Closed-Loop Driving Scenario Simulation From Real-World Traces
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current simulation methods for autonomous driving systems struggle to recreate real-world driving scenarios accurately, especially in closed-loop simulations where both the ego and non-ego agents exhibit autonomous, reactive behavior, making it challenging to test the safety and performance of autonomous vehicles in all possible driving conditions.
Innovation Solution
A computer system is developed to automatically derive driving scenarios from real-world data, incorporating a trace extraction component, driving scenario extraction component, simulator, and agent decision logic to create closed-loop simulations where both the ego and non-ego agents exhibit autonomous behavior, allowing for the evaluation of autonomous vehicle stacks in a simulated environment.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If physical world testing is used to ensure safety of autonomous vehicles, then safety validation can be performed, but it is expensive and time-consuming
Solution Approach 1:
The patent creates virtual copies of real-world driving scenarios by extracting traces from actual driving data and reconstructing them in a simulation environment. This allows safety testing to be performed on these virtual copies rather than requiring extensive physical testing, thereby reducing time and cost while maintaining validation effectiveness
Solution Approach 2:
The system performs preliminary extraction and structuring of driving scenario data from real-world traces before simulation testing. By preparing the test scenarios in advance through automated extraction and organization of driving data, the system enables efficient simulation-based validation without requiring time-consuming physical test setup
2Ease of operation
If open-loop simulation is used where non-ego agents follow predefined traces, then simulation simplicity is maintained, but realistic autonomous behavior cannot be evaluated
Solution Approach 1:
The patent transitions from static, predefined agent trajectories to dynamic, reactive agent behavior. Non-ego agents are equipped with decision logic that allows them to adapt their behavior in real-time based on interactions with the ego agent and changing environmental conditions, creating realistic autonomous behavior while maintaining simulation manageability
Solution Approach 2:
The system implements feedback mechanisms where non-ego agents continuously monitor the state of the ego agent and adjust their behavior accordingly. This feedback loop enables realistic autonomous reactions to ego agent actions, transforming simple trace following into sophisticated autonomous behavior that maintains simulation simplicity through structured interaction protocols
3Reliability
If closed-loop simulation with autonomous non-ego agents is implemented, then realistic behavior evaluation is achieved, but simulation complexity increases
Solution Approach 1:
The patent divides the complex simulation system into distinct modular components: trace extraction modules, scenario reconstruction modules, agent decision logic modules, and evaluation modules. Each non-ego agent is implemented as an independent module with specific behavioral rules, allowing the complex closed-loop simulation to be managed through systematic segmentation of functionality
Solution Approach 2:
The system creates a universal simulation framework that can handle multiple types of agents (pedestrians, vehicles, cyclists) and multiple driving scenarios using a common architecture. The agent decision logic and interaction protocols are designed to be multi-functional, accommodating different agent types and behavioral patterns without requiring separate simulation systems for each case
Data Source
AI summary
Abstract: A driving scenario is extracted from real-world driving data captured within a road layout. A simulation is run based on the extracted driving scenario, in which an ego agent and a simulated non-ego agent each exhibit closed-loop behaviour. The closed-loop behaviour of the ego agent is determined by autonomous decisions taken in an AV stack under testing in response to simulated inputs, reactive to the simulated agent. The closed-loop behaviour of the non-ego agent is determined by implementing an inferred goal or behaviour, reactive to the ego agent. The goal or behaviour is inferred from an observed trace of a real-world agent extracted from the real-world driving data.