A Grid-Based Framework for Collective Perception in Autonomous Vehicles.

Area of Science:

• Automotive Engineering
• Robotics
• Computer Vision

Background:

• On-board sensors for automated vehicles have limited perception due to line-of-sight and occlusions.
• Vehicle-to-Everything (V2X) communication enables cooperative perception among vehicles.

Purpose of the Study:

• To propose a perception framework that fuses data from on-board sensors and Collective Perception messages (CPM).
• To enhance the environmental awareness of automated vehicles by overcoming sensor limitations.

Main Methods:

• Modeling the environment using an occupancy grid to represent occupied, free, and uncertain spaces.
• Calculating independent grids for each sensor (on-board and V2X) and fusing them based on occupancy and confidence.
• Implementing a Particle Filter for object tracking and cell occupancy evolution over time.

Main Results:

• The proposed framework demonstrated good performance in fusing sensor data and CPM.
• Effective perception of static and dynamic objects was achieved even with significant uncertainties and communication delays.
• Validation through experiments using real vehicles and infrastructure sensors.

Conclusions:

• The developed perception framework is viable for Collective Perception applications.
• Fusion of on-board sensors and V2X data significantly improves automated vehicle perception capabilities.
• The approach enhances road safety by providing a more comprehensive understanding of the driving environment.