Version 7. Served two years and three international competitions with YURS, at the 2023 Fall, 2024 Winter, and 2024 Fall Canadian International Rover Competition (CIRC).
May 2023 to Aug 2023
A low-cost modular rover gripper built around reused lead screws, scavenged old rover hardware, machined aluminum parts, and 3D-printed task-specific pads. This was my first contribution to the mechanical subteam at the York University Robotics Society rover team.
Quick Read
- Built a low-cost rover end effector for CIRC-style manipulation tasks.
- Reused old rover lead screws, motor hardware, aluminum stock, and printed parts instead of buying a new mechanism.
- Iterated through seven versions, from a CAD 400 McMaster concept to a final machined/printed assembly.
- Designed around swappable task pads, printed gears, wire routing, camera/laser mounting, and competition repair constraints.
- Bench-tested grip behavior and estimated lead-screw clamp force, but did not measure actual grip force with a load cell.
- Mounted on the rover, though the full rover arm/control stack was not stable enough for a clean competition task demo.
Team Need
The rover needed an end effector for CIRC-style tasks, not just a generic open-close claw.
The end effector had to handle several different task objects without becoming expensive, heavy, or hard to rebuild between events.
Requirements
- Grab and lift 4 inch PVC pipe.
- Press buttons.
- Hook or latch onto loops.
- Turn a ball valve.
- Switch on a water pump.
- Handle unknown “uranium rod module” geometry.
- Leave room for possible tool or soil-collection attachments.
The “uranium rod modules” at competition. The geometry was not something I could design around exactly, so the gripper needed some tolerance for unknown shapes.
The astronaut plushy rescue task was one reason the gripper had to close around soft, awkward objects instead of only rigid parts.
Constraints
- Keep the cost low, ideally under CAD 70.
- Reuse old rover hardware where possible.
- Avoid relying too much on 3D-printed load-bearing parts.
- Machine only a small number of aluminum parts.
- Make the task pads swappable instead of redesigning the whole gripper for every task.
- Keep printed parts small enough to use the university Sandbox print service when possible.
The first few versions were mostly about finding a mechanism that could do the job without becoming expensive or hard to machine.
Iterations
v1 was the expensive McMaster version. It helped define the mechanism, but it was not realistic for the team budget.
v1 - McMaster Version
- Left-handed lead screw.
- Double-shaft worm gearbox 12V DC motor.
- Mostly McMaster-Carr parts.
- Around CAD 400 before tax and shipping.
- Dropped because the cost and left-handed hardware sourcing made no sense.
Found Parts
v1 was too expensive, so the project changed direction during the annual hangar cleanup.
Parts found during the annual hangar cleanup in the Bergeron Centre. This changed the project from a McMaster-Carr shopping list into a reuse-what-we-have gripper.
The hangar was the garage/workspace we used in the Bergeron Centre at YorkU. While cleaning it out, we found old rover hardware that could be reused for the gripper.
That changed the design completely. Instead of buying a left-handed lead screw, worm gearbox, and new hardware, I started designing around what was already in the room.
- Motor from the 2022 gripper.
- Lead screw from the 2022 wrist.
- Existing aluminum stock.
- Existing linear bearings / guide hardware.
- Printed parts only where they made sense.
v2 was the first serious low-cost version, built around reused rover parts and two right-handed lead screws.
v2 - SolidBugs
- Two right-handed lead screws, made by cutting the reused lead screw into two pieces.
- Printed gears inverted one side so the two sides could move together.
- Motor from the 2022 gripper.
- Interchangeable printed task pads.
- Almost CAD 0 except bearings and bolts.
- Dropped because the SolidWorks assembly became unreliable.
v3 simplified the gear and shaft-coupler interface instead of treating the printed gears as separate parts.
v3 - Coupler Gear Version
- Printed gears built around available shaft couplers.
- Orange PLA parts for gears and pads.
- Still needed bearings.
- Kept the two-lead-screw layout.
- Main goal was to reduce part count and make assembly easier.
v4 to v6 were mostly buildability iterations: bearing fits, shaft retention, CAD rebuilds, drawings, printed gears, and the first moving assembly.
v4-v6 - Making It Buildable
After v3, the main mechanism did not change as much as the CAD and fabrication details did. I kept restarting the assembly because each rebuild made the layout cleaner. At the time, every version felt like the cleanest one I could make, then I would start over and find a simpler way to constrain the parts or package the mechanism.
- Kept the two right-handed lead screw layout.
- Reworked the printed gear and shaft coupler interface.
- Worked through linear bearing fits and shaft retention.
- Updated the CAD around the actual scavenged parts.
- Prepared drawings and STLs.
- Moved from CAD to six machined aluminum parts.
- Reached the first moving proof of concept.
The first moving version used the 2022 wrist lead screw cut into two pieces, about 14 cm and 16 cm long. Those were mounted to opposite-rotation gears so both moving jaw bodies travelled together.
v7 - Final Version
v7 was the final integrated version. It kept the reused lead screw mechanism, but cleaned up the packaging, covers, mounting points, and printed task geometry.
v7 was the version that actually got built and stayed with the rover.
