A Myopic Monte Carlo Strategy for the Partially Observable Travelling Salesman Problem
Andrew R. Buck, James M. Keller
2016 IEEE Congress on Evolutionary Computation (CEC), Vancouver, BC, 2016, pp. 632-639
WCCI, CEC, Pathfinding, Mental Map
Abstract
In this paper, we present two greedy, myopic algorithms for solving the partially observable travelling salesman problem. Although not optimal from a decision-theoretic viewpoint, these strategies are shown to perform reasonably well under the uncertain conditions of the environment. The first algorithm is a strictly greedy algorithm and has no tunable parameters, whereas the second algorithm uses Monte Carlo sampling to determine likely configurations of the environment and applies value iteration to pick an action. We present both approaches with illustrative examples and empirically demonstrate their relative strengths and weaknesses.
Media
A greedy algorithm for solving the partially observable travelling salesman problem (PO-TSP).
A myopic Monte Carlo (MMC) algorithm for solving the partially observable travelling salesman problem (PO-TSP).
Files
[paper]
[slides]