and be adaptive to the dynamic environment and tasks.