Analyse, conception et contrôle de robots collaboratifs
|Abstract:||The project behind this thesis is about improving the working conditions of workers on assembly lines. A research team from General Motors came to us with a problem to work on: the workers on their assembly lines must use body postures that are difficult to bear when performing some tasks. These postures can become health hazards because of the repetitive nature of their work. The proposed solution to this problem is to make the operator work in a safe remote area from the assembly area, where he can keep a comfortable posture and remain far from the potential hazards of the assembly line. To make the link between the safe area and the assembly area, the solution of using a robot has been suggested. This thesis presents the work accomplished to design such robot. This thesis is composed of four chapters, each corresponding to an article dealing with a different subject. Chapter 1 deals with two different types of assembly tasks, presents their analysis and discusses solutions to introduce mechanisms in their process. A tool is designed and tested to perform snap-fit assembly tasks. Motion strategies are explored as well as vibrating mechanisms to deal with hose assembly tasks. Test phases and data measurements are presented in a video. In chapter 2, the issues associated with performing a task remotely are raised. Indeed, distance enhances greatly the required concentration and perception for a human operator to perform the task, resulting in a failure or a greater operating time, reducing productivity. Therefore, the accuracy correction mechanisms were considered. First, solutions with springs are presented to design compliant mechanisms that return to a neutral configuration when unloaded. These solutions bring the originality to introduce force thresholds to keep the compliance passive when not needed. Then, accuracy correction mechanisms are introduced through RCC mechanisms. Several rotational and translational mechanisms adapted to human collaboration are presented. Tests to validate the concept are shown in a video. After the design of passive mechanisms, the scope of the project turned to active solutions. The search for an effective architecture led to the contents of chapter 3. It presents a sixdegree-of-freedom 6-ꝐUS parallel robot. The kinematic analysis of the robot is presented, followed by an algorithm to determine the geometrical workspace of the robot. A singularity locus analysis and a force capability analysis are then presented. Even if this architecture was not selected in the end, the methods developed were used for the design of the final architecture. After several iterations, an architecture was chosen for the active robot. This architecture is presented in chapter 4. After a design process based on the work shown in chapter 3, three control schemes are presented. The first one is a classical position control which is a requisite for more advanced schemes. The second control scheme introduces concepts previously raised with the passive mechanisms discussed in chapter 2. The model simulates a RCC mechanism with the advantage of being reconfigurable without hardware modifications. The last control scheme introduces computer vision. An ARUCO marker is placed on the robot and the information it provides are injected in the control scheme. The objective is to simulate an environment where the robot detects the pose of the parts to assemble and adjusts itself in real time to compensate for the errors of the human operator.|
|Document Type:||Thèse de doctorat|
|Open Access Date:||8 February 2021|
|Collection:||Thèses et mémoires|
All documents in CorpusUL are protected by Copyright Act of Canada.