| Control of networked multi-vehicle systems designed to perform complex coordinated tasks is currently an important and challenging field of research. This dissertation addresses a cooperative search problem where a team of Unmanned Air Vehicles (UAVs) seek to find targets of interest in an uncertain and adversary environment. We present a practical framework for on-line planning and control of a group of UAVs to do the cooperative search, which is based on two inter-dependent tasks: (i) using "cognitive maps" as the environment representation and incrementally learning it through new sensor readings; and (ii) based on the obtained information, planning a guidance path for the vehicle to follow. This dissertation investigates both the cognitive map building problem and the cooperative path planning problem. We present two evidential methods to incrementally build cognitive maps for directing UAVs to perform cooperative search tasks, where the sensor reading is represented probabilistically and is fused into the maps using evidential rules. By utilizing the generated cognitive maps, a finite horizon optimal control problem is formulated to generate physical feasible paths for the vehicles to follow, where certain constraints on arial vehicles including physical maneuverability, fuel consumption, etc, are considered and the dynamic nature of the environment is taken into account. The emergent coordination among UAVs is enabled by letting each UAV to consider other vehicles' actions into its path planning decision process. The effectiveness of the proposed multi-vehicle cooperative search scheme is illustrated via simulations in various dynamic, uncertainty, and risky environment. |
