← Back to the desk SHEET 01 OF 06

01 · ENGR 122

Autonomous navigation robot

A tracked robot that drives the Stevens Smart Arena on its own, steering from overhead camera position data.

Role
Team lead. CAD, wiring and firmware
When
Jan to May 2026
Where
Stevens Smart Arena, team of 3
Tools
SolidWorks, WeMos D1 R2, C++, MQTT, HC-SR04

Notes

  1. 4 of 4 targets in about 1 min, 3 min limit
  2. Printed chassis, 7 revisions in SolidWorks
  3. MQTT position plus atan2 heading control

The problem

The robot had to visit four targets in a set order inside the Stevens Smart Arena, avoid walls and obstacles, and finish in under three minutes with no one touching it. The rules: at least three ultrasonic sensors, an OLED, a 3D printed body under five hours of print time, an ArUco marker the overhead camera could track, and no glue or tape anywhere.

What I did

I designed the printed body in SolidWorks, wired the power and sensor harness, and wrote the navigation code. I ran the team of three.

Exploded view of the robot: ArUco holder, motor control board, OLED, WeMos D1, three HC-SR04 sensors, printed body and Zumo chassis

Final assembly drawing with bill of materials

How it works

The arena camera tracks the ArUco marker and publishes x, y and heading over MQTT. The robot computes the angle to the next target with atan2, takes the shortest turn, and picks a gentle steer or a full pivot depending on how far off it is. It slows down near a target and holds for two seconds when it gets there.

Ultrasonic readings handle obstacles. If the front stays blocked for more than a second, it switches into a wall-follow recovery mode for 2.5 seconds instead of bumping the same box forever. The OLED shows target, position, heading and all three distances live, which made testing much faster.

Wiring diagram: WeMos D1, motor controller, three ultrasonic sensors, OLED, battery, switch and a 5 to 1 power bus

Result

In the final run it reached all four targets in about one minute against a three minute limit. It bumped one obstacle, recovered, and earned full marks for the demonstration.

What I’d change

A fourth sensor to cover the blind spots, the switch mount designed in from the start, and a small custom board instead of jumper wires.