Fully Autonomous ROS2 Kid's Electric Jeep

UC San Diego — MAE 148 · Fall Quarter 2025 (September – December)

Overview

The goal of this project was to transform a full-size kid's electric Jeep into a ROS2-controlled autonomous robot platform, providing a larger robot car base for future MAE 148 final projects. All stock electronics besides the drive motors and batteries were replaced with custom robotics components. The result is a vehicle that is fully controllable via ROS2, with closed-loop steering, skid steering, forward/reverse drive, and a hardware emergency stop.

RoboJeep full car — The completed RoboJeep platform.

Team

Anderson Compalas — ECE
Jobanpreet Mutti — MAE
Saul Armenta — ECE
Bastian Müllner — visiting student from TUM (UPS)

Accomplishments

Full ROS2 control stack — steering angle (closed-loop), skid steering, and throttle
Closed-loop steering with AS5600 absolute encoder and P-controller
Encoder hardware installed on all four wheel motors (AS5600 Hall-effect sensors)
Hardware E-stop relay wired directly to the RC remote, bypassing the Jetson Nano — safe operation at all times
Complete serial bridge between Jetson Nano and dual Arduino Mega 2560s

Hardware

⚠ Warning: Charge the battery regularly — there is no undervoltage protection. It should always stay above 22.5 V. Pull out a power cable at the 24 V splitter and use a multimeter to check.

ℹ E-Stop: The emergency stop is the top-right button on the remote; the green light indicates it is triggered. All H-Bridges require a 5 V enable signal to move. If nothing moves, check that the E-stop is not triggered (red LED on relay board = triggered). All enable wires are yellow.

Electronics

Compute: Nvidia Jetson Nano
Motor control: 2× Arduino Mega 2560 — drive H-bridges, read two wheel encoders each, read steering encoder
Encoder reading: 2× Arduino Micro Pro — one per remaining wheel encoder (AS5600 I²C address conflict workaround)
Motor drivers: 5× H-Bridge (MTDELE)
Wheel feedback: 4× AS5600 Hall-effect encoders
Steering feedback: REV-11-1271 absolute encoder
RC receiver: Radiomaster receiver → Arduino Micro Pro (USB joystick via ELRS Controller) → Jetson Nano
Safety: Relay board cuts 5 V enable to all H-bridges when E-stop is triggered

Wiring

Mechanical

Custom 3D-printed brackets were designed to mount all electronics. A printed steering spacer was also added to take up manufacturing tolerances in the steering assembly, significantly improving steering accuracy and repeatability.

Software

Quick Start

1. SSH into the Jetson

ssh -Y jetson@ucsdrobojeep-148-01.local

2. Start Docker container

docker start test_container
docker exec -it test_container bash

3. Build and launch (custom bash functions pre-loaded in ~/.bashrc)

# Build the workspace
build_robojeep

# Launch all nodes
launch_robojeep

The left joystick controls steering; the right joystick controls throttle and skid steering.

ROS2 Package Structure

robojeep_ws/
├── src/
│   ├── robojeep_msgs/            # Custom message definitions
│   ├── robojeep_base/            # Control logic, Ackermann + tank steering
│   ├── robojeep_arduino_bridge/  # Serial communication with Arduinos
│   └── robojeep_bringup/         # Launch files and robojeep.yaml config

Debugging Topics

ros2 topic echo /joy              # Raw joystick input
ros2 topic echo /cmd_vel          # Velocity commands
ros2 topic echo /steering_cmd     # Steering angle commands
ros2 topic echo /wheel_cmd        # Wheel velocity commands
ros2 topic echo /serial_status    # Formatted serial commands to Arduino

Challenges

Encoder speed: Reading encoders fast enough to capture motor RPM — optimized up to ~9,000 RPM, short of the 13,500 RPM motor maximum
Encoder I²C conflicts: AS5600 sensors share a fixed I²C address, requiring a separate Arduino per sensor
Stripped steering gears: Gears inside the steering motor assembly were stripped mid-project; salvaged replacement gears from the bed-lift motor
Serial latency: Passing encoder feedback from Arduinos to the Jetson in real time required careful serial protocol design

Demo Videos

Full system demo

Steering closed-loop control

Gallery

Front electronics (without battery)	Rear electronics
AS5600 wheel encoder mount	REV absolute steering encoder

Future Work

Battery voltage sensor with automatic cutoff below 22.5 V
Artificial differential — compute per-wheel speed from steering angle during turns
Wire existing dash and steering wheel switches into the Arduino for software-accessible inputs
Full autonomous laps using camera or LiDAR once the base platform is stable

← Back to Portfolio