This document summarizes Maja Mataric's research on using communication to address problems in multi-robot learning. It discusses two key problems - hidden state where robots can't sense all information, and credit assignment where reward must be divided among agents. The research demonstrates communication can help by allowing robots to share sensory data to overcome hidden state, and share reinforcement signals to overcome credit assignment. Two experiments are described: box pushing by 2 robots, and social rule learning by 4 robots. In both cases, local broadcast communication helped the robots learn tasks more effectively by addressing the two problems.