To address the trajectory tracking and self-balancing control challenges in unmanned motorcycle systems, this paper proposes a dual-loop control strategy based on cascade PID control, integrated with a Multi-subgroup Improved Marine Predator Algorithm (MS-MPA) for parameter optimization to enhance control performance. The outer-loop control utilizes position error and yaw angle error as inputs, outputting the target roll angle through PID regulation, while the inner-loop adjusts handlebar steering torque based on roll angle deviation to achieve dynamic balance and trajectory tracking. Additionally, an independent speed closed-loop PID controller is designed to ensure stable operation under various working conditions.To overcome the complexity of traditional PID parameter tuning, the algorithm is employed to globally optimize nine control parameters. The fitness function comprehensively considers path tracking error, maximum deviation penalty, and instability penalty, effectively improving convergence speed and avoiding local optima. Simulation experiments conducted in Simulink using a circular trajectory test scenario demonstrate that the optimized control parameters reduce the maximum tracking error to 2.0945 m with a steady-state error of only 0.17 m, significantly enhancing tracking accuracy and stability. This study provides a viable solution for balance control and trajectory tracking in unmanned motorcycle systems, validating the applicability of intelligent optimization algorithms in complex control systems.