Decentralized Object Presence Prediction Using Map-Based Data Compression
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing decentralized learning methods primarily focus on short-term situational awareness for vehicles, failing to provide long-term awareness that accounts for the presence of objects beyond the range of their own sensors.
Innovation Solution
The system employs decentralized data collection and city map data-based information compression to predict the presence of objects. This involves obtaining a 2-dimensional matrix representing object presence, filtering it based on map information, converting it to 1-dimensional data, and inputting this data into a trained prediction machine learning model to forecast future object presence.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If decentralized learning is used to enable long-term situational awareness, then vehicles can perceive environment beyond sensor range, but communication complexity and data processing requirements increase significantly
Solution Approach 1:
The patent extracts only the essential information needed for prediction (object presence in grid cells) from the full sensor data, separating critical spatial-temporal patterns from redundant details. This extraction approach enables long-term awareness while minimizing communication overhead by transmitting only compressed grid representations rather than raw sensor streams.
Solution Approach 2:
The patent transforms continuous sensor data into discrete grid cell representations with binary presence/absence parameters. This parameter transformation simplifies the data structure from complex multi-dimensional sensor readings to manageable grid matrices, reducing communication complexity while preserving essential spatial-temporal information for prediction.
2Measurement precision
If full 2-dimensional object presence data is transmitted and processed, then prediction accuracy is maintained, but computational resources and communication bandwidth are consumed
Solution Approach 1:
The patent segments the continuous monitoring area into discrete grid cells, dividing the large-scale environmental data into manageable small units. Each grid cell independently tracks object presence, allowing selective processing and transmission of only relevant grid data. This segmentation maintains prediction accuracy by preserving spatial relationships while dramatically reducing the total data volume that requires processing and transmission.
3Reliability
If decentralized learning is implemented without central server, then system autonomy and robustness improve, but each vehicle must identify and communicate with multiple other vehicles and edge servers
Solution Approach 1:
The patent merges the prediction functionality into a standardized module that can be deployed across multiple vehicles and edge servers with identical grid-based interfaces. This merging approach allows any vehicle or edge server to function as a prediction node without requiring complex peer identification logic, as they all speak the same grid data language and can process predictions independently based on received grid inputs.
Data Source
AI summary
A method for predicting presence of objects in an area is provided. The method includes obtaining a 2-dimensional matrix representing presence of objects in an area, each of values of the 2-dimensional matrix representing presence of objects in corresponding sub-region of the area, filtering the 2-dimensional matrix based on map information, converting the filtered 2-dimensional matrix to 1-dimensional data, inputting a series of the 1-dimensional data to a trained prediction machine learning model to obtain 1-dimensional data for future presence of objects, and converting the 1-dimensional data for future presence of objects to a 2-dimensional matrix representing the future presence of objects.