机器人-动物交互研究 z-bio 2025-01-14 843

　　Example of three interaction modes for robot rat: pinching, pouncing, and social nose contact

　　Under the support of National Natural Science Foundation projects (approval numbers: 62022014, 62088101), Professor Shi Qing's team from Beijing Institute of Technology has made progress in the research of robot animal interaction through cross collaboration with overseas collaborators. The research findings, titled "Modulating emotional states of rats through a rat like robot with learned interaction patterns," were published online on December 5, 2024 in the journal Nature Machine Intelligence. Paper link: https://www.nature.com/articles/s42256-024-00939-y .

　　Robots integrate into biological systems in the form of social agents, serving as observers and stimuli to interact with animals, helping humans explore how individuals or groups of animals respond to controlled environmental changes. This vitality fusion system has controllability and helps to elucidate potential biological intelligence that cannot be revealed through traditional methods. However, existing interactive robots still find it difficult to transmit multiple sensory information in vital fusion systems, and therefore cannot effectively regulate complex interaction processes. They mainly face three challenges: 1) integrating multiple sensory perceptions for communication with animals; 2) To integrate into animal communities, it is necessary to realistically imitate diverse and complex animal behaviors; 3) How to directly change the internal state of animals (such as emotions) has not been deeply explored.

　　The research team drew inspiration from the anatomical structure, dynamic movement, and social interaction of rats to develop a miniature mouse like interactive robot that can mimic various animal social behaviors and autonomously interact with rats. By establishing a main motion joint mapping model, a highly flexible multi joint spine structure was designed. Four social behavior patterns were generated using imitation learning, and an autonomous control architecture with custom interaction rules was constructed. The proposed mouse like robot achieved an average of 30 continuous interactions within half an hour, and was able to attract the attention of rats and significantly stimulate their interest.

　　The research results demonstrate that this type of mouse robot has achieved for the first time the regulation of emotional states in rats (Figure), overcoming the limitations of natural social interactions within biological systems. It is expected to help reveal the brain functions and neural circuit mechanisms behind these decisions or predictions, providing new ideas for understanding the "social" interactions between humans and artificial intelligence.