Autonomous Track Explorer

An autonomous wall-following robot built on the TM4C123 (ARM Cortex-M4) microcontroller. The car uses three Sharp IR distance sensors and hardware PWM to navigate a walled track — centering itself between two walls during normal travel, arcing through curves, and pivoting in place to find a new path when cornered.

---

Hardware

Component	Description
Microcontroller	TI TM4C123GH6PM (ARM Cortex-M4, 80 MHz)
Distance Sensors	3x Sharp GP2Y0A21YK0F analog IR sensors
Motors	2x DC motors controlled via hardware PWM
Motor Driver	H-bridge interfaced through PWM0
Indicators	Onboard RGB LED (Port F)
Controls	Onboard push buttons SW1 / SW2 (Port F)

Sensor Placement

  --------- wall ---------
        \    |    /
    [L]  \  [F]  /  [R]
          [Robot]
             |
           
  --------- wall ---------

• PE1 — Forward-facing sensor (collision detection)
• PE4 — Right diagonal sensor (right wall distance)
• PE5 — Left diagonal sensor (left wall distance)

---

How It Works

Startup

• SW1 starts the car, SW2 stops it
• On startup, the median filter is primed with 9 sensor reads before motion begins to ensure stable initial readings

Steering Logic

The steering() function runs every main loop iteration and evaluates the following cases in priority order:

Priority	Condition	Behavior
1	Ahead sensor reads > 20cm (open space)	Pivot in place toward closer wall to reorient
2	All three sensors < 80cm (boxed in)	Stop both motors, LED blue
3	Ahead + right blocked, left open	Arc right (left wheel only)
4	Ahead + left blocked, right open	Arc left (right wheel only)
5	Normal driving	Proportional centering (see below)

Proportional Centering

During normal driving, the car computes the difference between left and right sensor readings and applies a three-tier correction:

|diff| < 500        → Go straight (equal power to both wheels)
|diff| < 1200       → Gentle correction (60% duty on the wider side)
|diff| >= 1200      → Aggressive correction (90% duty on the wider side)

A higher ADC value means a closer object, so if left_dist > right_dist, the left wall is closer and the car steers right by increasing power to the right wheel.

Sensor Filtering

Raw ADC readings from the GP2Y0A21YK0F sensors contain occasional erroneous voltage spikes. A median filter (y(n) = median(x(n), x(n-1), x(n-2))) is applied to all three sensor channels every loop iteration to reject outliers before they affect steering.

LED Status

LED Color	Meaning
Red	At least one sensor reads beyond 10cm (open space nearby)
Green	All sensors within 10cm (tightly enclosed, well centered)
Blue	All sensors within 80cm (normal driving range)
Off	Car is stopped (SW2 pressed)

---

ADC Distance Calibration

The GP2Y0A21YK0F outputs an analog voltage inversely proportional to distance. ADC values were calculated from datasheet voltages and tuned experimentally:

Distance	Voltage	ADC Value
10 cm	~2.6 V	3200
15 cm	1.8 V	2400
20 cm	1.39 V	1780
30 cm	0.9 V	1190
40 cm	0.74 V	860
80 cm	0.4 V	700

---

Project Structure

TrackExplorer/
├── TrackExplorer.c     # Main file: system init, main loop, steering logic
├── Motors.c / .h       # Hardware PWM setup and duty cycle control
├── Sensors.c / .h      # ADC init and FIR/IIR/median filter implementations
├── LEDSW.c / .h        # RGB LED and push button GPIO configuration
├── PLL.c / .h          # Phase-locked loop: sets bus clock to 80 MHz
└── tm4c123gh6pm.h      # TM4C123 register definitions

---

Build & Flash

1. Open TrackExplorer.uvprojx in Keil µVision 5
2. Build the project (F7)
3. Connect the TM4C123 LaunchPad via USB
4. Flash using the onboard ICDI debugger (F8)
5. Press SW1 on the board to start the car

---

Course Context

Developed for CECS 347 – Microprocessor-Based System Design at California State University, Long Beach. Based on starter code originally authored by Dr. Daniel Valvano and Jonathan Valvano (UT Austin), with steering logic and control algorithm implemented independently.