Realizing a screwdriving operation for human-robot collaboration : During the assembly of a truck engine valve-hood

Abstract: According to Folgiatto, da Silveira and Borenstein (2012), the current trend in the industry indicates a change from a mass production to a mass customization manufacturing. This mass customization means an increase on the need of flexibility, as the production line must be ready for frequent changes in order to adapt for producing different product variants. Human-Robot collaboration has demonstrated to be a great tool for increasing not only the flexibility but also the productivity, as either the human operator or the robot will be working on the task under issue. Within this collaborative environment, equipment for providing the robot with the information about where to move, pick and place a part is needed. Multiple types of sensors are used in the industry to fulfill this need. However, the adoption of this type of technical equipment might result very expensive. This thesis focuses on the assembly of a valve-hood on top of a truck diesel engine, for which a collaborative robot will be used in order to remove the task of lifting this part from the human operator. Later both parts will screw in conjunction the bolts of the valve hood, hence, a solution for finding the bolts is needed. Despite providing a brief study of the current options in the market, regarding high resolution cameras for robots, this thesis aims to provide a low-cost alternative by not making use of any of them. A blind solution is proposed. This solution consists in the implementation of coordinate geometry equations in the robot software in order to create a program that allows the robot to find the bolts without regard to the orientation with which the valve-hood is placed in front of the robot. This code is used later for creating a sequence in which the robot is assigned to find the entirety of the bolts. In order to evaluate the performance of the adopted solution an accuracy experiment is conducted. This experiment is divided in three parts in which three different positions of the valve-hood are tried. Within each position ten repetitions are made at the 50% and 100% percent of the speed. The main metric used for measuring the accuracy is the deviation (in mm), which is measured between the target bolt position and the actual contact position found by the robot. The overall results of this experiment show a good accuracy, with a precise enough alignment to the center of the bolt position so to allow a correct screwing operation. As a conclusion, the adopted blind solution shows to be feasible for being implemented in this scenario.