Active Sensing by Unmanned Aircraft Systems in Realistic Communication Environments

This paper describes an information-theoretic framework for active sensing by unmanned aircraft systems in realistic communication environments. The position dependency of wireless networked communication is modeled using the standard empirical radio model in combination with Shannon capacity or a packet erasure channel. The covariance of the extended information filter is used to model sensed information content. The two information-theoretic concepts are then combined into a single optimization metric. Decentralized control laws are developed for simplified network scenarios. Stochastic approximation and extremum seeking control techniques are used to estimate the gradient of the sampled objective function. Experimental flight results demonstrate the capabilities of the active sensing framework.