A Review of the Bayesian Occupancy Filter.

Area of Science:

• Robotics and Autonomous Systems
• Computer Vision
• Artificial Intelligence

Background:

• Environment perception is crucial for autonomous vehicles to navigate safely.
• Existing methods for environment perception face challenges in accurately assessing surroundings.
• The Bayesian Occupancy Filter (BOF) has emerged as a key technique for environment occupancy evaluation.

Purpose of the Study:

• To provide a comprehensive review of the Bayesian Occupancy Filter (BOF) and its variants.
• To introduce a novel, detailed taxonomy for the BOF, structuring it into five progressive layers.
• To offer practical insights into BOF applications through implemented use cases.

Main Methods:

• Systematic review of existing literature on Bayesian Occupancy Filter (BOF) methods.
• Development of a hierarchical taxonomy categorizing BOF components from sensor-level to risk-assessment.
• Analysis of real-world use cases demonstrating the practical implementation of BOF and its taxonomy.

Main Results:

• The review consolidates current knowledge on BOF techniques and their evolution.
• A five-layer taxonomy is proposed, offering a structured understanding of BOF from raw sensor data to high-level risk assessment.
• Use case studies illustrate the effectiveness and versatility of the BOF framework in autonomous systems.

Conclusions:

• The Bayesian Occupancy Filter (BOF) is a vital tool for enhancing environment perception in autonomous driving and robotics.
• The proposed taxonomy provides a standardized framework for understanding, developing, and applying BOF methods.
• Further research and application of BOF, guided by the taxonomy, can significantly advance autonomous system capabilities.