Multi-Agent Learning for Resilient Distributed Control Systems