Robust nonlinear trailing control for multiple mobile autonomous agents formation.

摘要

Formation control of multiple autonomous agents has gained intense interest from the control community due to its importance in military and civilian applications. Leader/follower trailing control, which drives a follower to a prescribed relative position with respect to its leader, is one of the most fundamental building blocks of formation control.; The focus of this dissertation is on the design of a decentralized trailing control architecture for pairs of autonomous non-holonomic vehicles based on generic vehicle models. A fundamental constraint in trailing control requires that each agent employs local sensor information to process data on the relative position and velocity between its neighboring vehicles, possibly without relying on communication with mission control. This constraint poses a challenge to the design of the control system because the reference trajectory to be tracked may not be known a priori.; The lack of the leader reference trajectory can be overcome by resorting to the "internal model paradigm". It is shown how a controller can be designed by embedding a model of the autonomous dynamics of the leader in a robust stabilizer obtained by using nonlinear gain assignment techniques. A benefit of the proposed design is that each agent can be confined into specified "sectors" to avoid possible collision even during transients.; An important aspect in assessing the stability of a multi-layered formation is the analysis of the internal dynamics of intermediate leaders. Although this aspect is often neglected in the literature, it is shown that the internal dynamics need to be well behaved, not only at the desired formation in steady-state but also during transients, to prevent agents from erratic motion which may lead to potential collisions.; An important tool for the design of robust trailing control architecture is the nonlinear small-gain theorem. The use of saturated controller with tunable gains has been shown to be instrumental in building controllers robust with respect to small slow-varying measurement errors for arbitrary initial conditions, or even small time delay. Sufficient conditions guaranteeing asymptotic stability of closed-loop systems are given and practical guidelines on how to tune those control parameters are provided for implementation.