Enabling Adaptive Capabilities for Robotic Sanding Systems

Southwest Research Institute has been developing Scan-N-Plan^TM technologies for many years with applications in painting, sanding, grit blasting, grinding, chemical depaint and laser ablation. Scan-N-Plan, a suite of ROS-Industrial tools, enables real-time robot trajectory planning from 3D scan data. This allows robots to use 3D data to respond dynamically to the geometry and position of an object, often with no human intervention. The technology is also helping to automate labor-intensive tasks that might otherwise cause repetitive strain injury in human workers.

Scan-N-Plan applications may involve the following base capabilities:

1. Automated intrinsic and extrinsic calibration of 2D and 3D cameras.
2. Automated scanning and intelligent surface reconstruction.
3. Mesh or point cloud processing including filtering, segmentation, and localization.
4. Automated tool path generation both for scan path and process paths.
5. Automated robot path planning including free-space and Cartesian trajectory planning.
6. Machine learning for detecting features and masking, and for process monitoring.
7. Computational geometry to perform tasks such as estimating applied paint thickness.
8. Deposition thickness or the expected contact area of a tool given the mesh and the associated tool path.
9. Interfaces to common industrial controllers including PLCs and force-compensated compliance devices.

Case Study: Robotic Sanding on an Aircraft Propeller

We recently had an opportunity to develop a Scan-N-Plan solution for a customer seeking robot force-controlled sanding over an aircraft propeller blade. The goal was to ensure sanding had been accomplished while not sanding more than necessary. The customer also noted they wanted a specific force-controlled sanding head. The curvature and length of the aircraft part, coupled with the robot force control requirement, led to development of a unique robotic sanding system solution.

We prepared a robot force-controlled sanding head and a 3D camera, which also included a 2D color camera. Prior to operation, the camera was calibrated using the automated capabilities available in an open-source industrial robotics calibration library used for this project.

First, the robot collected a mesh of a portion of an aircraft propeller blade designated for sanding. The solution uses a software function that monitors the robot's pose as 3D data clouds stream in.

The next task was the generation of tool paths for sanding. For this operation, an off-line tool was employed. An edge following path and three distinct area-covering paths were generated. The off-line user interface provided an interactive interface for a variety of processing capabilities including segmentation, edge, and raster path generation.

For this request, the goal was to have the sanding solution operate within a desired boundary, and not sand beyond that boundary. To ensure this, we used machine learning and an edge adjustment solution which micro-adjusted the edge-following waypoints.

To obtain uniform sanding results, three separate techniques were employed. First, we generated a crosshatch pattern of “covering” strokes. Next, we computed the estimated contact area of the sanding head base and dynamically reduced the commanded force applied to maintain a constant sanding pressure continuously. Finally, we employed a machine learning algorithm trained to detect when sanding was complete and stop sanding at any way point observed to be completed.

Automating Robot Arm Trajectory Planning

Lastly, automated robot trajectory planning is an enabling technology that was working behind the scenes. The ROS-Industrial-based Tesseract and TrajOpt software repositories were able to automatically compute collision-free paths to and from the part and to all the designated waypoints. The algorithms take advantage of any kinematic redundancy in the system to provide collision-free and smooth joint trajectories that follow complex process paths.

The blending of advanced manufacturing software tools, capabilities, and hardware are enabling new applications that were previously not possible. These tools extend beyond sanding, making a range of surface processing options available for a range of coatings and substrates.

Stay tuned for additional updates, as we work on tougher substrates and heavy-industry parts. Contact SwRI to discuss robotic sanding system solutions or to learn how we can help you solve your advanced robotics application needs.