Nonuniform Line Coverage from Noisy Scalar Measurements
We study the problem of distributed coverage control in a network of mobile agents arranged on a line. The goal is to design distributed dynamics for the agents to achieve optimal coverage positions with respect to a scalar density field that measures the relative importance of each point on the line. Unlike previous work, which has implicitly assumed the agents know this density field, we only assume that each agent can access noisy samples of the field at points close to its current location. We provide a simple randomized protocol wherein every agent samples the scalar field at three nearby points at each step and which guarantees convergence to the optimal positions. We further analyze the convergence time of this protocol and show that, under suitable assumptions, the squared distance to the optimal coverage configuration decays as $O(1/t)$ with the number of iterations $t$, where the constant scales polynomially with the number of agents $n$. We illustrate these results with simulations.