Autonomous Car

Overview

This project challenged our team to design and build a fully autonomous mobile robot for a ROBA-style head-to-head game, capable of navigating a structured competition field, interacting with physical game elements, and operating reliably across multiple match scenarios. The system needed to support both autonomous behaviors and real-time manual intervention, while integrating sensing, control, actuation, and embedded software into a single cohesive platform.

We ultimately designed and built a two-wheel differential-drive autonomous robot powered by an ESP32, integrating Time-of-Flight sensors for wall following, Vive-based localization for absolute position awareness, PID-controlled motor actuation, and a Wi-Fi–based control framework to support seamless switching between autonomous and manual modes. The robot was engineered as a complete mechatronic system, with mechanical design, embedded software architecture, sensing, and control developed in parallel to meet competition constraints.

Our robot competed in a class-wide robotics tournament involving 27 teams, where it successfully system advanced to the quarterfinals!

Code Architecture

At a high level, the system consisted of a centralized state machine, non-blocking control loops, and dedicated sensing and control modules for wall following, absolute-position navigation, and Wi-Fi–based manual control.

Wall-Following Behavior

• Implemented in C++ using three Time-of-Flight sensors (1 front-mounted VL53L0X, 2 right-mounted VL53L4CD) to maintain a consistent lateral offset from field boundaries.

• The two side-mounted sensors were compared differentially to estimate angular misalignment relative to the wall, while the front sensor provided forward distance awareness.

• Distance error relative to a 100 mm target offset was converted into heading corrections applied through PID-controlled differential wheel speeds.

• This approach enabled sustained wall tracking without oscillatory drift, allowing the robot to follow long boundary segments as part of autonomous navigation routines.

  

Vive-Based Autonomous Navigation

• Used two Vive photodiode trackers to provide absolute position and orientation estimates within the competition field.

• Raw position data was filtered using a median filter to reduce jitter before being passed into the control loop.

• Autonomous navigation was executed as a sequence of orientation alignment followed by translational motion toward predefined coordinate targets.

• This structure allowed the robot to navigate to three distinct field objectives (low towers, high towers, Nexus) using closed-loop feedback rather than open-loop timing.

  
  

Wi-Fi–Based Manual Control

• Implemented a Wi-Fi control interface by hosting an HTTP server directly on the ESP32.

• Browser-issued commands were mapped to motor speed, turning, and mode-switching inputs in real time.

• Manual control operated as a first-class state within the system’s state machine, allowing instant transitions between teleoperation and autonomous behaviors without restarting control loops.

• This enabled live operator intervention during competition matches while preserving system stability and responsiveness.