Geo-Fenced Driving Policy Simulation for Realistic Traffic Interaction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for improving autonomous vehicle driving policies, such as online and offline simulations, are limited in providing realistic interactions and scenario diversity, especially for location-specific situations, leading to suboptimal safety and confidence in real-world driving.
Innovation Solution
A method involving an autonomous vehicle collecting data at specific locations, transmitting it to a data center for offline traffic simulations using realistic and interactive traffic generators, which updates the driving policy by simulating various scenarios to maximize failure rates and optimize safety and confidence measures.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If online simulations with virtual vehicles are used to improve safety, then safety and confidence of driving policy are improved, but the interactions with virtual vehicles are limited because virtual vehicles take decisions based on hard coded rules and cannot handle multiple real drivers
Solution Approach 1:
The patent creates copies of real vehicles and drivers within the simulation environment. Virtual vehicles are equipped with learned driving policies that replicate real driver behavior patterns, allowing multiple realistic driver personalities to interact within the simulation. This enables the system to test driving policies against diverse, realistic traffic participants rather than simple hard-coded virtual agents.
Solution Approach 2:
The patent introduces a learned policy model as an intermediary between real driver data and the simulation environment. This model translates real-world driving behavior into virtual agent behavior, enabling realistic interactions without requiring direct connection to real vehicles during simulation. The intermediary allows the system to leverage real driver patterns while maintaining simulation safety and scalability.
2Adaptability or versatility
If offline traffic simulation with simulated agent substitution is used, then scenario diversity is improved, but the original agent from the traffic cannot interact realistically with the simulated one because they just replay logs with simple safety rules
Solution Approach 1:
The patent transforms static logged trajectories into dynamic interactive agents by changing key parameters. Instead of simply replaying fixed log data, the system learns continuous driving policies from logs that can adapt to new situations. This allows simulated agents to respond dynamically to perturbations and interactions while maintaining realistic behavior patterns observed in real traffic.
Solution Approach 2:
The patent converts static log replay into dynamic policy-based agents. The learned policies enable simulated agents to adapt their behavior in real-time during simulations, responding to new situations and interactions rather than following predetermined paths. This dynamic approach maintains interaction realism while enabling diverse scenario generation through policy perturbations.
3Reliability
If extensive real-world data collection is performed to improve driving policy, then safety and confidence are improved, but time consumption and human intervention requirements increase
Solution Approach 1:
The patent performs preliminary learning of driving policies from real-world data in advance, creating reusable policy models. Once learned, these policies can be repeatedly used in simulations without requiring additional real-world data collection. This preliminary action separates the expensive data collection phase from the iterative simulation phase, enabling rapid policy improvement through simulation alone.
Solution Approach 2:
The patent creates virtual copies of real-world driving scenarios and agents within the simulation environment. By replicating real traffic patterns, road geometries, and driver behaviors in simulation, the system can conduct extensive testing without returning to the real world. This copying approach maintains safety benefits while eliminating the time costs of repeated real-world data collection.
Data Source
AI summary
A method updates a target driving policy for an autonomous vehicle at a target location. The method includes the steps of obtaining, by the vehicle, vehicle driving data at the target location; transmitting, by the vehicle, the obtained vehicle driving data and a current target driving policy for the target location to a data center; performing, by the data center, traffic simulations for the target location using the vehicle driving data to obtain an updated target driving policy; and transmitting, by the data center, the updated target driving policy to the vehicle.