Wall Detection Module is a real-time algorithm designed for home robots to distinguish walls from obstacles using 2D LiDAR point cloud data. Implemented in C++ for efficiency and real-time performance, this module enables robots to:
- Switch to wall-following mode smoothly
- Separate wall-mounted obstacles from walls
- Serve as a foundation for obstacle avoidance strategies
- 🚀 Real-time processing (single-frame LiDAR analysis)
- 📏 Sliding window segmentation with least squares line fitting
- 🔍 Adaptive thresholding for dynamic environments
↔️ Bidirectional segment extension for maximum wall coverage- 📊 Multiple output formats (point clouds, line parameters, visualizations)
- 🧩 Modular design for easy integration with robot navigation systems
graph TD
A[Raw LiDAR Point Cloud] --> B[Preprocessing]
B --> C[Sliding Window Segmentation]
C --> D{Linear Fit Valid?}
D -->|Yes| E[Bidirectional Segment Extension]
D -->|No| F[Next Window]
E --> G[Save Wall Segment]
G --> H[Collision Point Check]
H --> I{Wall Detected?}
I -->|Yes| J[Trigger Wall-following]
I -->|No| K[Obstacle Avoidance]
- C++17 compatible compiler
- Eigen3 library (for matrix operations)
- Opencv (optional for visualization)
- 2D LiDAR sensor (e.g., RPLidar A1, Slamtec T5)
# Clone repository
git clone https://github.com/your-org/wall-detection-module.git
cd wall-detection-module
# Build with CMake
mkdir build && cd build
cmake .. -DCMAKE_BUILD_TYPE=Release
make -j4
# Integrate into your project
target_link_libraries(your_project PRIVATE WallDetection)#include "wall_detector.h"
// Initialize with sensor parameters
WallDetector detector(
0.01f, // max_fit_variance (m²)
0.8f, // min_span_distance (m)
0.05f, // min_distance (m)
0.15f, // max_distance (m)
20 // segment_window
);
// Process LiDAR frame
std::vector<Point2D> lidar_points = getLidarData();
Point2D collision_point = getCollisionPoint();
// Detect walls
bool is_wall = detector.isWallObstacle(lidar_points, collision_point);
// Get detailed results
WallDetectionResult result = detector.getDetailedResult();| Parameter | Default | Description |
|---|---|---|
max_fit_var |
0.01 m² | Maximum allowed fitting variance |
min_span_distance |
0.8 m | Minimum wall segment length |
min_distance |
0.05 m | Minimum point distance threshold |
max_distance |
0.15 m | Maximum point continuity distance |
segment_window |
20 points | Sliding window size |
- Boolean wall flag:
Trueif collision point belongs to wall - Wall segments: List of line parameters (slope, intercept, variance)
- Visual markers:
- Green lines: Detected wall segments
- Red points: Noise points
- Blue markers: Collision point context
- ⏱️ Processing time: < 5ms per frame
- 📈 Accuracy: 92% across 15 test scenarios
- 💾 Memory: < 2MB heap usage
- Multi-frame temporal fusion
- Curved wall detection (spline fitting)
- Multi-sensor fusion (RGB-D + LiDAR)
- Hardware acceleration support
Apache 2.0 License. See LICENSE for details.
Developed with ❤️ by ZXJ | Wednesday, May 28, 2025