Learning coverage control of mobile sensing agents in one-dimensional stochastic environments

This technical note presents learning coverage control of mobile sensing agents without a priori statistical information regarding random signal locations in a one-dimensional space. In particular, the proposed algorithm controls the usage probability of each agent in a network while simultaneously satisfying an overall network formation topology. The proposed control algorithm is rather direct, not involving any identification of an unknown probability density function associated to random signal locations. Our approach builds on diffeomorphic function learning with kernels. The almost sure convergence properties of the proposed control algorithm are analyzed using the ODE approach. Numerical simulations for different scenarios demonstrate the effectiveness of the proposed approach.