Vehicle Sensor Data Sharing for Occluded Scene Perception
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous vehicles face challenges in perceiving their environment due to insufficient sensor data, particularly in regions obstructed by external objects or when sensor redundancy is low, leading to reduced confidence in navigational actions and increased collision risks.
Innovation Solution
The method involves forming a local ad hoc wireless network between autonomous vehicles to share and request supplemental perception data, allowing vehicles to broadcast and receive raw or derived sensor data to enhance their perception of the scene, thereby improving sensor redundancy and confidence in navigational decisions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If autonomous vehicles rely solely on their own sensors for environmental perception, then device complexity is reduced, but measurement precision and reliability of scene perception deteriorate due to insufficient sensor data in obstructed regions
Solution Approach 1:
The patent merges sensor data from multiple autonomous vehicles to create a composite perception of the environment. Vehicles share sensor data through wireless communication, combining their individual sensor readings to achieve more complete and accurate scene perception, particularly in regions where individual vehicles have obstructed views or insufficient sensor coverage.
Solution Approach 2:
The patent introduces wireless communication as an intermediary mechanism between autonomous vehicles. This intermediary enables the exchange of sensor data without requiring direct physical sensor contact or complex integrated sensor systems, allowing vehicles to obtain supplemental perception data from nearby vehicles while maintaining operational independence.
2Reliability
If autonomous vehicles increase sensor redundancy through data sharing, then reliability of navigational decisions improves, but loss of time for data processing and communication increases
Solution Approach 1:
The patent implements preliminary action by having autonomous vehicles continuously collect and buffer sensor data before it is needed for navigation decisions. Vehicles maintain ready-to-share sensor data in memory, so when a vehicle requests supplemental data for an obstructed region, the receiving vehicles can immediately provide pre-captured sensor readings without requiring real-time data collection or extensive processing during the critical decision moment.
3Measurement precision
If autonomous vehicles broadcast queries for supplemental perception data, then measurement precision in obstructed regions improves, but device complexity and energy consumption increase due to wireless communication infrastructure
Solution Approach 1:
The patent applies local quality by having autonomous vehicles broadcast queries and share data only within local geographic regions where vehicles are proximal to one another. Instead of attempting global data exchange, the system limits communication to nearby vehicles that are likely to have relevant supplemental sensor data for obstructed regions, reducing unnecessary energy consumption from wide-area communication attempts.
Data Source
AI summary
One variation of a method for accessing supplemental data from other vehicles includes, at an autonomous vehicle: recording a scan image of a scene around the autonomous vehicle at a first time; detecting insufficient perception data in a region of the scan image; in response to detecting insufficient perception data in the region, defining a ground area of interest containing the region and wirelessly broadcasting a query for perception data representing objects within the ground area of interest; in response to receiving supplemental perception data—representing objects within the ground area of interest detected by the second vehicle at approximately the first time—from a second vehicle proximal the scene, incorporating the supplemental perception data into the scan image to form a composite scan image; selecting a navigational action based on objects in the scene represented by the composite scan image; and autonomously executing the navigational action.