Robot Behavior Control Through Multimodal Emotional Interaction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing technologies face challenges in forming emotional connections between robots and humans, as they rely on limited input methods like touch panels or voice commands, and struggle with learning through trial and error, which can be dangerous and costly, while limiting performance and social interaction.
Innovation Solution
A behavior control device and method that includes a perception unit to acquire person information, a learning unit for multimodal emotional interaction, and an operation generation unit to generate behaviors based on learned interactions, using implicit and explicit rewards, social norms, and psychological knowledge to facilitate autonomous learning.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If reinforcement learning from evaluation feedback is used, then the robot can learn through trial and error, but the learning process becomes dangerous and costly
Solution Approach 1:
The patent applies preliminary action by using inverse reinforcement learning to extract reward functions from human demonstrations before the robot executes actions. This allows the robot to learn desired behaviors by observing human actions and inferring the underlying reward structure, rather than learning through dangerous trial-and-error in the real environment. The reward function is established in advance through analysis of human behavior patterns.
2Productivity
If learning from demonstration is used, then learning speed increases, but the trainee's performance is limited
Solution Approach 1:
The patent implements feedback by combining inverse reinforcement learning with explicit reward evaluation. The system not only learns from human demonstrations but also receives explicit feedback when the robot's actions deviate from desired behavior. This dual feedback mechanism allows the robot to correct performance limitations and achieve higher quality outcomes while maintaining fast learning speed through the pre-established reward function.
Data Source
AI summary
A social ability generation device includes a perception unit that acquires person information on a person, extracts feature information on the person from the acquired person information, perceives an action that occurs between a communication device performing communication and the person, and perceives an action that occurs between people, a learning unit that multimodally learns an emotional interaction of the person using the extracted feature information on the person, and an operation generation unit that generates a behavior on the basis of the learned emotional interaction information of the person.