This study presents a synthetic dataset designed to support the validation of joint-level trajectory extraction for pedestrian dynamics measurements. Existing pedestrian datasets commonly provide pedestrian-level trajectories, but rarely include frame-wise 3D body-joint coordinates that can be used to evaluate joint-level trajectory extraction and microscopic step measurements. To address this gap, this study developed a dataset containing 147 rendered video sequences, comprising 18,669 video frames in total, of a single animated pedestrian walking along a straight path at three walking speeds. The same walking motions were rendered from 49 virtual cameras placed at different distances, heights, and viewing angles, enabling investigation of camera configuration effects. For each video, the dataset provides frame-wise ground-truth 3D body-joint coordinates in both the world and camera coordinate systems, together with camera intrinsic and extrinsic parameters. By combining rendered videos, known camera parameters, and frame-wise body-joint coordinates, this dataset provides a controlled benchmark for developing and validating vision-based methods for extracting 3D joint-level pedestrian trajectories and deriving step measurements from video. The controlled variation in camera configurations can provide guidance for camera placement in real-world pedestrian experiments.

Real-time human pose estimation and tracking on monocular videos is a fundamental task in computer vision with a wide range of applications. Recently, benefiting from deep learning-based methods, it has received impressive progress in performance. Although some works have reviewed and summarised the advancements in this field, few have specifically focused on real-time performance and monocular video-based solutions. The goal of this review is to bridge this gap by providing a comprehensive understanding of real-time monocular videobased human pose estimation and tracking, encompassing both 2D and 3D domains, as well as single-person and multi-person scenarios. To achieve this objective, this paper systematically reviews 68 papers published between 2014 and 2024 to answer six research questions. This review brings new insights into computational efficiency measures and hardware configurations of existing methods. Additionally, this review provides a deep discussion on trade-off strategies for accuracy and efficiency in real-time systems. Finally, this review highlights promising directions for future research and provides practical solutions for real-world applications.

As more buildings around the world are becoming accessible, the question of egressibility, i.e., accessible evacuation, is important to assess. To achieve egressibility for people who are unable to use the stairs, areas of refuge or occupant evacuation elevators are the two dominant evacuation strategies identified in previous research. A key challenge with these strategies is that they require certain behaviours from the evacuees to be successful. Thus, the aim of this scoping review is to summarise and analyse the research performed within the area, with a focus on the behavioural aspects, in order to identify gaps in the knowledge and provide a foundation for further research. Over 5,000 papers were screened, and a total of 34 papers were selected for in-depth analysis. The review concludes that behavioural aspects related to areas of refuge have not been given much research attention, and the few studies published are based on evaluation of hypothetical scenarios. Evacuation elevators have received more research attention, but the majority of the studies published involve hypothetical scenarios. Hypothetical scenario experiments, sometimes called behavioural intent experiments, can be argued to have low validity. Thus, this review highlights the need for further research on behavioural aspects and design of systems that support the use of both evacuation elevators and areas of refuge. It also highlights that where the preferred methodologies of field or case studies are not readily available to fill the knowledge gap, the use of laboratory and VR methodologies, particularly where they gather perspectives from intended end users, have significant potential to progress understanding on how the use of areas of refuge and occupant evacuation elevators can contribute to egressibility.

Natural hazards, such as wildfires, pose a significant threat to communities worldwide. Real-time forecasting of travel demand during wildfire evacuations is crucial for emergency managers and transportation planners to make timely and better-informed decisions. However, few studies focus on accurate travel demand forecasting in large-scale emergency evacuations. To tackle this research gap, the study develops a new methodological framework for modeling highly granular spatiotemporal trip generation in wildfire evacuations by using (a) large-scale GPS data generated by mobile devices and (b) state-of-the-art AI technologies. Based on the travel demand inferred from the GPS data, we develop a new deep learning model, i.e., Situational-Aware Multi-Graph Convolutional Recurrent Network (SA-MGCRN), along with a model updating scheme to achieve real-time forecasting of travel demand during wildfire evacuations. The proposed methodological framework is tested using a real-world case study: the 2019 Kincade Fire in Sonoma County, CA. The results show that SA-MGCRN significantly outperforms all the selected state-of-the-art benchmarks in terms of prediction performance. Our finding suggests that the most important model components of SA-MGCRN are weekend indicator, population change, evacuation order/warning information, and proximity to fire, which are consistent with behavioral theories and empirical findings. SA-MGCRN can be directly used in future wildfire events to assist real-time decision-making and emergency management.

