Miss Caroline Raymundo
PublicationsGoogle Scholar and Research Gate.
PhD Project Summary
Fear learning is an essential skill that is ubiquitous in nature and plays an important role in adaptation and cognitive tasks for a large range of animal species. With the accelerated advances in robotics, robots have become increasingly present in our everyday life. Consequently, robots now have to deal with and adapt to the threats of our physical world and to the rules of our society. For this reason we believe that the ability to fear became a necessary skill for modern robots and essential for the future of robotics.
In addition, the ability to fear may also contribute to increasing the believability of robots. As humans, we expect others to be able to identify environmental factors that can represent a threat to themselves and act accordingly. Therefore, being able to properly express fear responses could highly increase the believability of a robot, leading to a better human-robot interaction experience.
This PhD project is dedicated to investigate and combine neuroscience findings behind the fear mechanism of the brain with computational algorithms in order to create a computational model of fear learning with main application to robotics. The result is a novel hybrid computational model, named SAFEL (Situation-Aware FEar Learning), which combines concepts of situation-aware expert systems with well-known neuroscientific findings on the brain fear-learning mechanism to allow robots to predict undesirable or threatening situations based on past experiences.
SAFEL allow robots to learn complex temporal patterns of sensed environmental stimuli and create a unified representation that can be later associated with a negative or positive "emotion", analogous to fear and confidence. This association is later used to predict the occurrence of aversive stimuli, thus allowing the robot the chance to react and avoid the threat before its occurrence.
The main contributions of SAFEL as compared to the state of the art are:
- Integration of a fear learning model with the concept of temporal context. SAFEL performs threat predictions based on complex temporal and contextual information. Existing fear memory models either focus in the contextual or the temporal aspect, overlooking the need of both skills for an artificial intelligent agent to properly react to real-world threatening situations.
- SAFEL is focused on real-world applications for artificial and autonomous intelligence in robotics. Many existing fear-learning models that are inspired by the real mechanisms of the brain focus on providing a close-to-real emulation of brain functions without addressing the practical usage of the model for artificial intelligence.
- The successful integration of a symbolic rule-based platform for situation management with classification algorithms (more specifically, artificial neural network and binary classification tree) for memorizing and predicting threats based on complex temporal context.
I am currently a member of the Computational Intelligence Research Group, at the School of Computing of the University of Kent. My research interests include computational neuroscience, machine learning, artificial intelligence, autonomous robotics, affective computing and situation-awareness.
I am also interested in context/situation-aware systems, rule-based systems and complex event processing, subjects with which I have worked during my Masters. For more information on my works in this area: