Assistive Navigation System for Visually Impaired Pedestrians

An assistive robotic navigation system that integrates ROS1, RGB-D perception, YOLOv5, SLAM, path planning to support safe navigation for visually impaired pedestrians.

Project Highlights

ROS1 Gazebo Simulation	A* Path Planning & Trajectory Tracking	YOLOv5-Based Tactile Paving Detection

The project integrates ROS1 Gazebo simulation, A-based path planning with discrete PID trajectory tracking*, and YOLOv5-based tactile paving detection into a unified assistive navigation framework for visually impaired pedestrians.

Project Overview

The objective of this project is to develop an assistive navigation system capable of providing safe and intuitive guidance for visually impaired pedestrians.

The system combines environmental perception, localization, navigation planning, and human-centered voice guidance into a unified robotic framework.

System Architecture

RGB-D Camera
      │
      ▼
ROS Middleware
      │
      ├── Environment Perception
      │      ├── YOLOv5 Detection
      │      ├── Obstacle Recognition
      │      └── Free Space Estimation
      │
      ├── Localization
      │      ├── SLAM
      │      ├── Occupancy Grid
      │      └── Pose Estimation
      │
      ├── Navigation
      │      ├── Global Path Planning
      │      ├── Local Path Tracking
      │      └── Discrete PID Control
      │
      └── Human Interface
             ├── Voice Guidance
             ├── Hazard Notification
             └── Navigation Assistance

Methodology

Environment Perception

The RGB-D camera provides depth and color information for obstacle recognition and free-space estimation.

Main components include:

• RGB-D perception
• Free-space estimation
• Obstacle recognition
• Environment understanding

Object Detection

YOLOv5 is employed to detect tactile paving blocks and surrounding obstacles in real time.

The detected information assists pedestrian navigation in sidewalk environments.

Localization

SLAM is used to estimate the robot pose while simultaneously constructing an occupancy grid map.

This enables robust localization during autonomous navigation.

Navigation

The navigation module generates a global path using the A* algorithm.

A discrete PID controller tracks the generated path while compensating for trajectory deviations during navigation.

Human Assistance

The planned navigation route is translated into voice instructions to assist visually impaired users.

Voice feedback includes:

• Direction guidance
• Hazard warnings
• Destination assistance

Software Stack

Category	Technologies
Middleware	ROS1
Programming	Python, C++
Perception	OpenCV, YOLOv5
Localization	SLAM
Navigation	A*, Discrete PID
Visualization	RViz, Gazebo
Sensors	RGB-D Camera

Key Contributions

• Developed an integrated assistive navigation framework.
• Implemented RGB-D perception and obstacle recognition.
• Developed tactile paving detection using YOLOv5.
• Implemented global path planning with A*.
• Designed a discrete PID controller for trajectory tracking.
• Integrated voice guidance for user navigation.

Future Work

flowchart TD

A[RGB-D Camera]
B[ROS Middleware]
C[Environment Perception]
D[SLAM & Localization]
E[YOLOv5 Detection]
F[Obstacle Recognition]
G[A* Path Planning]
H[Discrete PID Tracking]
I[Voice Guidance]
J[User Navigation]

A --> B
B --> C
B --> D
C --> E
E --> F
D --> G
F --> H
G --> H
H --> I
I --> J

Loading