This study explores the electricity consumption patterns in institutional buildings at a university campus in Greater Paris, employing time-series analysis to assess the effects of major disruptions like the COVID-19 pandemic and the subsequent energy crisis triggered by the war in Ukraine. Utilizing dynamic linear modeling, our research achieved a Mean Absolute Percentage Error of 6.2 %, demonstrating the model’s efficacy in capturing complex consumption patterns under fluctuating conditions. We observed a 13 % reduction in electricity usage during the initial pandemic lockdowns, with varying responses in subsequent phases due to the energy crisis. These results underscore the critical role of adaptive energy policies and advanced modeling techniques in enhancing energy efficiency and resilience in educational institutions. By providing detailed insights into the factors influencing energy usage, our study assists policymakers and energy managers in devising robust strategies to improve energy management and meet sustainable development goals, particularly in response to external shocks.

Extensive research exists on implementing new algorithms for population synthesis; however, there remains no consensus on how to evaluate a generated synthetic population. The generated population must be similar to the real population. That involves ensuring a similar distribution at a macroscopic level, and that each generated individual is realistic at the microscopic level. Given that population synthesis involves real data, another concern is to respect the data privacy, which must be also evaluated. In this paper, leveraging insights from population synthesis and tabular data generation literature, we propose three metrics, that address the current limitations in evaluation. To support our metrics, we provide some mathematical demonstrations, and the robustness of the metrics are tested, whenever possible. In addition to these metrics, we propose an open-source dataset that we consider well fitted for model evaluation. This allows proposing a complete evaluation framework for population synthesis promoting reproducible science. As an application, we present an extensive benchmark with probabilistic models, and deep generative models, including-the current state-of-the-art generative model for tabular data synthesis-diffusion model. In this benchmark, we simulate two data scenarios corresponding to real-world use cases. The results indicate that, under these settings, Bayesian networks are the most interesting models, performing well for all criteria. Regarding deep generative models, the diffusion model gives promising results, as it is the only model achieving similar results to those of the Bayesian Network, and could be a great option in more complicated use cases, with more attributes. Our code is available here, with our evaluation methodology and implementation of models, for future research: https://github.com/vdarsel/PopulationSynthesis.

Shared Mobility Services (SMSs) are transforming urban transportation systems by offering flexible travel options. These services, which help reduce the number of cars on the roads, have the potential to enhance the transportation system’s performance, leading to improvements in travel times and emissions. This emphasizes the importance of assessing their impact on the system and users’ choices, particularly when integrated into complex multimodal systems that include public transport (PT). However, many studies overlook the synergies between SMSs and PT, leading to inaccurate traffic estimations and planning. This research presents an extensive review of multimodal transportation system models incorporating SMSs. It then introduces a multimodal traffic assignment model including a broad range of mobility options in urban transportation systems applicable in both continuous and integer settings, leading to a Mixed-Integer Bilinear Programming (MIBLP) formulation. The model comprises diverse travel options, including SMSs, and accounts for intermodality by allowing commuters to combine modes to optimize time and monetary expense. An in-depth examination of commuters’ mode and path choices on two test cases and an analysis of the price of anarchy reveals the disparities between user equilibrium and system optimum in such complex networks.

In multimodal transportation systems, shared mobility services (SMSs) are promoted for their potential to enhance flexibility and reduce congestion. However, SMS demand is often concentrated in high-density areas, which can limit the effectiveness and accessibility for various commuter groups. This uneven integration challenges transportation system efficiency, especially in terms of emissions and spatial equity. Addressing these issues requires coordination among multiple stakeholders whose objectives frequently conflict. Whereas authorities aim to ensure sustainable and equitable mobility, SMS providers focus on revenue maximization, and travelers seek to minimize personal travel costs. This paper proposes a multi-agent deep reinforcement learning framework that captures these interactions through dynamic pricing and incentivization strategies for SMSs and public transport. The framework integrates two reinforcement learning (RL) agents: (i) a public authority that allocates spatio-temporal public transport incentives to improve equity, emissions, and efficiency, and (ii) an SMS provider that dynamically adjusts fares to optimize revenue. The agents interact with the transportation system and adapt strategies in response to evolving demand, congestion, and network conditions. Numerical experiments conducted over a three-hour morning peak period show that dynamic incentivization effectively reduces congestion peaks, lowers commuters' costs by around 20

Belief Functions Theory (BFT) supports rich uncertainty modeling and information fusion, but its effectiveness hinges on the management of unreliable or conflicting evidence. Discounting operators address this issue by adjusting source contributions according to their reliability. However, the field still lacks a structured and up-to-date survey that clearly distinguishes existing operators and provides guidance on their selection. This paper fills that gap through two complementary taxonomies: a multidimensional characterization of operator behavior and a classification based on the information used to derive discount rates. We review representative operators, analyze their mathematical structure and practical requirements, and compare their qualitative behavior across key operational dimensions. The survey provides clear guidance for selecting and deploying discounting operators in systems that must handle unreliable evidence under diverse constraints.

