Model and Collision Avoidance for Motion Planning of Rigid-Soft Robot With Continuous Expression and Input Mapping

This paper presents a motion planning approach for rigid-soft hybrid robots that incorporates continuous expression and input mapping, aiming to achieve both tracking and collision avoidance. The conventional piecewise constant curvature model utilized in soft robots suffers from issues of discontinuity and singularity. These inherent drawbacks pose significant challenges when attempting to address the nonlinear problems associated with motion planning. To overcome these limitations, we establish a relationship between the length of drive cable and the coordination of robot terminal. Through the application of Taylor transformation, we effectively eliminate the control variation present in the denominator of this expression. Then, the nonlinear model predictive control (NMPC) with this model expression and input mapping is proposed for motion planning. By directly constructing the current input and output of the motion planning based on the combination of previous data and substituting the linearized part of the model with this data, we mitigate the impact of model inaccuracies on the solution process. Both the obstacles and robot are modeled as polyhedra, and the collision problem is reformulated in the form of continuous nonlinear inequalities. These expressions are then incorporated into the NMPC optimization function as collision avoidance constraints. This approach enables the attainment of a larger feasible movement area and more precise tracking results during the motion planning process. Finally, experimental and simulation results conducted on a rigid-soft hybrid robot demonstrate the feasibility and superiority of the proposed method, validating its effectiveness in practical applications.