Smooth trajectory planning for anthropomorphic industrial robots employed in continuous processes

The first robotic arm was employed in an automotive assembly line in the 1961. Since then, the industrial robots became progressively more widespread and now they can be commonly found even in small manufacturing enterprises. The trajectory planning is an essential aspect of the motion programming for a robotic arm. In collaboration with Comau, it was decided to focus on the study of a particular type of process, namely continuous process. Continuous processes are tasks that require a continuous motion of the tool in terms of positions and velocity profile and are very common in industrial manufacturing. Some examples are arc welding, spray painting and adhesive sealing. A smooth execution and a precise control of the Cartesian speed is essential to guarantee a good quality of the final product. The objective of the thesis is to propose a novel trajectory planner for this kind of processes. After a brief overview of the techniques currently employed in the trajectory planning for some common industrial processes, part of the efforts are dedicated to the geometrical analysis of the trajectory itself. This is important to find the critical parts of a path that may be difficult for the machine to follow with high precision. In this parts a lower speed can be imposed in order to obtain smoother results. This kind of work is generally done manually by trial and error, but it requires a lot of work and it can not achieve high precisions. Instead a method is proposed to simulate and automatize this process. After that, the actual trajectory planner and its implementation is described. It is based on the spline interpolation and different techniques are analysed in order to study their performance in terms of quality of path obtained and velocity profile. The planner is proposed in two versions, one that considers all the points of the trajectory at once an another that works only on a small subset at a time. The second version is proposed having in mind a possible on-line implementation, which limits the number of points that can be used to plan the trajectory. The proposed methods are implemented in MatLab and have been tested on real processes proposed by Comau. The trajectory computed with the novel planner have been properly simulated and then fed to a real robot bypassing the on-board planner. The results obtained have been analysed and verified in order to validate the approach. The work done until now do not fully resolves the complex problem of trajectory planning for continuous processes, but the results are encouraging and they can be used as a basis for a future development and implementation.