Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Enhanced Visual SLAM for Collision-free Driving with Lightweight Autonomous Cars

Zhihao Lin, Zhen Tian, Qi Zhang, Hanyang Zhuang, Jianglin Lan

TL;DR

This work tackles obstacle avoidance for lightweight autonomous cars using a CPU-only, vision-based pipeline. It integrates ORB-SLAM3 enhanced with optical flow for perception and a CLF-CBF-QP-SRP framework for globally safe and stably generated trajectories, supplemented by SRP-obstacle reconstruction and local TEB planning. In simulation, the proposed approach delivers collision-free, smooth, and efficient paths and outperforms benchmark methods in safety and trajectory quality, while maintaining low computational load. The camera-based strategy reduces reliance on heavier sensors, enabling cost-effective indoor autonomous driving with broad practical implications.

Abstract

The paper presents a vision-based obstacle avoidance strategy for lightweight self-driving cars that can be run on a CPU-only device using a single RGB-D camera. The method consists of two steps: visual perception and path planning. The visual perception part uses ORBSLAM3 enhanced with optical flow to estimate the car's poses and extract rich texture information from the scene. In the path planning phase, we employ a method combining a control Lyapunov function and control barrier function in the form of quadratic program (CLF-CBF-QP) together with an obstacle shape reconstruction process (SRP) to plan safe and stable trajectories. To validate the performance and robustness of the proposed method, simulation experiments were conducted with a car in various complex indoor environments using the Gazebo simulation environment. Our method can effectively avoid obstacles in the scenes. The proposed algorithm outperforms benchmark algorithms in achieving more stable and shorter trajectories across multiple simulated scenes.

Enhanced Visual SLAM for Collision-free Driving with Lightweight Autonomous Cars

TL;DR

This work tackles obstacle avoidance for lightweight autonomous cars using a CPU-only, vision-based pipeline. It integrates ORB-SLAM3 enhanced with optical flow for perception and a CLF-CBF-QP-SRP framework for globally safe and stably generated trajectories, supplemented by SRP-obstacle reconstruction and local TEB planning. In simulation, the proposed approach delivers collision-free, smooth, and efficient paths and outperforms benchmark methods in safety and trajectory quality, while maintaining low computational load. The camera-based strategy reduces reliance on heavier sensors, enabling cost-effective indoor autonomous driving with broad practical implications.

Abstract

The paper presents a vision-based obstacle avoidance strategy for lightweight self-driving cars that can be run on a CPU-only device using a single RGB-D camera. The method consists of two steps: visual perception and path planning. The visual perception part uses ORBSLAM3 enhanced with optical flow to estimate the car's poses and extract rich texture information from the scene. In the path planning phase, we employ a method combining a control Lyapunov function and control barrier function in the form of quadratic program (CLF-CBF-QP) together with an obstacle shape reconstruction process (SRP) to plan safe and stable trajectories. To validate the performance and robustness of the proposed method, simulation experiments were conducted with a car in various complex indoor environments using the Gazebo simulation environment. Our method can effectively avoid obstacles in the scenes. The proposed algorithm outperforms benchmark algorithms in achieving more stable and shorter trajectories across multiple simulated scenes.
Paper Structure (23 sections, 23 equations, 10 figures)

This paper contains 23 sections, 23 equations, 10 figures.

Table of Contents

  1. Introduction
  2. RELATED WORKS
  3. Geometric Methods Enhanced SLAM Approach
  4. Trajectory Generation Methods
  5. SYSTEM OVERVIEW
  6. Vehicle Model
  7. Perception Based on Vision
  8. Point Features Matching and Attributes Updating
  9. Outlier Features Removing
  10. Point Features Optimizing Algorithm
  11. Global Path Planning by Using CLF-CBF-QP-SRP
  12. The formulation of CLF
  13. The stability control of moving car using CLF
  14. The formulation of CBF
  15. Safe control of moving car using CBF
  16. ...and 8 more sections

Figures (10)

  • Figure 1: A general depiction of our method. (a) The graph depicts the key points extracted from the scene by ORB-SLAM3. (b) The graph shows the obstacle avoidance path planning performed based on the information provided by ORB-SLAM3. (c) and (d) The graphs display the specific details of the simulated scene.
  • Figure 2: Our System Workflow. This system comprises two main components: environment perception and path planning. Initially, a PGM map and cost map are constructed. The vehicle, equipped with a visual sensor, extracts point features from the environment and uses relocation to ascertain its position and identify obstacles. A static map is inflated for navigational safety. The vehicle's pose is dynamically updated by tracking map points, and a global path is mapped using CBF. For local path planning, the TEB algorithm is employed. The system updates the vehicle's pose in real-time, calculates safe passage areas with CBF, and facilitates optimal, obstacle-free path selection to the destination.
  • Figure 3: The Intel D435i RGB-D camera utilizes the structured light triangulation method for depth sensing.
  • Figure 4: Illustration of the point feature matching process in two frames using grid IDs, Euclidean distance, and cosine similarity to ensure alignment and temporal consistency between consecutive frames.
  • Figure 5: Illustration of SRP for an obstacle.
  • ...and 5 more figures