A ROS 2‑based autonomous navigation stack for scaled vehicles that fuses LiDAR and depth‑camera data to generate SVM‑inspired virtual barriers and steer safely around obstacles.

This project implements a real‑time obstacle detection and avoidance system on a 1:10 RC vehicle (MacAEV) using:

• Sensor fusion of a 2D SLAMTEC RPLIDAR A2M8 and an Intel RealSense D435 depth camera
• Ground‑plane filtering (RANSAC) and point‑cloud replacement to merge camera and LiDAR ranges
• Convex optimization to compute parallel virtual barriers that maximize clearance
• Finite‑state control (normal, backup, turn) with a PD‑based steering controller for smooth navigation

1. RealSense Node publishes /camera/depth and camera info
2. RPLIDAR Node publishes /scan
3. VESC Driver & Odom Node handle motor commands and odometry
4. Ackermann Mux Node routes manual or autonomous /drive commands
5. Nav Node subscribes to /scan, /camera/depth, and /odom, runs fusion & control, and publishes /drive

1. Ground‑Plane Filtering
   • Apply RANSAC on the depth image to remove ground points
2. Frame Transformation
   • Transform remaining depth points into the LiDAR frame
3. Data Replacement
   • For each beam θᵢ, if camera range r_cam < LiDAR range r_lidar, replace r_lidar with r_cam

This yields a fused 1D scan that captures obstacles both above and below the LiDAR plane.

• Barrier Parameters
  • Compute heading θ_head as weighted average of ranges
  • Define left (α_l, β_l) and right (α_r, β_r) sectors
• Optimization
  Solve: $$\begin{aligned} \min_{w,b}\ &\frac{1}{2}\|w\|^2 \\[6pt] \text{s.t.}\quad & w^\top p_i + b \ge 1, && \forall\,p_i \in \mathcal{P}_{\mathrm{left}},\\\ & w^\top p_j + b \le -1, && \forall\,p_j \in \mathcal{P}_{\mathrm{right}}. \end{aligned}$$ This maximizes the margin (sum of left/right distances) between the two obstacle sets
• Finite‑State Machine
  • Normal: drive forward, steer to maximize clearance (PD on d_lr)
  • Backup: reverse when blocked
  • Turn: pivot in place if needed

• Chassis: 1:10 RC Car (ARMA Granite 4×4)
• Compute: NVIDIA Jetson Orin Nano AI computer
• Sensors:
  • SLAMTEC RPLIDAR A2M8 (2D LiDAR)
  • Intel RealSense D435i (RGB‑D camera)
  • BNO055 9‑DOF IMU
  • TRAMPA VESC 6 MKV motor driver
• Power: VENOM 3S LiPo battery
• Software: Ubuntu 20.04, ROS 2 Foxy, C++17

1. Install dependencies

   sudo apt update
   sudo apt install \
     ros-foxy-ros-base \
     ros-foxy-ackermann-msgs \
     ros-foxy-rplidar-ros
   

2. To run the full system:

   ros2 launch auto_nav nav_launch.py