Doctoral Project: Anticipation and prediction in robot interaction
Part of research project: Anticipation, Prediction and Behavioral Reliability in Social Interactions
Description of the research project
Agents with predictive skills can anticipate future events and thus have a major advantage over agents that cannot (Pezzulo et al., 2008; Winfield and Hafner, 2018). In a social setting, the ability to anticipate an interaction partner’s future actions will allow to adaptively modulate own behavioral strategies (Moussaid et al., 2011). In a cooperative context, anticipation may help to enhance communication among agents while in a competitive context it gives room for manipulative strategies. The proposed project will focus on the following four main objectives: In a first step, we will focus on developing analytical methods to find heuristics of anticipation in live fish that vary also in social responsiveness (objective a). In a second step, we will create situations in which live fish
have to cooperate or compete with Robofish in order to achieve a goal effectively. This will allow us to estimate costs and benefits associated with anticipatory strategies (objective b). The aim of our experimental data is to identify cognitive heuristics that play a role in anticipation in moving pairs that will then be used to develop a ubiquitous synthetic behavior of anticipation when social responsiveness in interaction partners varies (objective c). In order to evaluate this synthetic behavior, it will be implemented into Robofish as well as in humanoid robots and then tested in situations involving either robot-only pairs (robot-robot) or, in case of Robofish, also pairs with one live agent (fish-robot) (objective d). Our project depends on SCIoI’s approach of bringing together analytic and synthetic disciplines as it involves back and forth switching between analytical phases (experiments and complex data analysis) and synthetic phases (building robots, developing and implementing algorithms) that ultimately lead to the creation of synthetic artefacts that use principles of social intelligence.
Description of the doctoral project
This project focuses on anticipation and prediction for social interaction in both animals and robots. The interactions can be between a pair of agents or a small group. We want to explore how prediction is used to anticipate future actions of a social partner by designing computational models of internal simulations of the agent, other agents and the environment. These models will first be explored using a small number of humanoid robots in a situation of cooperation and competition. Successful models will then be transferred to the RoboFish, adapting the models to empirical data from the fish experiments.
Project start date: October 1, 2019
Applicants must hold a Diploma/Master’s degree in Computer Science or related sciences and should have proven skills/background in following topics:
- Computational modelling / Machine learning / Developmental and Bio-Robotics
- Programming skills
- Interest in interdisciplinary collaboration
- Interest in performing behavioural experiments