Abstract: In the 21st century, with the rapid development of computing and sensing technology, autonomous driving has become a hot and important research topic. The vast market for bicycles has created numerous opportunities for driverless bikes. An unmanned bicycle robot has the characteristics of flexible movement and narrow body, thus it can be widely used in disaster area-rescue operations, entertainment performances, and transportation scenes. Therefore, several scholars have studied and focused on this type of bicycle. For the lateral self-balancing problem of bicycle robots, a new balance control method has been studied for a class of bicycle robots that are equipped with an angular momentum wheel. The kinematics constraint of the robot balance control is constructed based on the lateral balance condition of the bicycle robot, and the balance constraint is regarded as the control target. Based on the Udwadia–Kalaba (U–K) theory, a torque analytical model satisfying the lateral balance of the robot was established, and a balance constraint following the controller based on the model was designed. The findings show that the proposed control method can achieve the lateral balance of the bicycle robot and overcome the disturbance caused by the initial deviation of the lateral roll angle θ. Through the calculation of the balance torque model, the bicycle robot is actively balanced. Compared with the traditional PD feedback control method, the control method based on the model design has the characteristics and advantages of fast system response, low overshoot, and ease of optimization of the control torque. The proposed control method is simulated and confirmed using MATLAB, and lateral self-balancing control of the bicycle robot is achieved at the initial roll angular velocities of 0, 1, 2, and 5°·s⁻¹. The simulation results confirm the stability and effectiveness of the control system. This study proposes a novel idea for the balance control of unmanned bicycle robots.