Parking Lot Space Prediction Using Graph Attention and Recurrent Networks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for predicting free parking spaces in parking lots are inaccurate due to coarse user evaluations and misoperations, leading to unreliable prediction results.
Innovation Solution
A method involving building a parking lot association graph, determining local and global space correlation information using graph attention and hierarchical graph neural networks, and predicting future free parking spaces using gated recurrent neural networks, based on environment context features and historical data from real-time sensors.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If user feedback is used to generate annotation data for predicting parking difficulty, then the prediction system can be implemented, but the accuracy of prediction results deteriorates due to coarse evaluations and user misoperations
Solution Approach 1:
The patent introduces sensor data as an intermediary to bridge the gap between user feedback and accurate parking space prediction. Sensors objectively measure parking lot occupancy and vehicle flow, providing precise data that mediates between coarse user evaluations and the need for accurate predictions, thereby resolving the contradiction between implementation feasibility and prediction accuracy
Solution Approach 2:
The patent replaces the mechanical system of manual user feedback with an automated sensing system. Instead of relying on users to manually evaluate parking difficulty, the system uses sensors to automatically collect and process objective data about parking lot conditions, eliminating human error and coarse evaluations while maintaining system implementability
2Device complexity
If only local space correlation information is used for prediction, then the model complexity is reduced, but the prediction accuracy deteriorates due to lack of global context
Solution Approach 1:
The patent segments the prediction model into two distinct components: local space correlation information and global space correlation information. This segmentation allows the system to process information at different spatial scales independently, managing model complexity while capturing both local parking lot conditions and global regional patterns, thereby improving prediction accuracy without overwhelming computational burden
Solution Approach 2:
The patent adds a global dimension to the prediction model by incorporating global space correlation information. This dimensional expansion allows the system to consider not only local parking lot conditions but also broader regional parking patterns, providing a more comprehensive view that enhances prediction accuracy while maintaining manageable model complexity through hierarchical processing
3Measurement precision
If historical sensor data is collected and processed, then the prediction accuracy is improved, but the data processing time and computational resources increase
Solution Approach 1:
The patent applies preliminary action by pre-processing and storing sensor data in an organized manner as it is collected. Historical sensor data is prepared and structured in advance, so when prediction is needed, the system can quickly retrieve and process pre-organized information rather than raw data, reducing real-time processing time while maintaining high prediction accuracy
Solution Approach 2:
The patent applies local quality by focusing computational resources on processing the most relevant local space correlation information first. The system prioritizes processing data from nearby parking lots and recent time periods, applying higher processing quality to locally relevant data while using simpler processing for global context, thereby improving prediction accuracy without uniformly increasing processing time across all data
Data Source
AI summary
A parking lot free parking space predicting method, apparatus, electronic device and storage medium are provided. The method comprises: building a parking lot association graph for parking lots in a region to be processed; aggregating environment context features of neighboring parking lots according to weights of edges between the neighboring parking lots and a parking lot i to obtain a representation vector of the parking lot i at a current time; and pre-training a graph attention neural network model using the environment context features of the neighboring parking lots and free parking space information, and a gated recurrent neural model according to the representation vector of the parking lot it at the current time.