Point-to-point motion controller for soft robotic manipulators

Accurate kinematic motion control of continuum manipulators based on soft robotic technologies is considered a challenging task due to non-linear material properties, high-dimensionality, and continuous state-space. This challenge becomes even more non-trivial in systems with multiple functional capabilities. We present a hierarchical controller that generates a trajectory within the reachable workspace of the manipulator in the top tier. It is then sampled at discrete points. A sample is fed sequentially to the lower-tier, which generates the required motor commands through a novel online parametric generalized policy iteration scheme. The main contribution of the work is to test the practicality of this controller on a simulated modular soft manipulator based on two performance criteria: accuracy and timing. A single module of the manipulator comprises of three radially arranged soft actuators that have the following hypothetical multifunctional capabilities: contraction, extension, and omnidirectional bending. The manipulator is moved through a sampled circular trajectory. The algorithm is able to generate a solution in 6.92s on average for an optimized within a 0.003m accuracy limit. The robustness of the algorithm is verified by increasing the number of samples and using an unbiased starting position. This controller paves way for real-world applications of these systems.