Gibbs Sampler Based Control of Autonomous Vehicle Swarms in the Presence of Sensor Errors

In earlier work of the authors[1], [2], [3], [4], [5], it was shown that Gibbs sampler based annealing algorithms could be used for vehicle swarms to achieve self-organization. Nevertheless, the earlier convergence analyses were based on the assumption that the Gibbs potential can be precisely evaluated. In practice, Gibbs potentials have to be measured via sensors, which usually introduce errors. The robustness of the stochastic algorithm under sensor errors is studied in this paper. Two types of sensor error, range-error and random error, are investigated. Analytical results on convergence are derived for sensor errors with limited support, and are further validated through simulations.