At this point the mechanism was mostly settled: two lead screws, printed gears, machined structure, and swappable task pads. The main change was making the assembly easier to build, mount, wire, and maintain.
I originally made the moving jaw bodies from aluminum. After machining them, I realized that was not a useful place to spend weight. The printed gears, budget, and packaging mattered more than making every visible part metal.
Earlier moving jaw bodies were machined from aluminum. After making them, I realized they were heavier than they needed to be and were not the part most likely to fail first.
First full assembly. The main structure was machined aluminum, while the orange printed parts handled gears, task pads, covers, and adaptable geometry.
The final design was less about making every part stronger and more about putting strength where it mattered.
- Two reused lead screws.
- 3D-printed gear set.
- Swappable printed task pads.
- Machined aluminum structure.
- Printed moving jaw bodies instead of aluminum ones.
- Central cover around the exposed gear area.
- Zip-tie holes for wire routing and competition fixes.
- Mounting points for camera, laser, and tool ideas.
- Motor encoder available, though it was not used in the final control setup.
The center cover was partly there to close the open gap around the gears. It also became a useful place to mount or route small things.
I added holes because competition hardware always ends up needing last-minute wire routing, strain relief, or temporary fixes. Zip ties were the realistic answer.
The task pads also had large openings so a valve handle could fit through them. The idea was to close around the handle and use the wrist rotation to turn the valve.
Grip Testing
Crude grip test. I was watching motor current as a rough signal for when the gripper had contacted the object and was clamping hard enough; this was not a calibrated force test.
Grip Force Estimate
I did a rough lead-screw force estimate before the first full assembly.
The assumptions I wrote down at the time were:
- 29 kgf-cm motor stall torque.
- T8-style 8 mm lead screw.
- 2 mm lead used in the original estimate.
- 0.1 assumed friction coefficient.
Using those assumptions, I estimated about 362.5 kgf of clamping force at the moving jaw body.
Later, after more assembly work, I described the design as roughly 300 kgf of grip force. I would treat both numbers as rough theoretical estimates, not measured results.
The basic idea was that motor torque gets converted into axial force by the lead screw.
$ F = \frac{2 \pi T}{l} $
A more realistic power-screw estimate includes thread friction:
$ T = F \frac{dm}{2} \left( \frac{l + \pi \mu dm}{\pi d_m - \mu l} \right) $
T is motor torque.F is axial force from the lead screw.d_m is mean screw diameter.l is lead per revolution.μ is the friction coefficient.
I used the nominal 8 mm screw diameter in the original estimate. A more careful calculation would use the thread mean diameter and would check the actual screw efficiency, so I would not present the 362.5 kgf number as a measured or verified result.
I also need to verify whether that screw was single-start or multi-start. The equation uses lead per revolution, not just thread pitch, so this assumption matters a lot.
The important part is that I checked the order of magnitude and realized the mechanism could generate much more grip force than the printed parts or task objects probably needed.
What I should have done next was test it with a load cell between the pads and record actual grip force against motor current.
The motor had an encoder, so position feedback was physically available. It was not used in the final rover setup because the interface did not match what the electrical/software side was using at the time.
Unfinished but useful test. The part that failed was not the gear train; the printed jaw body walked off the brass lead screw nut because I had not screwed it in yet. I'm recording this video.
The printed gears were still something I worried about. I had to use printed gears because proper gears did not fit the budget, and I printed spares in case they failed at competition.
By the final version, they held up better than I expected.
Integration
The gripper was mounted to the rover and tested separately for opening and closing. It was physically ready enough to reach the rover, but using it in a task depended on the rest of the arm and rover stack being ready too.
Gripper mounted during a 2023 CIRC indoor panel task. This is not a clean gripper demo; it shows the rover trying to approach the task panel while the full arm, wrist, drive, and control stack was still being debugged.
I never got a clean competition run where the gripper completed a task.
That was disappointing, but it is also a normal systems problem. The gripper depended on the drive base, arm, wrist, wiring, power, motor controllers, and software all working at the same time.
The electrical and software teams were working through custom Vroom motor controllers, CAN communication, drive reliability, circuit protection, power issues, and later Spark MAX controllers for the arm.
For this page, the honest outcome is that the gripper reached the rover, but the full rover was not stable enough to show it properly in a competition task.
What Changed
The biggest change was going from a bought-parts design to a found-parts design.
The second biggest change was realizing that metal was not automatically better. Some parts needed to be machined aluminum. Other parts were better as printed geometry because they were lighter, faster to replace, and easier to adapt for tasks.
The final gripper was not the design I would make with more time and money. It was the design I could actually build with the team’s budget, tools, and timeline.
What I Would Improve Now
I would measure the actual gripping force instead of relying on a theoretical lead-screw calculation.
I would use a load cell between the task pads and record force against motor current, so the current limit could be based on real data instead of a rough bench test.
I would also spend more time on integration earlier. The gripper had an encoder available, but it was not used. If I were doing it again, I would define the electrical and software interface earlier instead of treating the gripper mostly as a mechanical object.
I would still keep the swappable task pads.
That part of the design made sense. The rover tasks were too varied for one fixed jaw shape, and changing printed pads was much cheaper than redesigning the whole gripper.