Proportional integral derivative (PID) control is widely used in the fields of industrial and robot control. However, it faces challenges such as complex parameter setting, difficulty in accurately modeling the system, and sensitivity to changes in the controlled object. To address these challenges, this study proposes a hierarchical adaptive PID control algorithm based on a deep reinforcement learning algorithm, named TD3-PID, for the automatic control of mobile robots. In this algorithm, the upper-layer controller adjusts the parameters and output compensation of the lower-layer PID controller by observing the current environmental and system status in real time to compensate for errors in real time and optimize system performance. This study applies the proposed TD3-PID controller to a trajectory tracking task of a four-wheel mobile robot and conducts real-scenario experimental comparisons with other control methods. The results show that the TD3-PID controller exhibits superior dynamic response performance and anti-interference ability. The overall response error is significantly reduced and significant advantages are seen in improving the performance of the control system.