Exocontrol Upgraded: Magnetic Encoders Power Gevo Robot's New Teleoperation Controller

Gevo Robot and the Next-Generation Exocontrol

In the world of humanoid robots and teleoperation, enabling operators to control robot movements with greater precision and naturalness remains a core challenge. Recently, a group of developers shared progress on Reddit about the new version of their Exocontrol (exoskeleton-style remote controller) built for their avatar robot Gevo, showcasing some key hardware upgrades in their component selection.

Exocontrol is essentially a wearable or handheld motion capture device that lets operators map their own movements in real time, driving the remote robot to replicate corresponding poses and actions. The precision and feel of such a device directly determines how smooth and controllable the teleoperation experience becomes.

Magnetic Encoders and Bearings: The Core Hardware Upgrades

According to the developers' post, the most significant improvement in this version of Exocontrol is the adoption of magnetic encoders and bearings. These two hardware changes deliver noticeable improvements during actual operation — both when physically moving the controller and in the tactile feel of using it.

Compared to traditional potentiometers or optical encoders, magnetic encoders offer non-contact measurement, resistance to dust and grease, long service life, and high resolution. For teleoperation joints that move frequently, magnetic encoders provide more stable and continuous angle readings, reducing the precision loss caused by drift and mechanical wear.

How Bearings Improve the Feel

Introducing bearings is another detail that's easy to overlook but critical to the user experience. Bearings significantly reduce friction during joint movement, making the controller's motion smoother, reducing operator fatigue during extended sessions, and making motion transmission more linear and predictable. The developers specifically highlighted "the improvement you can clearly feel when moving and handling the Exocontrol" — and that's precisely the cumulative result of these kinds of mechanical refinements.

Magnetic encoders work by mounting a permanent magnet on a rotating shaft and pairing it with a Hall-effect sensor chip fixed to a circuit board (such as the AS5048 or AS5600 series) to detect changes in magnetic field direction and calculate the shaft's absolute or incremental angle. Compared to potentiometers, magnetic encoders have no physical contact between a brush and resistive film, eliminating resistance drift from wear. Compared to optical encoders, they're insensitive to dust, grease, and minor obstructions, making them better suited for teleoperation devices in complex mechanical environments. A typical 12-bit magnetic encoder provides 4,096 resolution steps, corresponding to an angular resolution of approximately 0.088° — sufficient for most humanoid robot joint precision requirements. In teleoperation scenarios, encoder response latency is typically in the microsecond range, far below the millisecond-level response time of human movement, so sensor latency won't become a bottleneck.

Building a Sensor Visualization Tool with Codex

Beyond the hardware upgrades, the development team also mentioned that they used Codex (an AI-assisted programming tool) to create a utility for reading and viewing sensor values. This means they can monitor the output data from each magnetic encoder in real time, using it for debugging, calibration, and verifying whether the controller's motion mapping is accurate.

This approach reflects an increasingly common trend in hardware development: developers using AI coding assistants to quickly build debugging and visualization tools, rather than writing tedious monitoring code from scratch. For teams focused primarily on robotics projects — with their energy concentrated on mechanical design and control logic — these kinds of tools can dramatically shorten iteration cycles.

Codex is a code-generation model developed by OpenAI and one of the underlying technologies behind GitHub Copilot. It can automatically complete or generate entire functions, scripts, or full programs based on natural language descriptions or existing code context. In hardware debugging scenarios, developers typically need to write boilerplate code for serial port reading, data parsing, and real-time plotting — tasks that are repetitive but prone to errors. With Codex, a developer can simply describe their needs in plain language (for example, "read AS5600 encoder values from serial port and display them in a real-time line chart") and quickly get working prototype code, requiring only minor modifications for specific hardware before it's ready to use. This workflow allows robotics developers with mechanical engineering or controls backgrounds to rapidly build the debugging infrastructure they need, even without deep software engineering expertise.

Summary and Takeaways

Although this was just a brief progress update, it reflects several practical directions in the DIY robotics and teleoperation community: using magnetic encoders to improve sensing precision, using bearings to optimize mechanical feel, and using AI tools to accelerate development on the software side. These are all foundational building blocks for constructing a reliable avatar robot control system.

For enthusiasts following humanoid robotics and teleoperation technology, the ongoing iteration of the Gevo project is worth watching — from sensor selection to visualization tooling, the entire process demonstrates how a small team can steadily nail down each engineering detail, one step at a time.