Emotion-Aware Robot Planning With Multi-Agent Behavior Simulation

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current robots are largely reactive and lack the ability to proactively understand human intentions and preferences, especially in multi-agent scenarios, leading to suboptimal behavior in interactions and limited adaptability beyond training data.

Innovation Solution

A robot system utilizing a multi-agent simulation and large language models to generate and evaluate behavior options, considering human profiles and preferences, and iteratively updating its knowledge base based on feedback to enhance proactive interaction.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If robots use frequentist approach for behavior planning based on statistical analysis of past data, then they can generate behavior predictions, but they require large amounts of training data and lack generalization ability beyond training data

Engineering Contradiction:
Improvebehavior prediction accuracyVSAvoidgeneralization ability
Core Design Contradiction:
ReliabilityVSAdaptability or versatility

Solution Approach 1:

The system performs preliminary simulations of multiple potential behaviors before actual execution. By pre-evaluating various behavior options through simulated environments and assessing their potential outcomes, the robot can make more informed decisions without relying solely on statistical patterns from training data, thus improving generalization to novel situations.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The behavior planning system dynamically adapts by continuously updating its understanding of human intentions and preferences during interactions. The simulation-based evaluation allows the robot to flexibly adjust its behavior choices based on current context and learned patterns, rather than relying on static statistical models, thereby improving adaptability to new situations.

Inventive Principle:
Principle #15Dynamics

2Ease of operation

If robots rely on explicit commands from humans for behavior planning, then they can execute instructed tasks, but they lack proactive ability to anticipate and meet human intentions

Engineering Contradiction:
Improvecommand execution accuracyVSAvoidproactive behavior capability
Core Design Contradiction:
Ease of operationVSProductivity

Solution Approach 1:

The system generates multiple potential behavior options in advance and evaluates them through simulations before execution. This preliminary planning allows the robot to proactively anticipate human needs and intentions rather than merely reacting to explicit commands, improving productivity while maintaining ease of operation through structured decision-making.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The system incorporates feedback mechanisms where simulation outcomes and actual execution results are used to refine future behavior planning. By continuously learning from interaction outcomes and updating its understanding of human preferences, the robot becomes increasingly proactive in anticipating intentions while maintaining accurate command execution.

Inventive Principle:
Principle #23Feedback

3Ease of operation

If LLM-based approaches are used for behavior planning, then robots can generate next best actions, but they struggle with exhaustive situation analysis and decision-making when multiple human agents have competing interests

Engineering Contradiction:
Improveaction generation capabilityVSAvoidsituation analysis complexity
Core Design Contradiction:
Ease of operationVSDevice complexity

Solution Approach 1:

The complex decision-making process is segmented into distinct components: generating multiple behavior options, simulating each option's outcomes, evaluating results against human preferences, and selecting the best action. This segmentation allows the system to handle complex multi-agent situations systematically by breaking down the analysis into manageable simulation and evaluation steps.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The simulation environment acts as an intermediary between the LLM-generated behavior options and the final decision-making process. The simulation evaluates potential outcomes of each behavior option in a controlled virtual environment, providing structured feedback that helps resolve complex situations involving multiple human agents with competing interests without overwhelming the LLM directly.

Inventive Principle:
Principle #24Intermediary (Mediator)

Data Source

PatentEP4578604A1Proactive robot using theory of mind, multi-agent simulation, and emotion-aware planning
Publication Date: 2025.07.02 HONDA MOTOR CO LTD
  • EP4578604A1 patent drawingFigure 1
  • EP4578604A1 patent drawingFigure 2
  • EP4578604A1 patent drawingFigure 3

AI summary

A robot system for executing at least one behavior in an interaction with at least one human agent in an environment comprising a robot configured to execute the at least one behavior, at least one sensor configured to obtain information on the environment, and a processing means configured to detect the at least one agent based on the obtained information. A memory of the robot system serves as a knowledge base including profiles of the at least one agent and information on capabilities of the robot. The processing means is configured to generate a plurality of options of behaviors for interaction with the at least one agent, to initialize models for the at least one robot and the at least one agent with the information on the capabilities of the robot and the profiles from the knowledge base, and to perform a simulation for each of the plural options for behaviors based on the initialized models. In the simulation, each of the models generates a rating for a satisfaction or a cost value for each of the plural options of behaviors. The processing means is configured to select one behavior from the plural options of behaviors based on the generated ratings and cost values, and to control the robot to execute the selected behavior.