Characterising the stop/start walking process of individuals in a crowded and congested space is an important consideration in modelling pedestrian movement. However, the reaction of pedestrians to speed changes, especially to the person in front, has not been fully characterised nor quantified for full adoption in computer models. This study, therefore, explored the different phases of the stop/start process through a series of novel experiments conducted at University College Dublin (UCD) in which individual movements were captured precisely using motion capture equipment. The overall aim of the study was to develop a novel methodology (inspired from vehicle traffic flow) to break down and quantify the components of the stop/start walking process, i.e., the perception-reaction time, slow-down time, and start-up time of individuals walking in a single-file. These times, together with the total stopping distance and inter-person distances of each follower to their leader at the beginning and the end of each phase were quantified successfully and their inter-relationships were explored. The results showed the mean perception-reaction time, slow-down time, and start-up time delay were 0.48, 0.58, and 0.39 s, respectively. Where applicable, normal, lognormal, Weibull, gamma, and log-logistic probability distributions were fitted to the data to determine the best fit. The novel methodology developed in this study can be used in the future to investigate pedestrians' behaviour in response to any changes in leaders' speed, i.e. quantify the reaction of individuals in different phases. The results of this study can inform the representation of the stop/start process in microscopic pedestrian models.

Background Live fuel comprises a significant portion of the fuel consumed in forest and scrub crown fires. However, its flammability remains poorly understood. Although live fuel differs from dead fuel in moisture content, chemical composition, cellular structure and physiological characteristics, its higher moisture content masks the effect of other characteristics on its flammability. Aims The aim of the study was to delineate and assess the effects of live/dead condition, moisture content and particle size on flammability of gorse (Ulex europaeus L.). Methods Live and dead gorse material of three size classes (0–3, 3–6, and 6–10 mm in diameter) at six moisture contents (0, 10, 25, 50, 75 and 100%) was tested in a cone calorimeter to evaluate its flammability using new sample preparation and moisture conditioning techniques. Key results On average, live fuel ignited 21% slower, reached 11% higher peak heat release rate, and had a 12% shorter burn duration than dead fuel of the same moisture content. These differences were most pronounced in coarser material. Conclusions For gorse, fine dead fuels increase the likelihood of ignition, fine live fuels contribute to high burning intensities, and coarser live and dead fuels prolong combustion. Implications These findings highlight the need to account for flammability differences between live and dead fuels in fire behaviour models beyond those driven by variations in moisture content.

Emergency fire situations in tunnels can be especially dangerous when occurring in long underground or subsea tunnels, particularly when evacuation on foot is the only alternative. This paper presents the results from a study comparing different visual and acoustic measures to facilitate efficient and safe emergency evacuation and their effect on people's self-rescue behaviour in response to a tunnel fire. Eighty-one participants evaluated seven different scenarios in virtual reality with or without visual and acoustic supporting measures (i.e. signs, lights, acoustic beacons) to find their way to emergency doors. Objective behavioural data, such as orientation, and walking speed, were collected. The results suggest that the distance between the emergency doors increases uncertainty and affects the time to self-rescue significantly, with four times longer times for 500 m than 250 m between doors. Additionally, the use of continuous guiding lights positively supported orientation and walking speed, with 97 % of the participants finding their way and showing a reduction of time to reach the emergency door of 10-20 s. The study underscores the importance in the proper visual and acoustic evacuation measures for the wayfinding of emergency exits, improving self-rescue of people.

Virtual reality allows creating highly immersive visual and auditory experiences, making users feel physically present in the environment. This makes it an ideal platform to simulate dangerous scenarios, including fire evacuation, and study human behaviour without exposing users to harmful elements. However, human perception of the surroundings is based on the integration of multiple sensory cues (visual, auditory, tactile, or/and olfactory) present in the environment. When some of the sensory stimuli are missing in the virtual experience, it can break the illusion of being there in the environment and could lead to actions that deviate from normal behaviour. In this work, we added an olfactory cue in a well-documented historic hotel fire scenario that was recreated in VR, and examined the effects of the olfactory cue on human behaviour. We conducted a between subject study on 40 naive participants. Our results show that the addition of the olfactory cue could increase behavioural realism. We found that 80% of the studied actions for the VR with olfactory cue condition matched the ones performed by the survivors. In comparison, only 40% of the participants’ actions for VR only condition were similar to the survivors.

