• A novel variable stiffness actuator based on a rocker-linked epicyclic gear train is proposed. • The simultaneous relocations of the pivot and torsion spring permit a wide-range stiffness adjustment. • The energy efficiency is improved by introducing a linear guide and a cam follower. • A dynamic single-neuron adaptive PID controller is presented for trajectory tracking. This paper presents a novel variable stiffness actuator based on a rocker-linked epicyclic gear train (REGT-VSA). The stiffness adjustment principle of the actuator is to change the lever ratio of force by converting the differential motion of the planetary gear train into the linear motion of the elastic element. The unique design of the rocker-linked epicyclic gear train ensures the compactness of structure and the accuracy of transmission. The simultaneous relocations of the pivot and torsion spring permit a wide-range stiffness adjustment. By introducing a linear guide and a cam follower, the internal friction is reduced, and the energy efficiency is improved. A prototype and its trajectory tracking controller are developed to evaluate the performance of REGT-VSA. An exponential function is used to train the neuron proportion coefficient of the conventional single-neuron adaptive proportional-integral-derivative (PID) controller to compensate for the position error in real time, thus creating the dynamic single-neuron adaptive PID controller. The stiffness identification and trajectory tracking experiments under different conditions prove the feasibility of the actuator and the proposed algorithm.