Delay estimation for wireless LAN control of nonlinear systems

Computationally intensive state estimation methods such as computer vision induce substantial delays when used in a wireless LAN control scenario. To facilitate the use of such estimation algorithms, the computation and communication delays can be compensated for. A method is presented for separating delays for a class of systems and delay models where a round-trip delay is known, but not the individual components. The solution involves traversal time computation using the system model in addition to the round-trip delay sums to form an under-determined system. Admissible delays are shown to be confined to one hyper-octant. Accurate delay reconstruction is achieved by applying projection onto convex sets. The implications are favorable for a robot motion control scenario where signals involving complex expressions are difficult to physically realize. In this situation, state estimates formed from the reconstructed delays enable simple control signals to achieve control objectives.