A fast cable-driven parallel robot for vision-assisted pick-and-place operations

Bin-picking (or 3D-picking) is a complex pick-and-place task that requires a robot to take products from a pile (usually contained in a box) and reposition them in the place area. This operation is often performed by manipulators with a serial architecture, which is characterized by high dexterity but limited dynamics. Marchesini Group S.p.a. proposed the development of a bin-picking cell for feeding an automated packaging machine with lightweight products taken from a pile with productivity equal to 100pieces/min. The aim of this thesis is to develop the manipulator to be used in the aforementioned bin-picking cell. Based on the constraints related to the coexistence of the robot with the vision system and the fast movements that the manipulator must perform to satisfy the required productivity, an overconstrained Cable-Driven Parallel Robot (CDPR) is proposed. The small footprint of the manipulator in the pick-up area facilitates the task of the vision system, and its parallel architecture in an overconstrained configuration guarantees high dynamic capabilities. The proposed CDPR has a configuration with 8 cables and 6 degrees of freedom, and it is called CORHDA: Cable-driven Overconstrained Robot for Highly Dynamical Applications. The kinematic and dynamic models used to describe the manipulator are taken from the literature, whereas the conceptual and mechanical design of the robot prototype is carried out based on an example application. A new strategy is developed to properly control the manipulator without force sensors. The experimental tests on the prototype show that movements up to 8.8m/s and 255m/s^2 of velocity and acceleration are feasible, with cables always remaining taut. The manipulator is not yet sufficiently accurate for executing the bin-picking task but is highly repeatable. Accordingly, it is assumed that a future calibration of the robot geometric parameters will increase its positioning precision.