Social robots to assist humans in their daily lives

They can detect when you're feeling unwell, remind you to take your medicine or help you deal with extremely stressful situations. Although “social robots”, powered by artificial intelligence (AI), are not yet a major part of our daily lives, they have long been a full-time occupation for Adriana Tapus. The lecturer and researcher at the Computer Science and Systems Engineering Unit (U2IS) at ENSTA designs robotic systems with cognitive capabilities that can interact with humans. Her aim is to improve people's daily lives by establishing a symbiotic relationship with robots capable of adapting their behaviour according to users’ profiles and changing needs.

Yet, even before thinking of helping people, one has to understand them. Adriana Tapus is therefore trying to “model” humans – in other words, to put into equations their gestures, postures and facial expressions, as well as certain indicators such as facial temperature or heart rate, which can provide information about a person's emotions. The applications are numerous, ranging from the medical field and the surveillance of sensitive infrastructures like nuclear power stations, to education and the automotive industry. 

“Humans are complex”

The main difficulty is that “humans are complex,” says the director of the doctoral school at the Institut Polytechnique de Paris. “No two people are identical and the same person can have different moods throughout the day, which have an effect on their social interactions and their performance in everyday tasks”. Moreover, these robots share space with people, so they need to be able to adapt to their environment.

“At the heart of our approach is the robot's ability to understand its human partner and adjust its behaviour according to the user's characteristics, including personality, preferences, emotional state, physical or cognitive abilities, and even fatigue, while taking into account the task at hand. This adaptation using learning methods and AI aims to improve the user's acceptance, confidence and performance”, says the researcher.

Adriana Tapus and her team are contributing to the development of robots designed to assist vulnerable people (the elderly, children, people with physical or cognitive disabilities). “The aim is by no means to replace healthcare professionals, but rather to assist them and complement their support,” says the researcher. The robots will be able to provide regular monitoring, offer physical or cognitive stimulation and accompany people over long periods, helping to lighten the load on professionals while guaranteeing ongoing support.

As part of the European ENRICHME project which was rolled out from 2015 to 2018, a new approach to care for the elderly was proposed, offering personalised services, improved human-robot interactions and a professional care infrastructure. Through multidisciplinary research, this project in which Adriana Tapus participated, aimed to prolong the independence of the elderly at home by improving their quality of life. The system was tested and validated in several facilities in Europe (Greece, Poland, England), with a significant impact on the well-being of elderly people requiring constant supervision. The robot helped them to remember to take their medication, monitored their temperature, and engaged them with memory games and physical exercise.

Another example is decision-making support in situations of extreme stress. As part of the European RAICAM (Robotics and Artificial Intelligence for Critical Asset Monitoring) project, the researcher is working on improving interactions between nuclear power plant personnel and robots designed to help them monitor the facility. The aim is to detect and interpret signals of cognitive load (stress), such as an increase in facial temperature, and to determine the type of assistance to provide depending on the situation. In the summer of 2025, Adriana Tapus' team will be taking part in a full-scale simulation exercise at the Zwentendorf power plant in Austria, which was built but never commissioned, as part of a robotics hackathon funded by the European Union (Enrich 2025).

Tailor-made humour

The sophistication of human modelling goes as far as teaching robots to develop... a sense of humour, which has proved to facilitate social interaction. The difficulty lies as much in the ability to detect the right moment to express humour as in the synchronisation between the verbal semantic content and the non-verbal elements (gestures, posture, facial expressions, etc.) of the robot. It is also a matter of adapting the type of humour to the interlocutor, which presupposes that the artificial intelligence recognises clues such as the person's gender, age or cultural references. This work will continue in the new European doctoral training network project EU SWEET (Social aWareness for sErvicE roboTs), which will begin in February 2025.

*U2IS : a research lab ENSTA, Institut Polytechnique de Paris, 91120 Palaiseau, France