Sensor location through linear programming with triangle inequality constraints

Interest in dense sensor networks due to falling price and reduced size has motivated research in sensor location in recent years. While many algorithms can be found in the literature, no benchmark exists and most papers fail to compare their results to other competing algorithms. To our knowledge, the algorithm which achieves the best performance in sensor location uses semidefinite relaxation of a quadratic program to solve for sensor location. We propose solving the same program, however without relaxing the constraints, but rather transforming them into linear triangle inequality constraints. Our linear program ensures a tighter solution to the problem. We benchmark ours against the competing algorithm, and provide extensive experimentation to substantiate the robustness of our algorithm even in the presence of high levels of noise.