Continuum manipulators are a new generation of robotic systems that possess infinite number of degrees of freedom associated with inherent compliance, unlike traditional robotic manipulators which consist of a finite number of rigid links. Because of this characteristic, controlling continuum manipulators is more complicated and difficult based on only traditional control theory. Soft computing techniques are solid alternative for improving the control performance of such kinds of robots. In this paper, we employ two types of neural dynamic approaches, i.e., gradient neural dynamics and zeroing neural dynamics, to solve the real-time Jacobian matrix pseudo-inversion problem, thereby achieving model-based kinematic control of multi-section continuum manipulators. Different kinds of neural dynamic models are investigated and their performances in terms of tracking accuracy are shown with and without noise disturbances. Simulation validations with a two-section and a three-section continuum manipulator demonstrate the feasibility and robustness of the proposed models.

Continuum机械手是新一代的机器人系统，拥有与固有顺应性相关联的无限多个自由度，这与由有限数量的刚性连杆组成的传统机械手不同。由于这种特性，仅基于传统控制理论，控制连续统操纵器变得更加复杂和困难。软计算技术是提高此类机器人的控制性能的可靠替代方案。在本文中，我们采用两种神经动力学方法，即梯度神经动力学和归零神经动力学，来解决实时雅可比矩阵伪逆问题，从而实现基于模型的多段式连续体机械手运动学控制。研究了不同种类的神经动力学模型，并显示了在有无噪声干扰的情况下其在跟踪精度方面的性能。用两部分和三部分连续介质操纵器进行的仿真验证证明了所提出模型的可行性和鲁棒性。