Background
Autonomous Mobile Robots (AMRs) combine robotic manipulator arms with highly controlled mobile platforms. The existing robotic technology often utilizes the mobile base to position a robot in a specific location to perform a manipulation task. An industry need exists to utilize these subsystems in a coordinated fashion. A target application has been identified for structural welding of a large structure. The project’s objective was to research and to evaluate existing robotic weld technology such as seam tracking algorithms that are used in highly constrained welding applications (such as a part precisely positioned in front of a robot) to determine how these capabilities could be scaled for integration with a highly unconstrained mobile robotic welding processes for large structures, such as ship hulls and heavy industry vehicle frames.
Approach
SwRI has experience in developing algorithms capable of detecting unique features within captured digital images. Additionally, the project team previously developed a range of trajectory control algorithms that use feedback for dynamic robot control. These two areas of expertise were used to evaluate the range of potential methods and to identify the leading approach and capabilities of an automated offset compensation control solution to better understand the real-world implications of tracking a weld seam on a mobile manipulator. To test the solution, an omni-directional platform was commanded to move in a variety of motion profiles (parallel, sinusoid, staggered, and sawtooth) along the length of a simulated steel plate’s weld seam to determine if the robotic arm and vision system could maintain positional accuracy.
Accomplishments
We designed, built, and tested a mobile system capable of tracking and providing real-time adjustment to robot position to maintain consistent positioning. The vision system feedback provided real-time directional offset information to the robot system during vehicle motion, which was then applied to a velocity-based PID control loop to allow the robot to minimize the error between its current position and the goal position while limiting positional overshoot. The robot maintained a positional accuracy of approximately ±1 mm of center of the weld seam during robot and vehicle motion, with a maximum error of 7 mm.