Multi-agent coordination by iterative learning control: Centralized and decentralized strategies

Iterative learning control (ILC), an approach to achieve perfect trajectory tracking for uncertain dynamic systems that are periodic or repetitive, can be viewed as a kind of coordination or planning algorithm. This paper exploits this view to provide two coordination algorithms for distributed multi-agent systems. First we show how to achieve formation control for a class of nonholonomic mobile agents though an iterative update of each agent's angular velocity along the trajectory. The algorithm required to achieve this result uses local measurements, but a centralized computation of the control input. Second, we show a decentralized coordination strategy for a set of simple first-order integrator dynamic systems. In this case the control updates are computed locally by each agent using only local information, yet through the iterative update process the group achieves the desired formation. Numerical simulations illustrate the results.