Backend Confidence Mapping for Occlusion-Aware Digital Road Maps
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current digital road maps for autonomous driving systems lack real-time updates and reliability, as they depend on static data and are not effectively adapted to changing environments, leading to potential inaccuracies and safety concerns.
Innovation Solution
An object confidence value generation system that utilizes a backend and object recognition devices in vehicles to continuously update digital road maps by detecting and evaluating environmental data, assigning confidence values to objects based on recognition and occlusion, ensuring that only reliable information is maintained, and adjusting these values over time.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If static data is used for digital road maps, then system complexity is reduced, but reliability and real-time accuracy deteriorate
Solution Approach 1:
The system implements feedback mechanisms where detection units in vehicles continuously monitor environmental data and transmit it to the backend. The backend processes this data and updates the digital road map accordingly, creating a closed-loop feedback system that maintains reliability through continuous validation and correction of map data against real-world observations.
Solution Approach 2:
The system enables self-service by allowing vehicles to automatically detect, evaluate, and report environmental data without human intervention. The backend automatically processes this data, updates confidence values, and maintains the digital road map, eliminating the need for manual map updates and reducing system complexity while improving reliability.
2Measurement precision
If real-time updates from multiple vehicles are integrated, then accuracy and up-to-date information improve, but data processing complexity and computational load increase
Solution Approach 1:
The backend serves as an intermediary that receives, processes, and integrates data from multiple vehicle detection units. It consolidates environmental data, evaluates confidence values, and maintains the digital road map, thereby managing data processing complexity centrally while enabling accurate real-time updates across the network.
Solution Approach 2:
The system dynamically changes parameters such as confidence values based on the quality and consistency of received environmental data. By adjusting confidence values according to detection reliability, the system can weigh different data sources appropriately, improving accuracy while managing computational load through parameter-based filtering and prioritization.
3Reliability
If confidence values are continuously adjusted based on detection data, then reliability of digital road map improves, but computational resources and processing time increase
Solution Approach 1:
The system applies partial action by selectively updating confidence values only for objects and regions where new environmental data is received and evaluated. Rather than continuously recalculating all confidence values, the backend performs targeted updates based on actual data changes, reducing computational resource consumption while maintaining reliability for relevant map elements.
4Measurement precision
If occlusions are detected and excluded from confidence value reduction, then accuracy of object presence determination improves, but complexity of data evaluation increases
Solution Approach 1:
The system applies local quality by treating occluded regions differently from unobserved regions. When occlusions are detected, the evaluation unit specifically excludes only the affected local areas from confidence value reduction, while maintaining normal confidence updates for unoccluded areas. This localized approach improves accuracy for occluded objects without requiring complex global reevaluation.
Data Source
Figure 1~3
Figure 4~6
AI summary
The invention relates to a system for generating confidence values for objects (31) in a digital street map (3) on a back end. To this end, the system comprises a back end (2) and an object recognition device (1) for a vehicle (4). The object recognition device (1) also comprises a detection unit (13), an evaluation unit (10), a positioning unit (11) and a transceiver unit (12). The evaluation unit (10) is designed to recognise the objects (31) and elements masking (21) the objects (31) in general situation data, and to provide them with position information. Said data is then sent to the back end (2). The back end (2) is designed to receive the data and to produce or update the digital road map (3). Furthermore, the back end (2) is designed to modify a confidence value of the object according to the received data.