AGIBOT — Faster Iteration for Robot Test Hardware

At AGIBOT, I designed fixtures, sensor mounts, and structural supports for testing the X2 humanoid robot. My work included head roll-pitch calibration fixtures, IMU mounts, acoustic-test supports, and modifications to robot components to accommodate test equipment.

I developed parts and engineering drawings incorporating GD&T in SolidWorks, iterated through printed prototypes and fit checks, and worked with other team members to resolve CAD and interface issues. A significant part of the internship was finding ways to keep testing moving while the hardware continued to change.

Due to copyright, only a limited number of non-CAD photos can be shown for this project.

Choosing the Manufacturing Method for the Test

Many test components were initially outsourced in metal because they were modified versions of metal parts already on the robot. For temporary sensor-mounting and test-support work, I questioned whether every part needed the same manufacturing approach as its production counterpart.

I proposed and implemented 3D-printed alternatives for suitable test components. These parts needed to fit the existing hardware, carry the intended test loads, and be available quickly. They were testing tools, not replacements for production robot components sold to customers.

Printing gave us a way to revise and check those parts without waiting for another outsourced delivery. Turnaround for suitable prototypes moved from days to hours. In the first few examples, the corresponding metal parts arrived roughly one to two weeks later, after testing with the printed versions had already finished or moved to the next stage.

Refining Strength Through Print Orientation

The substitution still required engineering judgment. Early prints built in an unsuitable orientation snapped under loading. I changed the orientation and refined the prototypes until they supported their intended test use.

That experience made print direction part of the design process. I checked the geometry together with how the part would be loaded, then used physical fit and loading checks to guide the next iteration. The team continued using printed alternatives for suitable later test parts.

I estimated the prototyping-cost reduction at roughly 30%, allowing for material use and printer cost. The more immediate benefit was keeping the test cycle moving: the team could use and revise a functional prototype while the metal version was still being manufactured or delivered.

Taking Parts Through the Engineering Workflow

Across the internship, I delivered about two parts per week, checked for fit and loading, and coordinated revisions with teammates. I also updated bills of materials, released part records in Teamcenter PLM, and initiated procurement requests for robot components.

This work connected the CAD model to the practical steps needed to make a part usable by another person: a clear drawing, a workable manufacturing method, a fit check, and an up-to-date part record.

What I Learned

This internship taught me to start with what a part actually needs to do. A test fixture can resemble a production component while having a different service life, loading requirement, and delivery priority. Recognizing those differences helped me choose a manufacturing method that supported the test rather than automatically repeating the original material choice.

I also learned that a fast prototype is valuable when the team can use it to make the next decision. The fit checks and print-orientation revisions were what made the shorter turnaround useful. Seeing testing progress before the metal parts arrived made the effect of that design choice tangible.