Back to search

IKTPLUSS-IKT og digital innovasjon

Predictive and Intuitive Robot Companion (PIRC)

Alternative title: Forutseende og intuitiv robotmedhjelper (PIRC)

Awarded: NOK 16.0 mill.

PIRC targets a psychology-inspired computing breakthrough through research combining insight from cognitive psychology with computational intelligence to build models that forecast future events and respond dynamically. The systems will be aware and alert for how to act best given their knowledge about themselves and perception of their environment. Humans anticipate many future events more effectively than computers. We combine sensing across multiple modalities with learned knowledge to predict outcomes and choose the best actions. Can we transfer these skills to intelligent systems in human-interactive scenarios? In PIRC, we will combine our machine learning and robotics expertise and collaborate closely with researchers in cognitive neuropsychology, to apply recent models of human prediction to perception-action loops of future intelligent robot companions. Our work will allow such robots to adapt and act more seamlessly with their environment than the current technology. We will equip the robots with these new skills and, in addition, provide them with the knowledge that users they are interacting with apply the same mechanisms. Studies of human perception and decision-making are of special relevance to modelling behaviour and forecasting future events and actions. This will include mechanisms for adaptive response time from quick and intuitive to slower and well-reasoned. The models will be applied in two robotics applications with potential for very wide societal impact: physical rehabilitation and home care robot support for older people. In the first phase of the project, research was done through a newly graduated master´s student and collaboration with Eindhoven University of Technology, respectively which has led to two peer-reviewed conference articles including Human-Robot Interaction (HRI) studies. The work has been on different prediction models and user studies with senior participants, respectively, to compare different forms of human-robot interaction. Another master´s student introduced a technique called “PINE: Planning and Identifying Neural Network" (2021). It can select the most relevant trained task for a robot and if unknown, it selects the most similar known task to build the new learning on. Further, a master´s student graduated in spring 2022 with a project implementing a rehabilitation task through game playing by a PAL Robotics TIAGo robot. Several master students are currently also working on PIRC-relevant master's thesis projects. We have also hired a PhD student (started August 2021) and a researcher (January 2022–July 2023). They have focused on understanding the role of human intuition in Human-Robot Interaction regarding non-verbal communication. That is, by combining methods such as gaze-tracking, pupil measurements, and surveys, we can better understand how humans understand and interpret different robot behaviours. Some of the findings were presented at a conference in 2022, while there are also a couple of more papers currently in review. Furthermore, we have put effort into how humans intuitively understand robot behaviour and studied how to design behaviour for robots to make them show different personalities. This study was done in collaboration with the University of Eindhoven. A study on robots as welfare technology to reduce falls among older adults has also been performed and published. The work with user studies will continue, and we have since 2022 collaborated with a Care+ type of housing facility for independent living for elderly people. There is also an activity centre for elderly people, and the staff and residents have expressed much interest in contributing to making progress in improving robot assistance technology. Several user studies with the TIAGo robot have been undertaken there as a complement to experiments in lab settings. Through an in-kind PhD student, we have also published two papers (2022 and 2023) on improved robot arm reaction time for collision avoidance by a novel way of performing user pose prediction (by combining traditional control and reinforcement learning). In addition, we have a number of other papers published by in-kind researchers. We have also focused our effort on creating a virtual environment where AI algorithms can be more easily developed and tested. Further, we have done a user study with senior participants on the difference in perception between a robot presented in augmented reality and a real-life robot assistant. This research will allow us to better adapt the algorithms developed in simulation so that they translate better when deployed into the real world. From a research dissemination perspective, we have also communicated the project´s goals and results through invited talks and tutorials at various international conferences and in media articles. See the most recent one here: https://elektronikk.prenly.com/p/elektronikk/2023-11/r/1/1/5683/1123083

PIRC targets a psychology-inspired computing breakthrough through research combining insight from cognitive psychology with computational intelligence to build models that forecast future events and respond dynamically. The systems will be aware and alert for how to best act given their knowledge about themselves and perception of their environment. Humans anticipate future events more effectively than computers. We combine sensing across multiple modalities with learned knowledge to predict outcomes and choose the best actions. Can we transfer these skills to intelligent systems in human-interactive scenarios? In PIRC, we will apply our machine learning and robotics expertise, and collaborate with researchers in cognitive psychology, to apply recent models of human prediction to perception-action loops of future intelligent robot companions. Our work will allow such robots to adapt and act more seamlessly with their environment than the current technology. We will equip the robots with these new skills and in addition, provide them with the knowledge that users they are interacting with, apply the same mechanisms. Studies of human perception and decision making are of special relevance to model behaviour and forecast future events and actions. This will include mechanisms for adaptive response time from quick and intuitive to slower and well-reasoned. The models will be applied in two robotics applications with potential for very wide impact: physical rehabilitation and home care robot support for older people. Psychology and biology have inspired several breakthroughs in artificial intelligence, such as the perceptron which underlies deep neural networks. We aim to develop similarly high-impact psychology-inspired predictive models. The unexplored nature of these models offers high potential gains, but significant risk. This is partly mitigated by the two different applications in PIRC, which help to uncover different benefits in the new methods.

Publications from Cristin

Funding scheme:

IKTPLUSS-IKT og digital innovasjon