Crowdsourced Radar Map Layer Filtering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current autonomous driving and advanced driver-assist systems rely on high-definition (HD) maps, but generating and maintaining high-quality HD maps is challenging, especially across different vehicle types and large geographical regions, due to limitations in sensor quality and the need for continuous deployment of fleets with high-quality sensors.
Innovation Solution
The proposed solution involves crowdsourcing techniques where vehicles in a geographical region publish radar data to a server, filtering it frame-by-frame to remove moving objects, and transmitting it only when the confidence metric of the vehicle's position estimate exceeds certain thresholds, enabling accurate HD map creation and updating.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If crowdsourced radar data from multiple vehicles is used to generate HD maps, then map quality and coverage are improved, but data filtering and processing complexity increases
Solution Approach 1:
The patent segments the HD map generation process into distinct layers (radar map layer, camera map layer, etc.) and processes radar data through systematic filtering stages (frame-by-frame filtering, batch filtering) to manage complexity while maintaining map quality
Solution Approach 2:
The patent introduces confidence metrics and reliance metrics as intermediary evaluation mechanisms to automatically assess and filter crowdsourced radar data quality, reducing manual processing complexity while ensuring map accuracy
2Measurement precision
If radar data is transmitted frequently to update HD maps, then map accuracy is improved, but communication bandwidth and energy consumption increase
Solution Approach 1:
The patent implements periodic data transmission based on threshold evaluations, where radar data is transmitted only when confidence metrics or reliance metrics exceed specified thresholds, rather than continuous transmission, thereby reducing energy consumption while maintaining position estimate accuracy
Solution Approach 2:
The system uses feedback mechanisms through confidence metrics and reliance metrics to dynamically determine when data transmission is necessary, creating an energy-efficient loop that maintains accuracy by transmitting only when needed
3Manufacturing precision
If frame-by-frame filtering is applied to remove moving objects, then map accuracy is improved, but processing time increases
Solution Approach 1:
The patent segments the filtering process into frame-by-frame filtering for immediate moving object removal and batch filtering for more comprehensive processing, allowing real-time accuracy while managing overall processing time through distributed computation
Data Source
AI summary
Creating and updating an accurate radar map layer for HD map using crowdsourcing may comprise a vehicle obtaining radar data and filtering the radar data on a frame-by-frame basis. In some embodiments, additional filtering may be made on a batch of frames. The vehicle can then transmit the filtered radar data responsive to a determination that a confidence of a position estimate of the vehicle exceeds a conference threshold level and/or a determination that a reliance of the position estimate of the vehicle on the radar data exceeds a reliance threshold level.