Abstract
In a general resource collection mission, it is important to plan optimal exploratory routes to maximize the total probability of finding resources. Very often, the resource distribution is given by a probabilistic landscape. This problem resembles a Selective Traveling Salesman Problem on a probabilistic map without well-defined routes or cities. In this thesis, we propose a new algorithm called the modified Elastic Net algorithm to provide a solution to this problem. Our algorithm incorporates the existing Elastic Net Algorithm with barrier method, which is able to produce optimal routes subject to node separation constraints. We then investigate the properties of this algorithm through simulations on several maps and conclude its dependency on the initial conditions, separation constraints and search range over the landscape. Next, we propose another algorithm - the Moving-Squares Algorithm, an approach based on simulated annealing, to serve as the benchmark to the modified Elastic Net. Finally, we carry out comparisons between the two algorithms and show that the modified Elastic Net outperforms the Moving-Squares Algorithm, proving its effectiveness in our problem.