Autonomous Vehicle Destination Selection for Targeted Data Collection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous vehicles are inefficient in data collection as they often drive randomly or in specific patterns when not in use for transportation services, resulting in uneventful miles that do not contribute significantly to map accuracy or training model data.
Innovation Solution
A targeted driving system that uses processors to select destinations from a set of predetermined locations based on relevance scores calculated from routes and specific driving goals, such as increasing miles with higher disengage rates or lane changes, and assigns these destinations to autonomous vehicles to travel to in autonomous mode.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If autonomous vehicles drive randomly or in specific patterns when not in use for transportation services, then they can cover various areas, but the data collection efficiency is low and the miles driven are uneventful and do not contribute significantly to map accuracy or training model data
Solution Approach 1:
The system pre-identifies and stores multiple potential destinations with associated route information before the autonomous vehicle needs to collect data. When data collection is needed, the system quickly selects from these pre-prepared options based on current conditions, avoiding the need for real-time complex decision-making and ensuring the vehicle always has relevant destinations to visit.
Solution Approach 2:
The system dynamically adjusts destination selection based on changing conditions such as current location, traffic conditions, and data collection priorities. The relevance scoring mechanism allows the system to adaptively choose destinations that maximize data collection efficiency while minimizing uneventful miles, making the routing flexible and responsive to real-time requirements.
2Loss of information
If autonomous vehicles are dispatched to multiple predetermined destinations, then targeted data can be collected, but the system complexity increases due to route determination and relevance scoring
Solution Approach 1:
The system segments the destination selection process into distinct components: route determination for each destination, relevance scoring based on multiple criteria, and bucket assignment for organized selection. This modular approach allows each component to be optimized independently while working together to achieve accurate map data collection.
Solution Approach 2:
The system introduces an intermediary processing layer between the vehicle and destinations that handles the complex calculations of route determination and relevance scoring. This intermediary system processes multiple factors (distance, traffic, data collection priority) to generate optimized destination selections, shielding the vehicle control system from complexity while ensuring high-quality data collection.
3Reliability
If autonomous vehicles focus on routes with higher expected disengage rates or lane changes, then training data quality improves, but the number of suitable destinations decreases
Solution Approach 1:
The system changes the parameters used to evaluate destinations by introducing relevance scores that incorporate multiple factors including expected disengage rates, lane change frequency, and distance. By adjusting these parameters and their weights, the system can flexibly prioritize training data quality while still maintaining a sufficient pool of suitable destinations through the bucket assignment mechanism.
Data Source
AI summary
Aspects of the disclosure provide a method of providing a destination to an autonomous vehicle in order to enable the autonomous vehicle to collect data according to a targeted driving goal. For instance, a current location of an autonomous vehicle may be received. A set of destinations may be selected from a plurality of predetermined destinations. A route may be determined for each destination. A relevance score may be determined for each destination based on the determined routes and the targeted driving goal. Each destination may be assigned to one of a set of two or more buckets based on the relevance scores. A destination of the set may be selected based on a predetermined sampling probability. The selected destination is sent to the autonomous vehicle in order to cause the autonomous vehicle to travel to the selected destination in an autonomous driving mode.