Autonomous Test Connection Arm for Vehicle Dash Components

Robotic automation is a growing process that PACCAR wants to add to its Manufacturing Support Team. Our solution frees up test operators to use their time more efficiently and reduces the time required to test each dash, saving time and money for the plant and company.

Our team incorporated a Kawasaki RS007N robot arm capable of accurately connecting a plate containing multiple cable connectors to the plugs of the rear of a semi-truck dashboard. The system uses ROS, MoveIt, Intel RealSense, and object detection software to locate and align the dashboard from camera video, enabling precise automated connection.

Objectives:
- Program a robotic arm capable of automating the connection of a tester cable to semi-truck dash assembly.
- Reduce errors and damage caused by incorrectly installed cables.
- Enable operators to utilize their time more efficiently.

Technical Pipeline:
- Created a ROS workspace integrating MoveIt, Intel RealSense Camera, and pose estimation.
- Used NVIDIA’s DOPE for 6-DoF pose estimation, trained on synthetic datasets from Unreal Engine 4.
- Developed a TensorFlow model for bolt detection, optimizing motion planning.
- Kawasaki arm successfully traveled to dashboard at an almost perfectly aligned position.
- Visualized estimated pose and arm’s end pose in ROS.

Challenges & Future Work:
- Environmental noise (lighting, mechanical parts) can affect model performance.
- The Kawasaki RS007N arm has 6 degrees of freedom; 7 joint arms or linear actuators could improve trajectory planning.
- Mechanical improvements can increase noise tolerance.

Skills & Tools: ROS, MoveIt, Intel RealSense, SolidWorks, NVIDIA DOPE, TensorFlow, Unreal Engine 4, Python, C, C++
Education: University of Washington, MS Electrical and Computer Engineering, 2023