AI-Driven Attention and Integer Programming for Optimal Motion Planning and Control in Autonomous Vehicles

AI-based motion planning in autonomous vehicles requires not only efficient path generation but also real-time adaptation under uncertainty, limited onboard resources, and stringent safety requirements. This paper proposes a hybrid framework that integrates Transformer-based attention mechanisms with Mixed Integer Programming (MIP) to enable intelligent, constraint-aware trajectory planning and control. The attention module learns to prioritize critical waypoints and generate informed initial solutions based on real-time traffic dynamics, vehicle states, and environmental context. These solutions are then globally optimized via MIP to ensure collision-free motion, optimal resource usage, and task fulfillment within safety and stability bounds. By combining AI perception with mathematical optimization, the proposed system demonstrates strong performance in fault tolerance, real-time scheduling, and safety-critical control, offering a viable solution for next-generation autonomous driving systems and intelligent transportation networks.