With increasing rates of elderly and obese people in the population, questions are being raised about the validity of inputs used by computer evacuation models to predict the movement of crowds in the built environment. The objective of this study is to examine the movement of individuals in a VR environment. Exploring individual movement in VR (where the individual is exposed to a virtual environment with virtual agents while actually moving alone in the physical environment) is a necessary step on the path to determining if VR is a useful tool to gather new crowd movement data. Specifically, this work presents the results of two experiments that were conducted to measure the correlation between inter-person distance (the distance from a participant to a virtual agent) and walking speed. Results show a positive correlation between walking speed and the inter-person distance for inter-person distances between 1.0 and 1.5 m. Above inter-person distances of 1.5 m, walking speed was not dependent on inter-person distance. An important finding from this work is no observed significant difference in the relationship between walking speed and inter-person distance across both experimental setups - 'pushing' or 'following' configurations. Finally, this work shows the potential of gathering individual movement data using VR.

Accurately estimating movement through smoke is critical for fire safety design. However, physical laboratory experiments using artificial smoke can be very expensive. Virtual reality (VR) can potentially be an alternative method to cheaply investigate movement through smoke. The primary objective of this study is to determine the relationship between movement speed and visibility in VR. The secondary objective was to compare movement through smoke in VR to existing experimental data. Five scenarios were tested: real-world unimpeded movement, unimpeded movement in VR, and movement in VR with virtual smoke reducing the visibility to 3.5 m, 2.5 m, and 1.5 m. A wireless head mounted display was used to immerse participants in the virtual environment. For the study, 42 participants experienced the smoke scenarios in a random order. The results indicate that movement speed decreased with visibility, but to a lesser extent than in previous physical laboratory experiments with artificial smoke. Unimpeded movement in VR was shown to be significantly slower than real-world unimpeded movement. Prior experience with VR was not shown to have a significant impact on unimpeded VR movement speed. Increasing the realism of the virtual environment could potentially better align results from VR experiments with past physical laboratory experiments.

To develop effective wildfire evacuation plans, it is crucial to study evacuation decision-making and identify the factors affecting individuals’ choices. Statistic models (e.g., logistic regression) are widely used in the literature to predict household evacuation decisions, while the potential of machine learning models has not been fully explored. This study compared seven machine learning models with logistic regression to identify which approach is better for predicting a householder’s decision to evacuate. The machine learning models tested include the naïve Bayes classifier, K-nearest neighbors, support vector machine, neural network, classification and regression tree (CART), random forest, and extreme gradient boosting. These models were calibrated using the survey data collected from the 2019 Kincade Fire. The predictive performance of the machine learning models and the logistic regression was compared using F1 score, accuracy, precision, and recall. The results indicate that all the machine learning models performed better than the logistic regression. The CART model has the highest F1 score among all models, with a statistically significant difference from the logistic regression model. This CART model shows that the most important factor affecting the decision to evacuate is pre-fire safety perception. Other important factors include receiving an evacuation order, household risk perception (during the event), and education level.

Elevator evacuation is associated with several safety benefits and can be a cost-efficient evacuation strategy in many buildings or facilities. Despite the advantages, elevator evacuation is not yet a common strategy in the built environment. In this paper, a design strategy that allows the designer to consider information needs of evacuees when designing the information and guidance measures for evacuation elevators is presented. The strategy is based on human behaviour theories and the Theory of Affordance. The use of the design strategy is also illustrated for a case study, i.e., an underground metro station with evacuation elevators. This case study identifies several important measures, which accommodate the information needs of the evacuees, e.g., (1) voice alarm containing information that elevators can be used for evacuation, and (2) a system in the elevator lobby indicating that the elevators are operational. The paper also highlights future research areas where the current level of knowledge related to elevator evacuation needs to increase.

The threat of wildfires is increasing at an alarming rate due to climate change and the expansion of the wildland-urban interface. It is critical to improve understanding of people's evacuation decisions during wildfire emergencies. Therefore, this study proposes a novel methodology to model evacuation rates using large-scale GPS data generated by mobile devices. We first overlay socio-demographic and built environment attributes-aggregated at the census-block-group-level-to the inferred home locations of the mobile device users. We then develop a linear regression model to examine how the socio-demographic and built environment variables affect evacuation rates across census block groups. We apply the GPS data (44.2 million signal records from over 5000 devices) collected during the 2019 Kincade Fire in Sonoma County, California to evaluate the proposed methodology. The results of the model are generally consistent with findings of a prior survey of the same fire event. We also include factors in our model that are rarely measured through previous methods and find several built environment factors such as distance to the fire, land parcel size, and residing in a high fire risk area to be correlated with evacuation rates. Another notable finding is that people living in urban block groups, block groups with a higher median age, and block groups with a higher average level of educational attainment are more likely to evacuate. This research shows that the use of GPS data is a valuable complement to existing methods in wildfire evacuation research, and provides new insights to improve evacuation planning.

