🏎️ Autonomous Wheeled Mobile Robot

MCT443 – Design of Autonomous Systems

---

📖 Project Overview

This project presents the design and simulation of an autonomous wheeled mobile robot capable of lane keeping and reactive obstacle avoidance, developed using ROS 2 Jazzy and Gazebo Harmonic.

The robot autonomously navigates a 10-meter track, avoids static obstacles through lane transitions, and detects a color-coded finish line to complete its mission. The work demonstrates a complete robotics pipeline from SolidWorks CAD design to real-time feedback control in simulation.

---

🏗️ System Architecture

The system follows a modular Sense–Think–Act architecture:

• Sense: Camera and GPU LiDAR sensors
• Think: Perception algorithms and finite state machine
• Act: Feedback control and differential drive actuation

---

🛰️ Core Modules

🔩 Hardware Layer

• URDF model derived from SolidWorks CAD
• Integrated with ros2_control using ign_ros2_control
• Accurate mass and inertia modeling

👁️ Perception Layer

Lane Detection

• OpenCV moment-based centroid tracking of the black lane

Obstacle Detection

• GPU LiDAR (ultrasonic-style range filtering)

Finish Line Detection

• HSV-based segmentation for green line recognition

🧠 Decision Layer

• 5-state Finite State Machine (FSM)
• Handles lane following, obstacle avoidance, and recovery

🎮 Control Layer

• Proportional (P) Controller
• Steering correction based on visual lateral error

---

🔄 Technical Workflow

1️⃣ CAD to URDF Pipeline

• Designed in SolidWorks for mechanical realism
• Geometry simplification for stable physics simulation
• Exported as STL meshes
• Manual migration to ROS 2 Jazzy / Gazebo Harmonic
• Added modern sensor plugins and ros2_control tags

---

2️⃣ Control Strategy (P-Controller)

Lateral Error Definition

Let the image width be ( W_{\text{image}} ) and the detected lane centroid be ( C_x(t) ):

$$ e(t) = \frac{W_{\text{image}}}{2} - C_x(t) $$

Control Law

$$ \omega(t) = K_p \cdot e(t) $$

Parameters

• Proportional gain:

K_p = 0.005

• Constant linear velocity:

v_x = 0.2 m/s

---

🧠 Obstacle Avoidance (FSM)

State	Action	Trigger Event
FOLLOW_LANE_1	PID lane tracking	Obstacle distance < 0.5 m
SWITCH_LEFT	Open-loop arc maneuver	Timer > 4.0 s
FOLLOW_LANE_2	Adjacent lane tracking	Obstacle distance < 0.5 m
SWITCH_RIGHT	Open-loop return maneuver	Lane / finish visible
FINISH	Deceleration and stop	Green area > threshold

---

▶️ Running the Simulation

Launch Gazebo World

ros2 launch autonomous_rob auto_rob.launch.py world:=project_world.sdf

### Start Autonomous Controller
ros2 run autonomous_rob phase2_node

🛠️ Built With

SolidWorks – Mechanical Design & CAD

ROS 2 Jazzy – Middleware & Control

Gazebo Harmonic – Physics Simulation

OpenCV – Computer Vision & Perception

✅ Conclusion

This project demonstrates a complete autonomous navigation pipeline for a differential drive robot. The modular ROS 2 architecture enabled robust lane following, reliable obstacle avoidance, and accurate mission completion in simulation.