This text is part of the special Research section
Since 2019, an innovative project led by Tiago Falk, from INRS, and Pierre Giovenazzo, from Université Laval, has been exploring the possibilities of artificial intelligence and machine learning with the aim of mitigating population decline. of bees and to facilitate the work of beekeepers.
Designed in collaboration with the Deschambault Animal Science Research Center (CRSAD) and Nectar Technologies Inc., the interdisciplinary collaborative project by Tiago Falk and Pierre Giovenazzo is based on multimodal sensor technology installed within the hives. These continuously measure temperature, humidity and sound through a built-in microphone, before sending the data to a cloud and then analyzing it through artificial intelligence.
“The goal of the project is to understand the dynamics of bees, to remotely monitor the impact of the genetic modifications developed at CRSAD and, in the medium term, to use artificial intelligence to be able to determine the health of the hive, predict its honey production and prevent infestations of various parasites”, sums up Tiago Falk.
The research group wants beekeepers to benefit from this valuable data by marketing the sensors and their technology to Nectar Technologies Inc., a Montreal company that already produces a software platform to help bee professionals establish better management practices. of the apiary according to their geographical location and their activities.
From genetics to statistical analysis
The CRSAD research group has been developing a bee genetic selection program since 2010 to promote its adaptation to the Quebec environment, which is, due to its short summer, difficult to navigate for the species, explains Ségolène Maucourt, post-doctoral student in biology and researcher at CRSAD. By measuring various parameters, such as spring development, honey production and overwintering, Pierre Giovenazzo’s team is able to select the colonies that have the best capacity for reproduction.
So far, the technique has proven itself, but the process is tedious and requires a lot of data and, above all, human resources to collect them. Fortunately, multimodal sensors could overcome this problem. The sound volume and the raw audio tapes collected by Tiago Falk’s laboratory, which specializes in voice recognition, allow him to assess, in particular, the quantity of honey produced, the number of bees working and the various environmental factors that affect – or not — beehives (a passing train, air conditioning running, etc.).
“We are developing tools that will bring the machine learning [l’apprentissage automatique] to understand all these conditions in the hive,” says Tiago Falk.Twice a month, the CRSAD team measures and categorizes the data by hand in order to teach the algorithm to recognize it. The goal is to replace this step with the automated system. “There is a lot of work to be done in acoustic analysis. This is our biggest challenge,” continues Marc-André Roberge, co-founder and CEO of Nectar Technologies inc.
Over time, the machine should be able to recognize what data indicates the presence of honey, the noise generated by human conversation or even the infestation of parasites. “Later, if a sound picked up is different, like the rain falling on the hive, for example, the algorithms will be able to process it without having to alert the beekeepers”, specifies Tiago Falk. A second generation of sensors, more resistant to the conditions of migratory beekeeping, should soon be made available to researchers.
save the bees
The parties involved make no secret of it: all this mobilization aims to facilitate the work of beekeepers and reduce their losses to a minimum, but also to save bee colonies, which have been doing rather poorly for several years. “About 60% of bees in Canada have died without warning this year. There are hypotheses, but we don’t know exactly why, ”underlines Tiago Falk.
Behind all this, the objective is also to understand the behavior of bees and the impact of the environment on them. The research team also plans to carry out an environmental analysis using the hives as a bio-indicator. “In the current agricultural system, which operates a lot in large monocultures with little floral diversity, we try to maximize yield without really considering local biodiversity. How can we give a voice to bees and pollinators in this system? This is the broader question we are asking ourselves,” adds Marc-André Roberge.