Robust Feedback Control of Reconfigurable Multi-Agent Systems in Uncertain Adversarial Environments.

摘要

In this last period, our most exciting contributions pertain to algorithms for estimation and synchronization that are suitable to networked systems. Motivated by the estimation algorithms with both performance and robustness guarantees developed during the second year of performance, we proposed a hybrid observer that not only guarantees very fast performance but also total rejection of perturbations. The class of noise signals our observer allows is given by the family of piecewise-constant signals with changes satisfying a dwell-time condition. A design procedure for the observer is also provided for the estimation of the state of a single system, but preliminary results show that the algorithm can be extended to networked systems. we also generated estimation and synchronization algorithms for two agents that operate under intermittent information. We were able to devise algorithms that only require information about the (noisy) outputs of two systems at sporadic times to estimate their states and synchronize them (asymptotically). The algorithms are dynamical (in fact, hybrid) and have a state variable that is appropriately reset at the arrival of new information. The algorithms are also robust to small external perturbations, to skewed time clocks, and to infrequent packet drops