With the increasing complexity of urban transportation systems, traffic congestion and safety issues have become key challenges in modern urbanisation. The rise of autonomous driving technology offers new possibilities for solving these problems, but there are still many technical bottlenecks in how to achieve efficient autonomous driving traffic control in complex traffic environments. This paper proposes a big data-driven multi-agent reinforcement learning framework that aims to improve the decision-making ability of autonomous vehicles in complex traffic environments through collaboration between multiple agents and real-time application of big data. We constructed a traffic flow prediction model based on big data and combined it with multi-agent reinforcement learning. The effectiveness of this method in urban traffic and highway scenarios was verified through experiments. The experimental results show that multi-agent reinforcement learning based on big data can significantly optimise traffic flow, reduce vehicle collisions, and improve overall traffic efficiency. The research in this paper provides a new solution for autonomous driving traffic control and demonstrates the potential of multi-agent reinforcement learning in intelligent transportation systems.