Obstacle avoidance for autonomous land vehicle navigation in indoor environments by quadratic classifier

Area of Science:

• Robotics and Artificial Intelligence
• Computer Vision
• Autonomous Systems

Background:

• Autonomous land vehicles (ALVs) require robust navigation systems for indoor environments.
• Obstacle detection and avoidance are critical challenges in ALV path planning.
• Existing methods may struggle with the dynamic and complex nature of indoor corridors.

Purpose of the Study:

• To develop and validate a vision-based obstacle avoidance system for ALVs in indoor corridors.
• To utilize a pattern recognition approach for identifying collision-free paths.
• To demonstrate the feasibility of the proposed method on a real ALV.

Main Methods:

• A vision-based approach employing a quadratic classifier for pattern recognition.
• Treating detected obstacles and ALV body sides as distinct patterns.
• Developing a systematic method to classify patterns for path planning.
• Defining the classifier's decision boundary as a local collision-free path.

Main Results:

• Successfully identified and classified obstacles within indoor corridor environments.
• Designed a quadratic classifier using pattern separation techniques.
• The classifier's decision boundary provided a viable local collision-free path.
• Real-world implementation on an ALV demonstrated successful navigation.

Conclusions:

• The proposed vision-based system effectively enables autonomous land vehicle navigation in indoor corridors.
• Quadratic classification of visual patterns provides a reliable method for obstacle avoidance.
• The approach is feasible and successful for real-world ALV applications.