The thesis addresses development of the robotic system for real time trajectory planning of robot manipulator, based on measurements from a line laser sensor. During motion, the system estimates the reference pose, depending on current surface geometry. For that reason the system doesn't require any pre-known surface model. The reference pose is determined so that the tool maintains desired distance from the surface and an orientation normal to the local surface. The system regulates selected degrees of freedom, while the remaining degrees of freedom can be controlled by an operator.
Online estimation of the reference pose is determined with estimation of local surface geometry, so that system during motion acquire depth profiles of unknown surface, from which the local cloud point is formed. Using PCA method and geometric relationships, the developed algorythm calculates the manipulator's reference position and orientation. The system supports two different modes for specifying the direction of motion in xy plane. Firsty, direction could be pre-defined and secondly specifying direction with teleoperation, where direction of motion is determined by operator with haptic device. The proposed approach is validated in the IsaacSim simulation environment and implemented on a real robotic system consisting of a Franka Emika Panda manipulator, an Omron line laser sensor, and a Phantom haptic interface.
Results demonstrate the proposed approach in the IsaacSim simulation environment and describe the operation of the implemented real robotic system. Next, comparisons between reference and actual trajectories are presented and described for both modes. During reference pose estimation, particular attention is given to measurement noise, which significantly affects real-time reference pose planning. Therefore, control errors were further analyzed with various smoothing parameters. It turns out that an appropriate level of filtering must be determined, since excessive filtering would result in poorer system response to surface variations.
|