Scalable Cooperative Tracking of Moving RF Ground Targets
Investigators: Nisar Ahmed (PI) and Eric W. Frew
Sponsor: Center for Unmanned Aircraft Systems
This work will develop a new approach to decentralized sensor fusion and trajectory optimization to enable multiple networked UAS assets to cooperatively localize moving RF signal sources on the ground in the presence of uncertainties in ownship states and sensing models. Our approach ties together model predictive planning with the recently developed idea of factorized distributed data fusion, which allows each tracker vehicle to ignore state uncertainties for other vehicles and absorb new target state and local model information without sacrificing overall estimation performance. This approach will significantly reduce communication and computational overhead, and allow vehicles to maintain statistical consistency as well as accurately predict expected local information gains to efficiently devise receding horizon tracking trajectories, even in large ad hoc networks. This effort will have two key focal points: (i) efficient filtering and factored fusion algorithms for accurately capturing nonlinear dependencies between ownship pose-attitude errors, unknown RF sensing model parameters and target states; and (ii) online model predictive trajectory optimization algorithms that account for expected information gains while incorporating constraints on communication and vehicle performance.
Existing ground target tracking methods for UAS typically ignore ownship/sensor model uncertainties, e.g. due to pose/attitude estimation errors or sensor gimbal biases. This often leads to poor results for cooperative tracking, and is exacerbated for highly nonlinear sensing models required for RF localization. Other work has tackled this issue by estimating the joint set of dynamical states and uncertain model parameters of all tracker vehicles and targets, although the optimal decentralized implementation of this approach is not computationally scalable and is generally limited to loop-free/tree network topologies to ensure estimation consistency. Finally, the challenging problem of optimizing vehicle trajectories for cooperative autonomous target tracking with ownship and sensor model uncertainties has to date only been partially resolved under special circumstances for cooperative unmanned aerial networks. This work will address these issues under fairly general conditions in a novel manner that is both theoretically and practically sound.