Iterative learning control for multi-agent formation