Unmanned Aerial Vehicle(s) Trajectory Planning for Target Searching and Tracking
MetadataShow full item record
In this thesis, we introduce three trajectory planning algorithms for nonholonomic mobile sensors used for target tracking and the extended problem target searchings and tracking. These problems are motivated by the real-world application ice management, which includes searching for and tracking of icebergs. In addition to ice management, there are multiple other applications that can utilize similar problem formulations such as search and rescue, border patrol, traffic monitoring, environmental monitoring, combat scenarios, and wild animal tracking. Chapter 2 of this thesis is a review of the literature on mobile sensor networks for target searching and/or tracking problems. In the review, the focus is on the trajectory planning and target filtering algorithms. The most common algorithms for each task are presented. After Chapter 2, the mobile sensor is assumed to be a fixed-wing unmanned aerial vehicle, and the number of mobile sensors is one for Chapter 4 and 5 and fewer than the number of targets for Chapter 3. The contribution of Chapter 3 is a practical implementation of the target visitation algorithm, which is a combinatorial optimization formulation. This implementation is demonstrated both in simulation and a practical experiment. Compared to similar approaches it applies a static combinatorial formulation to a dynamic problem by making heuristic adjustments. In addition to demonstrate the result in a practical experiment. In Chapter 4, we use a cascaded formulation of an optimal control problem and show how to merge the equality constraints into the objective function. Then, we combine collocation and single shooting to get a implementation that can be more computationally effective than using collocation alone. The contribution of Chapter 5 is twofold. First, we derive a result that enable us to use set time constraints for how often a target must be revisited in the target searching and tracking problem. Second, we use this result combined with the two techniques from Chapter 3 and 4 to make a path planning algorithm. We demonstrate the performance of the algorithm compared to multiple base cases in simulations. Compared to other approaches this algorithm applies both combinatorial and optimal control formulations. The motivation for mixing the two techniques is that optimal control problems often cannot solve non-convex problems for global optimality. Furthermore, mixed integer linear programming can solve smaller instances of these problems, but it is difficult include nonlinear constraints in the formulation. Combining these approaches manages to utilize the strengths of both techniques.