Argus V2 — Modular Hardware for a Spherical Robot

I’m developing the hardware for Argus V2, a spherical robot intended to move on land and in water. My work covers the structure and shell, actuator integration, thermal management, sealing, and motor configuration. Other members of the lab develop and train the learning-based control system.

Argus explores dynamic symmetry: how uniformly a robot can accelerate in different directions. V2 builds on that research with new actuation mechanisms and amphibious-design requirements. General Robotics Lab Argus research

This project is ongoing, and the images shown here reflect the robot’s latest stage of development.

One Body, Two Actuation Configurations

When I joined, an earlier V2 prototype used translating internal masses. My work focuses on two newer configurations: rotating eccentric masses inside the shell, and a related mechanism that also moves exterior shell sections to interact with the environment.

Rather than developing two independent robot bodies, we converged on a shared base structure and motor/shaft arrangement with interchangeable face assemblies. One configuration uses fixed caps; the other replaces them with compartments that allow the exterior sections to move with the actuation. This keeps the common structure while letting the team explore different mechanisms through changes at the faces.

I developed the mechanical hardware in Fusion 360 and refined it through many physical prototype iterations. I also removed unused material and revised the internal supports, motor positions, and electronics placement to reduce size and weight. The final architecture combines off-the-shelf components with custom PLA parts, prioritizing straightforward fabrication and assembly in a fast-paced research setting.

Extending Operation Inside a Sealed Shell

Overheating was a practical limitation of the earlier linear-mass prototype. Because the actuators and electronics operate inside an enclosed shell, thermal management had to be considered alongside the mechanism and packaging.

I selected motors with higher-ratio gearing to reduce motor loading. Their integrated controllers also eliminated separate motor-driver boards, simplifying the internal hardware. I added conductive thermal interfaces and structures to carry heat away from the motors and electronics board.

Across the redesign, operating time before overheating increased from roughly 5 minutes to 20–30 minutes, an approximately fivefold improvement using the midpoint. Both versions operated within sealed shells; the result reflects the combined mechanical, actuator, and thermal changes between prototypes.

Integrating the Rest of the Robot

The hollow central structure houses the electronics, with actuators arranged around six faces. I worked through the competing needs of internal space, structural support, camera views, wiring clearance, and battery access. Wiring had to accommodate the moving mechanisms, while the battery and other serviceable components needed to remain accessible.

For the shell and face assemblies, I considered sealing together with surface features intended to support ground traction and water propulsion. We tested the seals at compartment interfaces and openings to check that those connections prevented water ingress into the protected internal spaces.

I also configured and tested the integrated servo hardware over CAN on an ARM-based Linux system, working through communication and feedback checks. This connected the mechanical design to the actuator hardware the control team would use.

What I Learned

This project taught me to think about the robot as a set of connected design decisions. Sealing changes how heat can escape. Actuator selection affects both the mechanism and the space needed for electronics. A compact layout still needs room for wiring, assembly, and maintenance. Improving one feature meant checking what it changed elsewhere in the robot.

I also learned the value of keeping the experimental platform adaptable. Sharing the body between the two actuation configurations let us concentrate the differences in the face assemblies. Repeated physical prototyping helped me turn those requirements into hardware that could be assembled, checked, and revised as the research developed.