Neural network application on cable driven parallel robots.

- Not high accuracy of trajectory tracking control because friction in the pulleys and cables, non linear characteristics of cables:   structural elongation, static and dynamic creep, hardening, long and short term recovery, hysteresis. => Complicate model         for accurate tension of cables or cable characteristic.

- To simply, assume complicated model of physic system is black box and use recurrent neural network to train and create          model with input data and output data.

Introduction and Objective

- Cable-driven parallel robots (CDPRs) have been developed as a special type of parallel robots in the last decades, which adopt flexible cables to manipulate the end-effector.

- However, As a result of cable actuation the large workspace makes the CDPRs easy to interference with the external environment. Therefore, the unknown moving obstacle such as moving particles, drones, or birds may enter the workspace to introduce the collision phenomenon.

- Our work is to find a feasible method to deal with moving obstacle avoidance problem for CDPRs, which means considers various constraints caused by the cable particularity must be considered.

Cable Driven Parallel Robot (CDPR)

- Control the end-effector with 8 cables

- limitations on moving system (The entire system must be moved)

Mobile robot

- Detect / avoid obstacles and arrived at destination

Mobile CDPR

- Mobile-Free CDPR with mobile robot

- Can be used in various industries such as architecture, space, etc