Scenario Parameter Optimization for Multi-Agent Responsibility Testing

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing techniques struggle to optimize scenario parameters in scenarios involving multiple agents, making it difficult to effectively inspect control logic for agents such as autonomous vehicles.

Innovation Solution

A scenario parameter optimization device that processes simulation results to calculate object function values and determine agent responsibility for events, optimizing parameters to minimize event occurrence and agent interaction, using methods like Hill Climbing, Simulated Annealing, or Genetic Algorithms.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If search-based testing is used to optimize scenario parameters, then the inspection efficiency of control logic is improved, but the technique cannot handle scenarios with multiple agents

Engineering Contradiction:

Improveinspection efficiencyVSAvoidmulti-agent scenario handling capability

Core Design Contradiction:

ProductivityVSAdaptability or versatility

Solution Approach 1:

The patent segments the complex multi-agent scenario optimization problem into distinct components: event detection module that identifies specific events, responsibility determination module that attributes events to specific agents, and optimization module that adjusts parameters. This segmentation allows the system to handle multiple agents by processing their interactions through discrete, manageable stages rather than treating the entire multi-agent system as a single complex optimization problem.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent introduces intermediary components between the optimization process and the multi-agent simulation: an event detection intermediary that translates complex agent interactions into detectable events, and a responsibility determination intermediary that attributes these events to specific agents. These intermediaries enable the optimization module to work with simplified representations of multi-agent scenarios while maintaining the ability to handle complex interactions.

Inventive Principle:

Principle #24Intermediary (Mediator)

2Device complexity

If scenario parameters are optimized without responsibility determination, then the optimization process is simpler, but the accuracy of identifying causative agents is reduced

Engineering Contradiction:

Improveoptimization process complexityVSAvoidagent responsibility identification accuracy

Core Design Contradiction:

Device complexityVSMeasurement precision

Solution Approach 1:

The patent applies preliminary action by determining agent responsibility before the optimization process begins. The event detection and responsibility determination modules operate first to identify which agent caused which event, creating a labeled dataset of agent-event relationships. This preliminary classification allows the subsequent optimization to focus on parameter adjustment rather than also having to determine causality, thereby maintaining optimization simplicity while ensuring accurate agent attribution.

Inventive Principle:

Principle #10Preliminary action

3Speed

If the optimization focuses only on object function value reduction, then the optimization is faster, but the ability to ensure specific agent responsibility is lost

Engineering Contradiction:

Improveoptimization speedVSAvoidagent responsibility assurance

Core Design Contradiction:

SpeedVSReliability

Solution Approach 1:

The patent implements feedback by using responsibility determination results to guide the optimization process. The system continuously monitors which agent is responsible for events and uses this information to provide feedback to the parameter optimization module. This feedback loop ensures that optimization not only reduces object function values quickly but also maintains or improves the accuracy of agent responsibility identification, as the optimization can adjust parameters based on observed agent behaviors and their causal relationships.

Inventive Principle:

Principle #23Feedback

Data Source

PatentUS20250355413A1Scenario parameter optimization device, scenario parameter optimization method, storage medium having stored therein a scenario parameter optimization program, and control logic inspection system

Publication Date: 2025.11.20 MITSUBISHI ELECTRIC CORP

AI summary

A scenario parameter optimization device includes: an object function value calculation unit that calculates an object function value of a predetermined event, and an event occurrence time at which the predetermined event occurs on the basis of an execution result of a simulation test of a scenario including a first agent and a second agent, executed on the basis of a scenario parameter; a responsibility determination unit that determines whether or not the first agent is responsible for the occurrence of the predetermined event from a situation at a responsibility determination time before the event occurrence time, and outputs a result of the determination as a responsibility determination result; and a scenario parameter optimization unit that optimizes the scenario parameter in such a manner that the object function value decreases and the first agent is determined to be responsible for the occurrence of the predetermined event.