Selected Work

A few of the builds behind Burf Robotics — real robots, real code, running in the field. Grounded in ROS 2, Raspberry Pi, Jetson, iOS, and a mixed fleet of humanoids and rovers.

▶ Watch our robots in action on YouTube

Sanbot Elf humanoid running a ROS 2 driver with a live map on screen

ROS 2 · World-first

A ROS 2 driver for the Sanbot Elf

Challenge: The Sanbot Elf is a closed Android service humanoid with no ROS 2 support — invisible to the standard robotics toolchain.
What we built: A ROS 2 driver that puts the Sanbot on the ROS 2 graph. Reverse-engineering the motion system revealed undocumented holonomic movement — its three omni wheels can strafe and spin in place, not just drive and orbit.
The result: Believed to be the first ROS 2 driver for the Sanbot Elf, unlocking a mass-market humanoid for standard mapping, navigation and teleop tooling.

A mixed fleet of robots: Sanbot, AfoBot, iPal, InMoov and a ROS 2 rover

Fleet · Control layer

One control layer for a mixed robot fleet

Challenge: Every robot family speaks its own language — vendor SDKs, ROS 2, bespoke firmware. Operating a mixed fleet usually means a different app per robot.
What we built: A single control stack that drives wildly different machines through the same primitives — move, speak, see, and stream telemetry — across a Sanbot humanoid, an AfoBot companion, an iPal, an InMoov, and a ROS 2 rover.
The result: Four-plus distinct robots operated from one interface, with a thin per-robot driver doing the translation — the differentiator behind Burf's fleet work.

PiBob, a 3D-printed Raspberry Pi humanoid robot

Open source · Raspberry Pi

PiBob — an open-source Pi humanoid

Challenge: Humanoid robots are expensive and closed — a hard place for a hobbyist or student to start.
What we built: A 3D-printable humanoid around a Raspberry Pi: 16-channel servo control, a built-in gesture library, live camera streaming over a web interface, a REST API, and STL files to print your own.
The result: A buildable open-source humanoid for under £100, published on GitHub for the maker community.

View on GitHub

A ROS 2 rover next to a live SLAM map in RViz

ROS 2 · SLAM · Nav2

A SLAM + waypoint-navigation rover

Challenge: Turning raw sensors — LiDAR, depth camera, IMU, odometry — into a robot that maps a space and drives itself to a point.
What we built: A mobile robot running ROS 2 with slam_toolbox for mapping and Nav2 for waypoint navigation, on Raspberry Pi and Jetson hardware, with the same driver reused across a rover and a TurtleBot 2 / Kobuki.
The result: Live mapping and autonomous waypoint navigation indoors — the hands-on ROS 2 experience that feeds Burf's driver and fleet work.

InMoov and Sanbot robots on a stand at the Robotics and Automation event

Live demos · Events

Interactive robots at live events

Challenge: Making robotics land with a non-technical crowd — engaging, reliable, and interactive on a busy show floor.
What we built: Bespoke, interactive robot demos — Sanbot face detection, an InMoov playing Rock, Paper, Scissors, and more — run reliably on busy show floors.
The result: Crowd-pleasing demos across UK shows — the NEC, MTC Coventry, MPTS London and Milton Keynes (with a BBC appearance). See all our events →

Consultancy

Ongoing robotics consultancy

Challenge: A UK robotics company needed extra hands on real robot integration and ROS 2 work, delivered reliably alongside their own team.
What we built: Steady, hands-on engineering support — robot bring-up, ROS 2 integration and driver work — delivered a few days a month as a trusted external contributor. (Client details kept confidential.)
The result: An ongoing consultancy engagement — the kind of dependable, real-world delivery Burf offers other teams.

An OpenStreetMap street view beside a LiDAR-reconstructed map from the ROS 2 wheelchair

ROS 2 · Outdoor SLAM

Mapping the outdoors on a ROS 2 wheelchair

Challenge: Most SLAM demos are a tidy indoor room. The real test is outside — uneven ground, GPS drift and no walls to lean on.
What we built: A ROS 2 rig on a powered wheelchair that records LiDAR, IMU and GPS as it drives, then reconstructs the route into a map by fusing GPS + IMU + LiDAR into a known-pose track. Trips upload and replay in the browser with synced video, map and sensor charts.
The result: Repeatable outdoor mapping runs you can scrub back through online — the same record → upload → replay pipeline now reused across Burf's robots.

An InMoov humanoid mirroring a person's arm, teleoperated live from a Meta Quest 3

ROS 2 · VR teleop

An InMoov humanoid, driven from a Meta Quest 3

Challenge: The InMoov is an open-source humanoid with its own control software — bringing it onto ROS 2 and driving it intuitively is not off-the-shelf.
What we built: A ROS 2 driver for the InMoov, plus a Meta Quest 3 app that teleoperates it — the robot mirrors your arm movements live from VR.
The result: Natural, low-friction teleoperation of a full humanoid from a consumer VR headset. Watch the demo →

Sophie — a friendly on-screen robot face for a care-home companion

Care companion · In development

Sophie — an AI care-home companion

Challenge: Care homes are stretched, and residents can go long stretches with little company. A companion has to be warm, genuinely simple to use, and safe — never pretending to be a carer.
What we built: Sophie: a friendly on-screen face that chats and has a call-for-help button that alerts staff. She runs on the Afobot — a low-cost screen-and-camera companion — with the conversation handled by our own AI. Always supervised; not a medical device.
Where it is: Built and verified on a real Afobot. We're now looking for a care home to run a small, supported pilot with. Talk to us about a pilot →

Got a robot to build, integrate, or bring to an event? Let's talk.

Book a call