Data-driven Visual Servoing of Flexible Continuum Robots in Constrained Environments

Abstract

Flexible continuum robots operating in constrained and dynamic environments face significant challenges, especially when interacting with uncertain and potentially unknown conditions. Traditional model-based methods face significant difficulties due to the inherent nonlinearities and uncertainties in robot dynamics, as well as the complexities introduced by environmental interactions. This work presents a new data-driven, model-free control strategy for flexible continuum robots operating in constrained environments, leveraging Lie bracket approximations to achieve effective regulation. The method enables effective visual servoing without requiring explicit kinematic or dynamic models, making it highly adaptable to diverse scenarios where environmental constraints and robot deformation impact system performance. Additionally, it does not rely on initial state estimation, further enhancing its suitability for dynamic, uncertain environments. The effectiveness of the proposed method is validated through simulations and experiments, showing enhanced robustness and adaptability in real-time control scenarios.