UC San Diego

This thesis is about the investigation of brain and body coupling with agents and objects at multiple scales in different contexts. We seek to characterize the reaction of behavioral and multi-region brain dynamics during interaction between rodents and other conspecifics, robots, and objects. Then we examine how these coupled agents and systems learn in the form of habituation during exploration of other agents. We highlight the importance of regulatory behaviors such as grooming, which may serve an important functional role in stabilizing the nervous systems using a phase alignment. The lessons learned from this empirical research are used to inform design principles for an autonomous interactive robot. These regulatory observations will act as a foundation for the proposal of a new learning framework which emphasizes the functional role of regulatory behaviors for maximizing safety. Recent studies introduce interactive robots as a dynamic comparison case or control condition for object and social interactions for the purposes of neuroscience. This dissertation will examine how interactive robots, as dynamic objects or potentially social others, can act as tools for probing questions related to agency, animacy and autonomy in social cognition, self-regulation, and perceptual exploration. Descriptions of the current state of interactive neurorobotics as a field are set forth, while also establishing design principles based on empirically-grounded interaction design studies. Chapter 1 is an introductory chapter introducing brain and body coupling as the basis of social cognition, animat and systems research and neural coupling during agent assessment. Chapter 2 examines how agent-based interactions perturb behavioral and brain states by characterizing the dynamics of behavioral and brain states that emerge during interaction with rats, robots, and stationary objects. To quantify dynamic interactions, we demonstrate the use of convolutional neural networks for offline animal and robot tracking. Chapter 3 examines olfactory habituation with rats and robots. Chapter 4 compares phase-amplitude coupling across brain regions during exploratory sniffing and regulatory self-grooming behavior. Chapter 5 provides a dynamical systems interpretation of the behavioral and neural data using a novel method for measuring dynamic coupling between neural systems, known as Convergent Cross Sorting. This lexicon of neurobehavioral dynamics will be used to inform design principles for interactive neurorobotics platforms. The results suggest that allostasis and autonomic regulation are crucial for designing interactive robots. In Chapter 6, exploration and regulation-based principles learned from the empirical portion of the dissertation are applied to developing autonomous algorithms for interactive robots, suggesting an ethologically grounded approach to learning. The Conclusion examines lessons learned overall and raises some ethical issues in the field of human-robot interaction.