Abstract
In recent years, the advancement of imitation learning has led to increasedinterest in teleoperating low-cost manipulators to collect demonstration data.However, most existing systems rely on unilateral control, which only transmitstarget position values. While this approach is easy to implement and suitablefor slow, non-contact tasks, it struggles with fast or contact-rich operationsdue to the absence of force feedback. This work demonstrates that fastteleoperation with force feedback is feasible even with force-sensorless,low-cost manipulators by leveraging 4-channel bilateral control. Based onaccurately identified manipulator dynamics, our method integrates nonlinearterms compensation, velocity and external force estimation, and variable gaincorresponding to inertial variation. Furthermore, using data collected by4-channel bilateral control, we show that incorporating force information intoboth the input and output of learned policies improves performance in imitationlearning. These results highlight the practical effectiveness of our system forhigh-fidelity teleoperation and data collection on affordable hardware.