Vehicle Uncertainty Maps for Low-Visibility Object Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current vehicle systems face challenges in accurately identifying and tracking objects in their environment, especially in uncertain conditions such as inclement weather or low visibility, which can impact safe and efficient operation, particularly in autonomous or semi-autonomous modes.
Innovation Solution
A method utilizing a deep neural network to process sensor data from vehicles, combining uncertainty data from multiple vehicles of the same type and environmental conditions to create a hotspot map that includes object locations and identification probabilities, which is then downloaded to vehicles to improve object detection and avoidance strategies.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If vehicle systems rely on individual sensor data and object detection algorithms, then each vehicle can operate independently, but object detection accuracy deteriorates in uncertain conditions such as inclement weather or low visibility
Solution Approach 1:
The patent combines uncertainty data from multiple vehicles of the same type operating in similar environmental conditions to create a comprehensive hotspot map. This merging of data from multiple sources improves object detection accuracy by providing a more complete picture of uncertain regions that individual vehicles cannot detect reliably on their own.
Solution Approach 2:
The patent introduces a server as an intermediary that collects, processes, and distributes uncertainty data and hotspot maps to vehicles. This intermediary enables vehicles to access aggregated uncertainty information from the fleet without direct peer-to-peer communication, resolving the contradiction by providing a centralized mechanism for information sharing.
2Reliability
If vehicles share uncertainty data across the fleet, then object detection accuracy improves, but data processing and communication complexity increases
Solution Approach 1:
The server acts as an intermediary that handles the complex tasks of collecting, processing, and distributing uncertainty data, thereby reducing the processing burden on individual vehicles. The server consolidates data from multiple vehicles, generates hotspot maps, and distributes them back to the fleet, simplifying the overall system architecture despite the increased data sharing.
Solution Approach 2:
The server performs multiple functions including data collection, uncertainty analysis, hotspot map generation, and data distribution to different vehicle types. This multi-functional approach consolidates complexity into a single centralized system rather than distributing it across multiple vehicles, reducing individual vehicle complexity while enabling fleet-wide improvement.
3Measurement precision
If vehicles download and process hotspot map data, then object detection in challenging conditions improves, but communication bandwidth and energy consumption increase
Solution Approach 1:
The server generates and prepares hotspot maps in advance based on aggregated uncertainty data from the fleet, so vehicles receive pre-processed information rather than raw data requiring extensive local processing. This preliminary action reduces the computational energy required in vehicles while maintaining improved object detection accuracy.
Solution Approach 2:
Vehicles receive copies of hotspot maps from the server rather than independently generating their own uncertainty analyses. This copying approach allows vehicles to benefit from fleet-wide data aggregation without duplicating the expensive processing required to generate such maps, reducing individual vehicle energy consumption while improving measurement precision.
Data Source
AI summary
A computer, including a processor and a memory, the memory including instructions to be executed by the processor to, based on sensor data in a vehicle, determine a database that includes object data for a plurality of objects, including, for each object, an object identification, a measurement of one or more object attributes, and an uncertainty specifying a probability of correct object identification, for the object identification and the object attributes determined based on the sensor data, wherein the object attributes include an object size, an object shape and an object location. The instructions include further instructions to determine a map based on the database including the respective locations and corresponding uncertainties for the vehicle type and download the map to a vehicle based on the vehicle location and the vehicle type.