This project will establish an inverse reinforcement learning (IRL) framework to learn the inter-robot interactions for multi-robot navigation, e.g., autonomous vehicles. While IRL has gained significant research interest, many developed works are limited to single-agent cases, making it difficult to be applied to the interactive environments in practical implementation and hence current solutions for multi-robot systems are either inefficient or too conservative. To deploy the multi-robot system in an efficient and non-conservative manner, an efficient IRL framework which effectively characterizes the multi-robot interaction must be developed to achieve efficient autonomous navigation.