[Belle histoire] Technology to enhance the relationship between elderly people and their pets

France, like many other Western countries, is undergoing a demographic transition, leading to an aging population. Against this backdrop, the issue of ageing well and keeping elderly people at home is a major one. The daily presence of a pet can have many beneficial effects on an elderly person: emotional support, preservation of social ties, reduced feelings of loneliness, physical activity...

Unfortunately, from the age of 65 onwards, the rate of ownership of such a companion falls by 50% every decade*. This is due to financial constraints, deteriorating physical capabilities, and even health conditions. Could technology help to slow this trend, by helping to maintain and support the bond between an elderly person and a pet?

How can technology be used in the relationship with a pet?

Hazar Zilelioglu's CIFRE thesis, co-directed by Yacine Amirat, Abdelghani Chibani and Ghazaleh Khobadandelou - from LISSI (Laboratoire Images, Signaux et Systèmes Intelligents) - and Ioana Ocnarescu - from Strate École de Design - addressed these issues. The CIFRE thesis also brought together two industrial partners: frog part of Capgemini Invent, with the help of BNP Paribas Cardif, and it is part of Robotics by Design Lab - the joint multidisciplinary laboratory launched by Strate and its partners in December 2019. " We're very interested in the possible contributions of technology to better aging, given the demographic transition," says Solène Le Bars, scientific manager of the health program at Capgemini Engineering. " We're trying to identify the most relevant use cases, particularly in nursing homes and hospitals. Here, it was more a question of contributing to the autonomy of elderly people at home. "

" The particularity of this thesis lies in its initial openness: at the beginning, we didn't really know what direction to take," recalls Hazar Zilelioglu, now a doctor in artificial intelligence at Capgemini Engineering. Consequently, the first phase of our research was to gain a better understanding of the needs of elderly people with regard to their pets. Eleven individuals were interviewed to find out more about a typical day and their relationship with new technologies (computers, smartphones, robots, etc.).

" The first thing we learned was that all the people surveyed routinely used their computers and smartphones, sometimes in conjunction with their pets," notes Hazar Zilelioglu. " These included finding information on the Internet, ordering products via their smartphone, and using a GPS tracker. On the other hand, they were more skeptical about the contribution of robotics to their relationship with their pet. " These results were then enriched by a survey of articles on the subject, to complement the difficulties and issues already identified.

Sensors to monitor animal activity

The data gathered during this first stage served as the basis for an ideation workshop, held at the Cardif Lab'. Indeed, design was the unifying factor in this ideation phase. The aim of the workshop organized by frog and Strate with BNP Paribas Cardif (data analytics, marketing, customer experience, CSR, sales) was to imagine new services around aging well by exploring the benefits of AI and social robotics on the relationship between an elderly person and their pet. " For example, we imagined the case of an individual needing to carry out analyses in hospital for several days," cites Ioana Ocnarescu, Director of Research at Strate École de Design. " In this case, the technology could be used to feed the animal while its owner is away. These scenarios were then submitted to three experts - a geriatric doctor, a canine behavior expert and a designer - who then recommended that the research team focus on preventing difficulties. In other words, the technology was to be used to monitor the relationship between an elderly person and their companion, to alert at the first signs of a possible imbalance.

But how do you do this? " We decided to monitor the animal's activity, which could be indicative of its owner's, " explains Hazar Zilelioglu. " We therefore defined an experimental protocol and developed a tracking solution installed on a collar and based on inertial units. " These devices, comprising three accelerometers and three gyrometers, tracked the animal's movements, supplemented by data collected via a dedicated mobile application. Using this app, participants could describe their companion's activities and answer questions relating to their own well-being and that of the animal. " At this stage, the support of Carnot TSN was decisive for the development of the prototypes and the mobile application ", emphasizes Ioana Ocnarescu.

Adapting AI models to the specificities of animal activity

The experiment was conducted with ten participants over a five-day period. First of all, it showed that the animals' activity decreased with age - both their own and that of their owners. "What's more, the results seem to show a correlation between pet activity and individual well-being," notes Hazar Zilelioglu. " This would confirm the idea that it is possible to indirectly track human beings through data relating to their pets. " These conclusions remain to be verified, however, due to the small size of the study sample.

In addition, the aim of the experiment was to feed artificial intelligence models designed to predict what the animal is doing, based on data from inertial sensors. " We had already worked on algorithms for recognizing human activity, but there are very few similar studies on animals," notes Hazar Zilelioglu. " While the principle of data processing is similar, we have to adapt it to certain particularities, such as the position of the sensors on the neck, which restricts the information we can gather. "

In addition, the deep learning models used here - like all systems of this type - required a training dataset for their learning phase. In animals, however, such a resource remains scarce. "That' s why the creation of a database on the activity of pets, particularly dogs, is one of the outputs of the thesis ", says Ioana Ocnarescu. To this end, in addition to the information gathered via sensors and the mobile application, the research team was able to rely on video streams: each participant in the experiment was filmed for an hour, at the start of the experiment. This provided annotated data that could be used directly.

Relevant technology need not be complex

The unlabelled data were not exploited as part of the thesis, but they represent an interesting avenue for further study. The research team plans to continue its work, in particular to consolidate the results obtained, through experiments involving more participants. " The idea would also be to move on to a higher level of knowledge, by analyzing not only the animal's activity, but also its behavior and interactions with its owner ", reveals Hazar Zilelioglu.

An orientation which does not necessarily imply the use of more sophisticated technologies. " Initially, for our study, we planned to develop complex and expensive robotic tools," points out Ioana Ocnarescu. " And in the end, it turned out that a 'simple' collar fitted with sensors already provided a great deal of information about the relationship between an elderly person and their pet. This is one of the lessons of this thesis: technology is not an end in itself. Sometimes, a solution that doesn't involve a great deal of technical complexity is enough to achieve the desired objective. "

*Source: https://www.frontiersin.org/articles/10.3389/fvets.2020.00293/full