This paper presents an advanced Federated Learning (FL) framework for forecasting complex spatiotemporal data, improving upon recent state-of-the-art models. In the proposed approach, the original Gated Recurrent Unit (GRU) module within previous Dynamic Spatial-Temporal Graph Convolutional Recurrent Network (DSTGCRN) modeling is first replaced with a Long Short-Term Memory (LSTM) network, enabling the resulting model to more effectively capture long-term dependencies inherent to time series data. The resulting architecture signif icantly improves the model's capacity to handle complex temporal patterns in diverse forecasting applications. Furthermore, the proposed FL framework integrates a novel Client-Side Validation (CSV) mechanism, introduc ing a critical validation step at the client level before incorporating aggregated parameters from the central server into local models, ensuring that only the most effective updates are retained and improving both the ro bustness and accuracy of the forecasting model across clients. The efficiency of our approach is demonstrated through extensive experiments on real-world applications, including public datasets for multimodal transport de mand forecasting and private datasets for Origin-Destination (OD) matrix forecasting in urban areas. The results demonstrate substantial improvements over conventional methods, highlighting the framework's ability to cap ture complex spatiotemporal dependencies while preserving data privacy. This work not only provides a scalable and privacy-preserving solution for real-time, region-specific forecasting and management but also underscores the potential of leveraging distributed data sources in an FL context. We provide our algorithms as open-source on GitHub.1

The implementation of real-time gait mode detection is paramount for providing tailored support to individuals utilizing actuated ankle-foot orthoses (AAFOs), enhancing their walking and mobility. However, existing systems often rely on multiple sensors and struggle with accurate and prompt detection of gait transitions, especially in varied environments. This study develops a novel real-time gait mode detection system that accurately identifies five daily living gait modes including level walking, ramp ascent and descent, and stair ascent and descent using only two foot-mounted inertial measurement units. A long short-term memory based algorithm, trained on data from ten healthy subjects, extracts six kinematic features to predict gait modes. The proposed method integrates this detection system with a taskoriented control strategy to adapt AAFO control according to identified gait modes. Real-time experiments with three healthy participants demonstrated robust gait mode detection, achieving an average accuracy of $98 \pm 1$% across the five modes, even under assistive torque. In trials mimicking abnormal gait, the system maintained an accuracy of $93 \pm 3$%. Additionally, transition delays were analyzed, showing detection can occur between transitions of the leading and trailing foot. The control strategy reduced dorsiflexor and plantar-flexor muscle activation, measured by electromyography, and improved swing phase tracking performance. Detection robustness was further evaluated by walking with obstacles and changes in environmental dimensions.

Congestion remains one of the most prevalent transport problems in major cities, with social equity being a pivotal aspect in shaping solutions. This study presents a two-stage framework that combines machine learning with social justice principles to identify critical commuters who can adjust their workplace arrival times. This framework addresses both traffic management and social equity concerns, considering synchronisation needs at the employer, household, and individual levels. Our proposed framework can identify critical commuters based on their basic information collected by the employer. We benchmark multiple machine learning approaches to model and predict an individual's ability to shift their workplace arrival times. Ultimately, we frame the problem as a classification task and select gradient boosting due to its superior performance. Using employee survey data from Rennes Metropole in France, we identify the key factors that influence individual's flexibility in their arrival times. Regular school drop-offs are the most significant factor, followed by theoretical arrival time contracts with employers and, to a lesser extent, age and income. Building on these findings, we apply the Rawlsian "Min-Max" fairness principle rooted in social science to refine the subpopulation of commuters with theoretical shift abilities and assess their practical likelihood of shifting within a socially equitable framework. By integrating machine learning insights with social equity considerations, this framework offers an interdisciplinary approach to potentially mitigate congestion, ensuring that policies not only address traffic demand management but also promote fairness and inclusion, supporting their long-term effectiveness.

