Shared-World Model for Autonomous Vehicle Sensor Fusion
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous vehicles face limitations in detecting objects beyond the range of their on-board sensors or those occluded by obstructions, which can lead to incomplete perception of the environment and potential safety hazards.
Innovation Solution
The method involves fusing on-board sensor data from multiple nearby vehicles to create a shared-world model, which includes absolute locations of vehicles and objects. This model determines detection-range overlap between vehicles and represents overlapping objects as a single entity, enhancing the perception of the environment.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If on-board sensors are used for object detection, then detection accuracy within sensor range is improved, but detection range is limited to several hundred meters
Solution Approach 1:
The patent combines sensor data from multiple connected vehicles to create a shared-world model. By merging detection data from vehicles at different locations, the system achieves extended effective detection range while maintaining detection accuracy, resolving the contradiction between limited sensor range and need for wide coverage.
Solution Approach 2:
The patent introduces a communication network and data-processing center as intermediaries between vehicles and the environment. These intermediaries enable vehicles to access detection information from other vehicles beyond their own sensor range, effectively extending the detection capability without modifying physical sensors.
2Speed
If on-board sensors are used for object detection, then real-time detection capability is improved, but occlusion by nearby vehicles causes detection failure
Solution Approach 1:
The patent merges detection data from multiple vehicles to compensate for occlusion. When one vehicle's sensors are blocked by another vehicle, the system combines this with data from other vehicles that have line-of-sight to the occluded objects, maintaining complete environmental awareness without sacrificing real-time detection capability.
3Loss of information
If sensor data from multiple vehicles is fused, then environmental perception capability is improved, but data processing complexity increases
Solution Approach 1:
The patent extracts only the essential detection information (object locations, distances, and identities) from raw sensor data for fusion in the shared-world model. By taking out only the critical data elements needed for environmental perception rather than processing complete raw sensor streams, the system reduces processing complexity while maintaining perception completeness.
Data Source
AI summary
A system receives, from a plurality of connected vehicles, respective GPS data indicating an absolute location of a respective connected vehicle and on-board sensor data indicating locations of nearby vehicles and objects relative to the respective connected vehicle. The system processes the data to create a shared-world model that includes at least the locations of the plurality of connected vehicles and the nearby vehicles and objects. Some implementations of the shared-world model further include the speeds and trajectories of each of the plurality of connected vehicles and the nearby vehicles and objects. The system transmits a representation of the shared-world model to one or more of the plurality of connected vehicles, which may utilize the shared-world model to provide advanced warnings for drivers or to provide improved path planning for autonomous vehicles. Some representations of the shared-world model include lane-level traffic functions such as traffic density, traffic speed, and traffic throughput.