A Gait Pattern Generator for Closed-Loop Position Control of a Soft Walking Robot.

Area of Science:

• Robotics
• Soft Robotics
• Biomimetic Engineering

Background:

• Controlling complex robotic systems, especially soft robots, presents significant challenges due to their high dimensionality.
• Gecko-inspired locomotion offers unique advantages for adhesion and mobility in unstructured environments.
• Existing control methods often struggle with real-time adaptation and precise trajectory planning for soft robots.

Purpose of the Study:

• To develop an efficient control strategy for a gecko-inspired soft robot to achieve precise position control in Cartesian space.
• To reduce the dimensionality of the robot's configuration space for simplified control.
• To enable the robot to navigate complex environments with arbitrary obstacles.

Main Methods:

• Formulating constraints under a constant curvature assumption to reduce the robot's nine-dimensional joint space to two generalized coordinates (velocity space).
• Approximating the direct kinematics mapping from velocity space to Cartesian task space using a bivariate polynomial.
• Developing an optimization problem for recursively generating optimal references to reach target positions.

Main Results:

• The joint space dimensionality was successfully reduced from nine to two, defining a velocity space.
• Direct kinematics were approximated by a bivariate polynomial, validated through simulations and experiments.
• The proposed gait pattern generator enabled the robot to master arbitrary obstacle courses in both simulation and experimental settings.

Conclusions:

• The simplified control framework effectively manages the gecko-inspired soft robot's position in Cartesian space.
• The dimensionality reduction and polynomial approximation provide an efficient method for trajectory planning and control.
• The robot demonstrated robust performance in navigating complex obstacle courses, validating the control approach.