This study presents a comparative analysis of three AI models for real-time human intention detection in the presence of assistive torque during the swing phase. The models were designed to classify five gait modes, including level walking, stair ascent, stair descent, ramp ascent, and ramp descent. The evaluated models include K-Nearest Neighbors (KNN), Random Forest, and Gated Recurrent Unit (GRU), all trained on kinematic features extracted from raw IMU data collected from eight healthy participants. Real-time validation was conducted with an unseen subject. Among the three models, GRU demonstrated the highest accuracy, achieving 96% in real-time conditions. The Random Forest model also exhibited strong performance, reaching an accuracy of 94%, while KNN showed a significant decline in accuracy. Additionally, an assistive torque based on trajectory tracking was applied during the swing phase for each detected gait mode. A proportional-integral-derivative (PID) controller was implemented to refine the tracking process, resulting in a 45% reduction in root mean square error across all gait modes. Copyright (c) 2025 The Authors. This is an open access article under the CC BY-NC-ND license (https://creativecommons.org/licenses/by-nc-nd/4.0/)

In this work, we investigate the relevance of IoT data use for one-day-ahead electric demand forecasting at the building level. In particular, we use hourly timestep experimental electric demand data in the case of an academic building. The dataset spans three years and includes ten measured data types from more than 40 sensors, such as temperature, relative humidity, illumination, CO2 levels, and occupancy. To evaluate the impact of each data type, we created 10 separate datasets corresponding to one of them. For each dataset, we implemented Long Short-Term Memory (LSTM). The models are evaluated based on two main criteria: statistical accuracy in regression performance and adherence to physical constraints, such as maximum power limits. Thus, we aim to assess the relevance of IoT devices use for electric demand forecasting independently of specific models used. Our results highlight the significant role of occupancy data in improving prediction accuracy, particularly those that measure building-level occupancy rather than local room-specific data. Additional findings include optimal times for forecasting daily energy consumption, which suggest initiating predictions either early in the morning or in the evening for more efficient forecasting. This analysis provides insights into the temporal relationships between different variables. It offers a deeper understanding of how IoT data can be used for medium-term predictions, extending up to one week. We observe that IoT data helps models extrapolate energy demand over more extended periods, thereby reducing cumulative error.

The distribution of high-quality drinking water is a major challenge for water distribution network operators. In this context, we propose an algorithm for contamination warning systems that partitions the network into areas of homogeneous water quality, thereby facilitating the identification of contaminated areas. To this end, we introduce a dedicated model-based approach that simultaneously clusters and aligns the conductivity signals measured at the network probes. We use the shifts inferred by the model to summarize the water propagation in a cluster. To produce physically meaningful clusters, we integrate expert knowledge about the physics of the signals into the curve clustering problem using a Hidden Markov Random Field (HMRF). We present a methodology for model selection and apply our approach to real conductivity curves from a large water distribution network.

The ACM SIGSPATIAL International Conference on Advances in Geographic Information Systems held a half-day workshop on Sustainable Urban Mobility. This was the second Workshop of this series, after a successful initial event at ACM SIGSPATIAL in 2023 in Hamburg [1]. The workshop explicitly considers urban mobility as an open, complex system, and hence points to a wicked problem: estimating the global impact of any intervention to a mobility system that is aiming for less emissions, less vehicles on the road and less single-person cars, or any other sustainability goal. Often, such interventions show unintended consequences. For example, efficiency gains in one mode of mobility may show rebound effects on that mode as well as feedback loops with other modes. While we typically ignore such interrelations as "too difficult", the workshop set out to explore what the spatial computing community can contribute.

The dynamics of urban transportation can be captured using activity-based models, which rely on travel demand data to get a comprehensive understanding of urban mobility. This data is usually derived from population samples and Household Travel Surveys (HTSs), which can be expensive and as a result, are conducted only every 5 to 10 years. Moreover, due to their limited reach, they are not adapted to represent the spatio-temporal structure of the flows of the total population. This calls for complementary data sources that could be used to update old surveys to cut costs and to estimate the global spatial mobility behavior of the population. In this paper, we propose steps in the state-of-the-art pipeline for travel demand synthesis with an approach for the temporal calibration and the location attribution based on time-dependent origin–destination (OD) matrices. These matrices describe the flows between zones of a city. This methodology is illustrated on the city of Lyon, France, with OD matrices estimated from the mobile phone activity of the subscribers of French telecom operator Orange. We explore how the spatialization can be performed using various probabilistic graph models whose parameters are evaluated via the OD matrices. The structure of the models enforces the consistency of the locations with the chains of activities, such as the fact that two “home” activities must have the same location. Multiple models are proposed, corresponding to different compromises between the two potentially incompatible sources that are HTS and mobile data. We show that while a very naive spatialization approach allows the generation of synthetic travel demand that perfectly fits the flows described by the OD matrices without respecting the consistency of the locations, the other proposed approaches offer much more realistic agendas at the expense of only small discrepancies with the mobile data.

