RC / Autonomous Car (Raspberry Pi)

A Raspberry Pi–based RC car that supports manual driving, collision-guarded driving, and basic autonomous navigation, built as a learning and experimentation platform for embedded systems and robotics.

This project started as a simple RC car and evolved into a multi-mode robotic system with real sensor integration, safety logic, and clean software structure.

The goal is not perfection — the goal is understanding, iteration, and reuse.

---

What This Project Does (At a Glance)

• Manual RC driving using an Xbox controller (Bluetooth)
• Collision-guarded driving using TF-Luna LiDAR
• Autonomous “Roomba-style” obstacle avoidance
• Clean separation between:
  • motor control
  • sensor input
  • control logic
• Designed so future versions can be built smaller or expanded without rethinking everything

---

Current Modes

1. Manual Mode

• Pure tank drive
• Left stick → left motors
• Right stick → right motors
• No sensor interference

2. Guard Mode

• Manual driving with safety
• TF-Luna monitors forward distance
• Behavior:
  • Far away → full speed
  • Getting close → gradual slowdown
  • Too close → forward motion blocked
• Reverse and steering away always allowed

3. Autonomous Mode

• No joystick driving required
• Behavior loop:
  • Drive forward
  • Detect obstacle
  • Reverse briefly
  • Turn randomly
  • Continue forward
• Simple, reliable, and expandable

---

Controller Controls (Xbox)

Button	Action
A	Arm / Disarm motors
B	Emergency stop (instant disarm)
X	Cycle modes (Manual → Guard → Auto)
LB	Decrease stop distance
RB	Increase stop distance
Y	Currently unused (reserved for future features)

---

Hardware Overview

• Controller: Raspberry Pi 3 B+
• Motors: 4× DC motors (tank drive)
• Motor Driver: SN754410 H-Bridge
• Sensor: TF-Luna LiDAR (UART)
• Control: Xbox controller over Bluetooth
• Power: External battery for motors, Pi powered separately

---

Why the Architecture Matters

This project intentionally separates responsibilities:

• Motor layer:
  Handles GPIO, PWM, and direction logic only

• Sensor layer:
  TF-Luna runs in its own thread and continuously streams distance data
  (prevents stale readings and blocking behavior)

• Control layer:
  Decides what the car should do based on:

  • mode
  • joystick input
  • sensor data

This structure is what made the system:

• stable
• responsive
• easy to extend

---

Key Lessons Learned

• Sensors should never block the main control loop
• LiDAR must be read continuously (threaded) to avoid stale data
• Static motor friction requires higher initial PWM (“kick”)
• Guard logic must fail safe, not optimistic
• Button mappings vary — always confirm them programmatically
• Good architecture matters more than fancy algorithms

---

Project Status

This project is functionally complete for its current goals.

Things that work reliably:

• Manual driving
• Collision guard
• Autonomous navigation
• Mode switching
• Emergency stop
• Live parameter tuning

---

Future Ideas (Optional)

These are nice-to-have, not requirements:

• Save tuning parameters to a config file
• On-screen telemetry (HUD)
• Camera-based perception
• Multiple LiDARs or sensor fusion
• Smaller chassis using the same control logic
• Cleaner module separation (motor.py, lidar.py, etc.)

---

Why This Project Exists

This project exists to:

• learn embedded systems by building, not just reading
• create something reusable for future robotics projects
• understand how software architecture affects real hardware
• have a reference system that “just works” when starting something new

---