A Sequential Multimodal Variational Autoencoder Framework for State Prediction in Synthetic Multi-Agent Honey Bee Colony Simulations

Automated bee colony monitoring is a maturing field of research in the Industry 4.0 approach to apiculture. Despite the great efficacy of models used for various tasks and some attempts at representation-based solutions for bee state modelling, no concrete framework has been proposed for time series predictions of bee colony states. This work provides a theoretical framework for modelling the behaviour of a simulated bee colony using a multimodal representation learning approach. For this goal, a synthetic bee colony model was devised, simulating the behaviour of individual bee agents at scale. The resulting model was used to create a dataset of simulated readings equivalent to measurements over multiple years in multiple apiaries. The synthetic dataset was then analysed using a MMVAE-LSTM (Multi-Modal Variational Autoencoder - Long-Short Term Memory) model. The multimodal approach was verified in comparison to classical methods of modelling and was found to outperform them significantly. The method has shown great resilience to missing modalities during inference, providing proof the latent space representation contains essential bee colony state information. The presented approach will allow for analysis of future large datasets combining colony weight measurement with the modular multimodal approach allowing for less rigid modality choice when planning large scale bee colony modelling experiments.

Authors

Institutions

Publication Details

Journal
Research Square
Published
2026-11-20
DOI
https://doi.org/10.21203/rs.3.rs-10831035/v1
Primary Topic
Smart Agriculture and AI
Type
preprint
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
preprint

A Sequential Multimodal Variational Autoencoder Framework for State Prediction in Synthetic Multi-Agent Honey Bee Colony Simulations

Piotr Książek, Jan Przepióra
Research Square
Smart Agriculture and AI
preprint

A Sequential Multimodal Variational Autoencoder Framework for State Prediction in Synthetic Multi-Agent Honey Bee Colony Simulations

Piotr Książek, Jan Przepióra
preprint en

Abstract

Automated bee colony monitoring is a maturing field of research in the Industry 4.0 approach to apiculture. Despite the great efficacy of models used for various tasks and some attempts at representation-based solutions for bee state modelling, no concrete framework has been proposed for time series predictions of bee colony states. This work provides a theoretical framework for modelling the behaviour of a simulated bee colony using a multimodal representation learning approach. For this goal, a synthetic bee colony model was devised, simulating the behaviour of individual bee agents at scale. The resulting model was used to create a dataset of simulated readings equivalent to measurements over multiple years in multiple apiaries. The synthetic dataset was then analysed using a MMVAE-LSTM (Multi-Modal Variational Autoencoder - Long-Short Term Memory) model. The multimodal approach was verified in comparison to classical methods of modelling and was found to outperform them significantly. The method has shown great resilience to missing modalities during inference, providing proof the latent space representation contains essential bee colony state information. The presented approach will allow for analysis of future large datasets combining colony weight measurement with the modular multimodal approach allowing for less rigid modality choice when planning large scale bee colony modelling experiments.

Research Square
Wrocław University of Science and Technology (PL), AGH University of Krakow (PL)
Smart Agriculture and AI
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon