For the week of September 27th – October 3rd, our team made significant progress toward finalizing the Iteration 1 design of our 6-DOF robotic arm. Our main focus was completing the mechanical design, validating the robot through simulation and torque analysis, finalizing the hardware parts list, and preparing for physical implementation when our ordered parts begin arriving.
A major area of progress was remodeling our robot around the dimensions and geometry of the UR3e model from last week. We used the UR3e’s dimensions as a reference for our own kinematic model, allowing us to use a proven robot configuration while adapting the design to our 3D-printed links and selected smart bus servos. The updated SolidWorks assembly now contains modeled representations of the purchased smart bus serial servos, power supply, shafts, and other components. The end effector was also modified to support a pen for our intended testing.
We also performed several simulations to validate the updated design. FEA stress testing was used to identify and strengthen weak areas of the printed components, while MuJoCo was used to visualize the robot and verify its motion and geometry with the SolidWorks assembly. Simultaneously, Python kinematic and torque analysis tools were used to evaluate loads at every robot joint and confirm that the purchased servos can support the weight of the robot arm. The results of these showed that the shoulder and elbow joints are near the practical limits of our selected servos at the full UR3e dimensions. To address this, the current design uses two servos at each of these joints, along with a reduction mechanism and counterbalance spring to reduce the load and side forces on the servos.
On the software side, we continued developing our ROS2 environment on the Jetson and established the project’s GitHub repository and initial software structure. We also determined that OpenCV AprilTag detection is a better fit for our vision system than using a YOLO26 model, since AprilTags can be detected directly (using the apriltags OpenCV library) without training an object detection model. This should provide a simpler and more efficient solution for our Jetson Orin Nano and Intel RealSense D455 camera setup.
Finally, we worked together to finalize the Iteration 1 master parts list and identify remaining hardware such as bearings, screws, shafts, tubing, and additional filament. With the mechanical design and simulations largely complete, our primary risk is now transitioning successfully into fabrication and assembly. The first print runs are expected to require approximately 35 hours and 900 g of filament, so we will need to carefully prioritize the initial parts and quickly iterate on any components that require redesign.
Our project is slightly behind the original schedule since the CAD design and subsequent validation took more time than anticipated. Therefore, we have had to push back most of our schedule by about a week, so printing and assembly will commence this following week.
For next week, our team plans to:
- Begin printing the finalized Iteration 1 components, starting with the base.
- Measure and test initial prints before committing to the full build.
- Order any remaining hardware and additional filament required for assembly.
- Begin assembling the printed components and servos as parts become available.
- Continue developing the ROS2 framework, including the URDF and tf2 structure.
- Begin prototyping AprilTag tracking using the Jetson and RealSense camera.