The $50 figure excludes the smartphone, and the AI features need a reasonably recent Android phone to run at a useful speed.
A small four-wheel 3D-printed car that uses an ordinary smartphone as its computer, camera and sensor suite, with an Arduino Nano handling the motors. Started at Intel's Intelligent Systems Lab and now run by the OpenBot Foundation, it ships Android and beta iOS apps for remote driving, person following, point-goal navigation and training driving policies from logged data.
At a glance
| Parts cost | $50 (2020) project · Stated by the project for the robot body only. The smartphone is not included source ↗ |
|---|---|
| Difficulty | 2/5 Easy · Printed body, off-the-shelf hobby electronics and simple soldering, though training the autopilot and phone compatibility add software friction. |
| Build time | Not stated |
| Tools needed | 3D printer (two body parts plus phone mount, needs a 240x150 mm plate for the regular body; slim and glue-together bodies exist for smaller printers), soldering (motor leads, optional switch and LEDs), screwdriver, optional custom PCB instead of hand wiring an L298N |
| Motors | 4 TT gear motors with wheels, driven by an L298N module or the project's custom PCB |
| Controller | Android smartphone (iOS app in beta) linked over USB OTG to an Arduino Nano |
| Sensors | Phone camera and IMU; optional wheel speed sensors, ultrasonic sensor and OLED display |
| Published | ✓ Parts list✓ CAD/STL✓ Assembly guide✓ Wiring✓ Software✗ Video GitHub READMEs (translated into 6 languages), STL/STEP files, KiCad-style PCB gerbers and parts lists, wiring diagram image |
| Activity | Last update 2026-09-25 · 3,516 GitHub stars · Slack workspace linked from the README; 673 GitHub forks; last tagged release v0.8.0 (2025-03-03) |
| Kits | Ready-to-run RTR-TT and RTR-520 vehicles are offered by the project (the buy link redirected to a Taobao listing on 2026-10-02) |
Licenses
| Design files | MIT (same repo-wide LICENSE covers the CAD and PCB files in the repo; no separate hardware license found) source ↗ |
|---|---|
| Code | MIT (repo-wide LICENSE, Copyright 2020 Intel ISL) source ↗ |
| Documentation | MIT (same repo-wide LICENSE; no separate docs or image license found) source ↗ |
Builder notes
Practical points from the project's issue tracker, forum and docs, each linked to its source.
- The repo moved from Intel's isl-org account to the ob-f organization, so older links to intel-isl or isl-org redirect there. Recent commits in 2026 rename the Android and iOS app packages to org.openbot. source ↗
- Low-end phones are a common source of trouble. In one early thread the maintainer said a budget Android 7/8 phone could drive the car but would struggle to run the AI models at a usable speed, and another builder saw controller input fail on an older Note 4 phone. source ↗
- Training a working autopilot takes several datasets and correct use of the high-level command input. The longest issue thread shows a builder getting a car that only crawled straight until retraining with more data. source ↗
- The phone-as-controller mode requires both phones on the same local network, and a school team reported that most of their mid-range phones would not pair, with no fixed list of supported models. source ↗
- The prebuilt Android APKs are signed only with a debug key, and the iOS apps currently have to be built from source in Xcode with a Firebase project. source ↗
Other wheeled plans
- Donkey Car · $250–$300 · difficulty 2/5
- LeKiwi · $480 · difficulty 2/5
- XLeRobot · $660 · difficulty 3/5
- F1TENTH (RoboRacer) autonomous race car · $3,000 · difficulty 3/5
- OpenMower · $790 · difficulty 3/5
- Upkie wheeled biped · $4,600 · difficulty 4/5
- TidyBot++ · $5,200–$6,300 (2024) · difficulty 4/5
- NVIDIA JetBot · $640 (2024) · difficulty 2/5
- MuSHR racecar · $1,100 (2022) · difficulty 3/5