Through a combination of regulations, fear of contagion, and changes in travelers’ habits, the COVID-19 pandemic affected the mobility of public-transit ridership worldwide. To understand the longer-term effects of the pandemic on public-transit ridership, we focus on the case of Paris, France, thanks to an open 5 year record of entries into more than 500 stations. To deal with the large volume of data, we use a statistical model that performs clustering and segmentation simultaneously while incorporating many exogenous variables, such as the day of the week or lockdowns, to account for their effect on the number of entries. We carry out an in-depth analysis of the results for the segments and clusters. Examining and comparing the regression coefficients across clusters and consecutive segments allows us to draw per-cluster and per-segment conclusions. We show that the number of weekday trips decreased in most clusters and that the reduction in weekly variations is proportional to the share of weekday trips in the volume of entries before the pandemic. In addition, we characterize the changes in the weekly profile: Thursday was replaced by Tuesday as the day with the highest ridership; because of teleworking, Friday became the least crowded weekday in clusters with strong differences between weekdays and weekends, while the lowest ridership weekday remains Monday in the other clusters.

In this paper, we analyze electricity consumption dynamics in institutional buildings, focusing on a university building in the Paris Region, France. Leveraging Dynamic Linear Modeling (DLM), we investigate temporal patterns and factors influencing electricity usage, including trends, yearly and weekly seasonality, and external variables. Our methodology, applied to field data collected over several years at a sub-hourly time step, enables accurate representation and prediction of consumption behaviors. Results highlight a discernible decrease in electricity consumption during COVID-19 lockdown periods, with reductions averaging 13.9 % during the initial lockdown and 6.1 % and 5.3% during subsequent phases. Variations in lockdown severity influence consumption patterns, emphasizing the building's electricity demand response to external shocks and the importance of adaptive energy management strategies. These findings contribute to understanding energy dynamics in institutional buildings and inform strategies for enhancing building energy performance and sustainability, benefiting from field data collected and provided by advanced energy metering infrastructure.

Forecasting counts data in transportation areas can enrich passenger information for public transport passengers, who may thus better plan their trips. Moreover, forecasting with uncertainty is particularly important in the transportation domain, where the risk of poorly managed high demand is to be avoided. In this paper, we propose a new probabilistic prediction model well-suited for multivariate, overdispersed, and possibly correlated count data. This model combines the strength of the deep learning framework with the modeling of counts data allowed by “sums and shares” distributions. Indeed, deep learning models can handle uncertainty by relying on an abstraction of contextual data and by assuming output distributions. Our model learns a latent representation of the input data with the help of a recurrent neural network and then translates it into multivariate count predictions with a sums and shares distribution, well suited to tackle multivariate overdispersed and correlated count data. An extensive benchmark of the proposed model is carried out. We compare this model with seven others from the state-of-the-art probabilistic forecasting models using five open-source data (bikes, taxis, railways, traffic, wikipedia) and a specific use case on pedestrian counts within a multimodal transport hub in the Paris Region. Our model outperforms other models in situations where the data present temporal regularities. The results also highlight the potential of our model in the specific use case. Moreover, this forecasting represents an interesting way to predict short-term pedestrian counts in response to different events, such as concerts or transport disruptions.

The Covid-19 pandemic drastically changed urban mobility, both during the height of the pandemic with government lockdowns, but also in the longer term with the adoption of working-from-home policies. To understand its effects on rail public transport ridership, we propose a dedicated Regression Mixture Model able to perform both the clustering of public transport stations and the segmentation of time periods, while ignoring variations due to additional variables such as the official lockdowns or non-working days. Each cluster is thus defined by a series of segments in which the effect of the exogenous variables is constant. As each segment within a cluster has its own regression coefficients to model the impact of the covariates, we analyze how these coefficients evolve to understand the changes in the cluster. We present the regression mixture model and the parameter estimation using the EM algorithm, before demonstrating the benefits of the model on both simulated and real data. Thanks to a five-year dataset of the ridership in the Paris public transport system, we analyze the impact of the pandemic, not only in terms of the number of travelers but also on the weekly commute. We further analyze the specific changes that the pandemic caused inside each cluster.