INESC-ID has been awarded $300K by the USA Air Force Office of Scientific Research (AFOSR) to investigate human-robot interactions. The funded project — “Trustworthy Ad Hoc Teamwork” — officially starts today, 01 September 2022, and will be funded for two years.

This project is led at INESC-ID by Alberto Sardinha, researcher in the Artificial Intelligence for People and Society Research Area and Assistant Professor at the Department of Computer Science and Engineering of Instituto Superior Técnico. Alberto explained: “Trust is an essential element for effective cooperation between humans and robots. However, to the best of our knowledge, no research work has analyzed trust between robots and humans within ad hoc teamwork scenarios or created tailored algorithms for trustworthy ad hoc teamwork.”

“Trustworthy Ad Hoc Teamwork” investigates how to create an autonomous agent that can efficiently and robustly collaborate with previously unknown teammates. The project plans to achieve this through novel ad hoc teamwork algorithms to build trust within human-robot interactions and by addressing a vital question: How can a robot learn to cooperate with unknown human teammates in complex domains while fostering trust development, whereby the robot does not have any pre-coordination protocol and must learn to cooperate on the fly?

By implementing a combination of sensors (networked within the environment – e.g., microphones, wall-mounted cameras – and the robot’s on-board sensors – e.g., lasers, cameras, RFID readers), an ad hoc robot (containing a combination of perception, decision and execution modules) and output modalities (different components or hardware that support the robot’s behavior or actions), this project aims to answer a series of additional unfolding queries: How can a robot learn, in a trustworthy way, the cooperative task in the absence of a reward signal and without full observability of states and actions? How to tailor the ad hoc teamwork algorithms in order to increase a human’s trust in robots? How can the multimodal capabilities of a robotic teammate (e.g., speech recognition and synthesis, recognition of people and intentions, social robot expression) provide valuable information to facilitate the development of trust within ad hoc teamwork scenarios?

“Trustworthy Ad Hoc Teamwork” is Alberto Sardinha’s third research project on this topic. The first project (Ad Hoc Teams: Ad Hoc Teams With Humans And Robots) was also funded by the AFOSR and the second (HOTSPOT: Human-robOt TeamS without PrecoOrdinaTion) by FCT. Trust and Influence is one of AFOSR’s interdisciplinary basic research funding portfolios, fitting the scope of the current grant.

Led by INESC-ID, “Trustworthy Ad Hoc Teamwork” will count with the additional participation of PUC-Rio (Brazil).