A Neural Network Approach for Building An Obstacle Detection Model by Fusion of Proximity Sensors Data

Area of Science:

• Robotics
• Artificial Intelligence
• Sensor Technology

Background:

• Proximity sensors are crucial for mobile robot obstacle detection.
• Traditional sensor calibration is manual, time-consuming, and sensor-specific.
• Existing models are nonlinear and sensitive to sensor type, environment, and obstacle properties.

Purpose of the Study:

• To develop a unique, accurate obstacle detection model for mobile robots.
• To overcome the limitations of traditional manual calibration methods.
• To leverage sensor fusion and artificial neural networks for improved performance.

Main Methods:

• Utilizing artificial neural networks (ANNs) for automatic calibration.
• Implementing sensor fusion techniques to integrate data from diverse proximity sensors.
• Developing a unified obstacle detection model.

Main Results:

• Achieved automatic calibration of proximity sensors.
• Created a single, robust obstacle detection model from fused sensor data.
• Demonstrated a more efficient and potentially more accurate approach to obstacle detection.

Conclusions:

• Sensor fusion combined with ANNs offers an effective solution for mobile robot obstacle detection.
• Automatic calibration via ANNs simplifies and enhances the traditional process.
• The proposed method addresses the limitations of individual sensor models and manual calibration.