Time-Delay Learning-Based Controller for Fully-Constrained Cable-Driven Parallel Robots

Abstract

Despite model-based approaches are capable of providing acceptable performance, but they require complex dynamics modeling, which make them difficult to apply. To overcome this problem, a time-delay learning-based controller as a model-free approach is proposed for fully-constrained cable-driven parallel robots (CDPRs) considering the positive cable force constraints and actuator limitations. The proposed method uses the memory of control effort as a learning element to linearize the system. In addition, using a saturation function in control law has provided the opportunity to ensure all the cables are bounded and remain in tension, to guarantee both the positive cable forces and actuator limitation. Moreover, in the proposed method, we are able to eliminate the requirement of measurement of the task variables, which is very costly by using expensive external positioning measurement systems. Finally, a fully-constrained CDPR is used to test the proposed method, and the results demonstrate how effective is the method in terms of tracking performance as a model-free controller.