On-Vehicle Behavior Modeling for Bandwidth-Efficient Autonomous Driving
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous vehicles face challenges in reacting to rare or unpredictable road and traffic conditions due to the complexity and variability of these environments, which makes it difficult to train them for every possible scenario, and current data collection methods are inefficient and costly in terms of communication bandwidth.
Innovation Solution
Implementing local vehicle behavior modeling where vehicles use onboard sensors to collect and process data, train a driving behavior model, and apply it for autonomous driving, while also uploading a compact model to a server for aggregation and distribution, thereby conserving bandwidth and enhancing robustness.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If large amounts of road condition data are transmitted from vehicles to remote servers for analysis, then the system can learn from existing conditions, but communication bandwidth is consumed and costs increase prohibitively
Solution Approach 1:
The patent extracts only the essential behavioral patterns and parameters from raw sensor data at the vehicle edge, rather than transmitting complete datasets. The edge computing device processes local sensor data to derive condensed behavioral representations that capture essential driving patterns without requiring transmission of voluminous raw data, thus resolving the bandwidth consumption problem while preserving learning capability.
Solution Approach 2:
The system creates simplified copies or representations of driving behavior data through edge processing. Instead of transmitting original high-fidelity sensor data, the edge computing device generates condensed behavioral models or feature extracts that replicate the essential information needed for fleet-wide learning, enabling efficient data sharing without exhaustive bandwidth usage.
2Reliability
If autonomous vehicles are trained for every possible rare condition or edge case, then they can react instinctively to extraordinary events, but the complexity and unpredictability of road conditions make this difficult
Solution Approach 1:
The system performs preliminary processing and pattern extraction at the edge device during normal operation, preparing condensed behavioral representations in advance. This allows the fleet system to accumulate diverse behavioral patterns from multiple vehicles over time, including rare edge cases, without requiring complex real-time processing when such events occur. The preliminary edge processing creates a repository of learned patterns that can handle rare events efficiently.
Solution Approach 2:
The edge computing platform provides universal processing capabilities across the fleet, enabling a single system to handle multiple vehicle types, diverse driving conditions, and various rare event scenarios. By creating a centralized learning system that aggregates behavioral patterns from across the fleet, the system achieves multi-functionality in handling different edge cases without requiring separate specialized training for each rare condition.
3Loss of information
If complete training data is uploaded to servers for fleet learning, then comprehensive analysis is possible, but communication bandwidth consumption becomes prohibitively expensive
Solution Approach 1:
The edge computing device extracts essential behavioral features and parameters from complete training data locally at the vehicle. This extraction process identifies and isolates the critical information elements needed for fleet learning while discarding redundant data. The result is a condensed dataset that preserves the essential learning value while occupying minimal transmission bandwidth.
Solution Approach 2:
The system transforms raw training data into different parameter representations through edge processing. By changing the data format from voluminous raw sensor streams to condensed behavioral parameters or feature vectors, the system maintains the informational content necessary for learning while dramatically reducing the data size suitable for transmission over communication networks.
Data Source
AI summary
This application is directed to on-vehicle behavior modeling of vehicles. A vehicle has one or more processors, memory, a plurality of sensors, and a vehicle control system. The vehicle collects training data via the plurality of sensors, and the training data include data for one or more vehicles during a collection period. The vehicle locally applies machine learning to train a vehicle driving behavior model using the collected training data. The vehicle driving behavior model is configured to predict a behavior of one or more vehicles. The vehicle subsequently collecting sensor data from the plurality of sensors and drives the vehicle by applying the vehicle driving behavior model to predict vehicle behavior based on the collected sensor data. The vehicle driving behavior model is configured to predict behavior of an ego vehicle and/or a distinct vehicle that appears near the ego vehicle.