Trajectory Planning for Contour Following and Robotic Deburring

Abstract

The work of this thesis is part of the RoboDeb project, aimed at investigating the application of industrial robots on deburring sheet metal parts. This thesis presents investigations, developments, and implementations of trajectory planning algorithms to automate trajectory design. This thesis first develops custom CAD/CAM software called the Planar Computer Automated Trajectory (PCAT) planning algorithm to establish baseline performance. Trajectory planning is next automated using the novel Planar Image-Space Trajectory (PIST) planning algorithm, which uses computer vision alone to generate machining trajectories. The PIST algorithm is completely automated, with no required CAD data of the workpiece, adapting to new workpieces or manufacturing imperfections. The feasibility of the PIST algorithm is investigated through robotic deburring experimentation, where it successfully deburred the sheet metal parts. The PIST algorithm proved to be the preferred candidate over the PCAT algorithm, as it provided an indistinguishable resulting surface finish while reducing the setup time.

Precision Robotic Machining: Modelling and Control Innovations for Improved Performance

“…The deburring tool is a 3/4" 120-grit flapwheel tool, rotating at 22 000 rpm. The path planning process from [79] is used to program the contouring path. The path commands a desired workpiece-tool contact force of f d = 2.5 N; however, this force is reduced to 1.25 N around corners to prevent rounding.…”

“…This contour is then fitted to workpiece geometry using an ICP algorithm. Further details are provided in [79]. This method is less accurate, with a mean position error of 2.5 mm (0.098"), but can be completed much more quickly, in approximately 15 seconds.…”

“…Camera System: The camera system is used to localize workpieces. The design and implementation of this system are discussed in [79]. The system uses two Basler both the robot manipulator and the workpiece.…”

“…This command calls a Python library written by Wade MacMillan, a fellow researcher within the same project [79]. The Python scripts photograph the part, then perform a series of steps, including rectification, lens warp correction, background subtraction, and edge detection, to isolate the workpiece contour.…”

“…This contour is then fitted to the workpiece geometry, stored in a DXF file, using an ICP algorithm, and the part transform is returned, representing the position of the part in the world frame. Further details can be found in [79].…”