Recently, wildfires have created severe challenges for fire and emergency services and communities in the wildland-urban interface (WUI). To reduce wildfire risk and enhance the safety of WUI communities, improving our understanding of wildfire evacuation is a pressing need. This study proposes a new methodology to analyze wildfire evacuation by leveraging a largescale GPS dataset. This methodology includes a proxy-home-location inference algorithm and an evacuation-behavior inference algorithm, to systematically identify different groups of wildfire evacuees (i.e., self-evacuee, shadow evacuee, evacuee under warning, and ordered evacuee). We applied the methodology to the 2019 Kincade Fire in Sonoma County, CA. We found that among all groups of evacuees, self-evacuees and shadow evacuees accounted for more than half of the evacuees during the Kincade Fire. The findings of this study can be used by emergency managers and transportation planners to better prepare WUI households for future wildfire events.

Urban-scale evacuation may take place because of disasters or emergencies. Efforts have been made to enhance the preparedness of communities for urban-scale evacuation. For instance, wayfinding systems are installed and implemented in tsunami-prone regions, indicating the evacuation routes to high ground or inland. However, communities tend not to familiarise themselves with wayfinding systems and the best evacuation routes because tsunami evacuation drills are not normally carried out given the challenges to plan and run them. This study proposes a rapid development approach for immersive virtual reality (IVR) training systems suited to urban-scale evacuation. This approach utilises 360-degree panoramas to represent an urban environment in IVR, getting rid of the process of 3D modelling or reality capture to reconstruct a virtual urban environment. The 360-degree panoramas used in this study were directly acquired via a 360-degree camera. Immediate feedback is applied as a pedagogical approach to inform users. The training objective is to make users capable of identifying evacuation signs and the best evacuation route. This paper outlines a development framework to demonstrate the prototyping workflow of a 360-degree panoramic IVR training system suited to urban-scale evacuation. 360-degree panoramic IVR requires low levels of development efforts and computational resources. Therefore, urban-scale evacuation drills become possible to be rolled out easily and quickly to a wider population using 360-degree panoramic IVR.

As the threat of wildfire increases, it is imperative to enhance the understanding of household evacuation behavior and movements. Mobile GPS data provide a unique opportunity for studying evacuation routing behavior with high ecological validity, but there are little publicly available data. We generated a highway vehicle routing dataset derived from GPS trajectories generated by mobile devices (e.g., smartphones) in Sonoma County, California during the 2019 Kincade Fire that started on October 23, 2019. This dataset contains 21,160 highway vehicle routing records within Sonoma County from October 16, 2019 to November 13, 2019. The quality of the dataset is validated by checking trajectories and average travel speeds. The potential use of this dataset lies in analyzing and modeling evacuee route choice behavior, estimating traffic conditions during the evacuation, and validating wildfire evacuation simulation models.

This study aims to quantify and develop a deeper understanding of the parameters that underpin the development of a new, predictive, microscopic model of pedestrian movement with the potential to accurately reflect the complexity of flow dynamics now and into the future. It presents the results and analyses of two single file experiments designed to quantify the physical space taken up by the extent of a person's stepping movement (maximum step extent) and the minimum distance between points of inter-person contact (contact buffer) across a range of walking speeds. The experiments successfully used high-resolution optical motion capture and enhanced video analysis to quantify the dynamic changes in gait and spatial parameters, which were manifested as overlapping steps, and changes to step extent, step length, step frequency, and contact distance. The sum of the step extent and contact buffer, at different speeds, was found to be within a few centimetres of the inter-person distance (headway), leading to the conclusion that these parameters are therefore key components for the derivation of inter-person spacing and, hence, overall crowd movement. The work informs the longer term aim of developing the mathematical model which has the potential to include pedestrian demographics, walking ability and cognitive capabilities. (c) 2022 The Authors. Published by Elsevier B.V.