Thomas Matyus

Can Walking Behavior Be Predicted? Analysis of Calibration and Fit of Pedestrian Models

Several models for simulation of pedestrian movement have been proposed in recent decades. These models are primarily used in the planning and evaluation of large pedestrian infrastructures, such as transportation hubs, with a focus to increase comfort and safety for pedestrians. Although the number of proposed simulation models is increasing at a fast pace, not much is known about the properties of calibration procedures or the transferability of the models estimated in one setting to other settings. This paper compares three calibration methods for a slightly adapted social force model. The main emphasis lies in the characteristics of the data-generation process and the information contained in the data sets. The sensitivity of the model parameters of the calibrated model were investigated, and the transferability of the model to different scenarios was tested. Results revealed that the quality of the data had a strong effect on the suitability of different calibration strategies and that the information content in the scene under investigation limited the transferability of the results to other scenarios. These results suggest that several data sets with different characteristics do not need to be included in the calibration process to achieve a model that performs well in a wider variety of settings.

Comparison of Different Calibration Techniques on Simulated Data

Pedestrian simulation models are used in a multitude of ways. While model development for Social Force models is quite advanced little is known about how to calibrate a Social Force model such that the parameters are reproducing real life data in many situations. This paper tests different calibration methodologies that were previously used in the calibration of Social Force models to give an indication of the usability of such techniques. This is done on simulated data to guarantee the comparability of the different methodologies. The result will give practitioners a guideline in choosing how to use real live data in the calibration process.

Realistic Interactive Pedestrian Simulation and Visualization for Virtual 3D Environments

We present a visual computing effort to realistically and interactively simulate and visualize aspects of human motion behavior in virtual 3D environments. It allows virtually changing the infrastructure of a layout and assessing the consequences in terms of motion paths and visibility (where will people look at?). We first create a virtual 3D model of an infrastructure with photogrammetric reconstruction of images and obtain highresolution video footage of real-world scenarios of the infrastructure. We calibrate video information with the 3D model in order to optimize an automatic human detection and tracking algorithm and to obtain real-world trajectories of people in world coordinates of major ground plane. These real world trajectories provide input to strategic and tactial movement aspects of pedestrian simulations and help validating the human motion simulations. The simulated pedestrian trajectories are the base for path visualizations and visibility analyses in the virtual environment. Our case study comprises an entry hall of a train station, but the approach is applicable to other environments such as museums.

Bridging the gap between visual exploration and agent-based pedestrian simulation in a virtual environment

We present a system to evaluate and improve visual guidance systems and signage for pedestrians inside large buildings. Given a 3D model of an actual building we perform agent-based simulations mimicking the decision making process and navigation patterns of pedestrians trying to find their way to predefined locations. Our main contribution is to enable agents to base their decisions on realistic threedimensional visibility and occlusion cues computed from the actual building geometry with added semantic annotations (e.g. meaning of signs, or purpose of inventory), as well as an interactive visualization of simulated movement trajectories and accompanying visibility data tied to the underlying 3D model. This enables users of the system to quickly pinpoint and solve problems within the simulation by watching, exploring and understanding emergent behavior inside the building. This insight gained from introspection can in turn inform planning and thus improve the effectiveness of guidance systems.

Simulation-Based Forecasts of Crowd Flows at Major Events Using Real-Time Measurements

The complexity and dynamic nature of large events arise the need for decision makers to assess the current situation and to derive multi-temporal forecasts in order to identify critical situations in a timely manner and to initiate appropriate countermeasures. In this work, we present a fast mesoscopic simulation model which incorporates measurements from counting and Bluetooth sensors, thus providing real-time forecasts of crowd flows at major events. With this approach already a sparse placement of sensors at strategic points on an event area is sufficient to achieve the necessary spatial and temporal resolution for a complete characterization of the current crowd flows. For model verification and validation, we investigated case studies from two music festivals in Austria in 2012 and 2013 where extensive measurements on human motion data were obtained to evaluate the deviations of the simulation results from the measured walking times.

Mind the gap: Boarding and alighting processes using the social force paradigm calibrated on experimental data

One of the few possibilities to increase the capacity of train lines is to make boarding and alighting more efficient. So far expensive experiments were needed to evaluate the effect of proposed changes to the train setup. This paper shows that a simulation model based on the social force paradigm and calibrated on measured data promises that in future simulations could be used to test new scenarios in a cheap and fast way.

Simulation of Handicapped People Finding Their Way Through Transport Infrastructures

This paper presents a research effort put into enhancing existing simulation models by including models for the motion and orientation behavior of handicapped people being unfamiliar with a transport infrastructure. On the tactical level the perception of guidance systems is modeled and makes it possible to simulate agent navigation through an unknown infrastructure using the present signage. The guidance information is determined against relevant influencing factors in a simulated virtual 3D environment. For the proof of concept the applicability of the wayfinding algorithm is demonstrated in three different scenarios. Results show that the proposed simulation model facilitates an agent to find its way autonomously through a transport infrastructure based on signage information only. This makes it possible to evaluate the visibility of the guidance system and can reveal areas lacking guidance information for people unfamiliar with the infrastructure especially for elderly and handicapped people with reduced reception capabilities.

Route-Choice Modeling for Pedestrian Evacuation Based on Infrastructure Knowledge and Personal Preferences

In recent years, pedestrian simulation has been a valuable tool for the quantitative assessment of egress performance in various environments during emergency evacuation. For a high level of realism, an evacuation simulation requires a behavioral model that takes into account behavioral aspects of real pedestrians. In many studies, however, it is assumed that simulated pedestrians have a global knowledge of the infrastructure and choose either a predefined or the shortest route. It is questionable whether this simplification provides realistic results. This study addresses the problem of human-like route-choice behavior for microscopic pedestrian simulations. A route-choice model is presented that considers two concepts: first, the modeling of infrastructure knowledge to represent the variations in the decision-making processes of pedestrians with different degrees of familiarity with the infrastructure (e.g., regular commuters versus first-time visitors). Second, for each pedestrian the internal preference for selecting a certain path can be calibrated to allow the choice for the fastest routes or the ones that are most convenient for the agent (e.g., by avoiding stairs). The approach here uses a hybrid route-choice behavior model composed of a graph-based macrolevel representation of the environment, which is augmented with local information to avoid obstacles and dense crowds in the vicinity. This method was applied with different parameter sets in an evacuation study of a multilevel subway station. The results show the impact of these parameters on evacuation times, use of infrastructure elements, and crowd density at specific locations.

Real-Time Estimation of Pedestrian Inflow Rates from Saturated Sensor Counting Data in a Complex Metro Station

Many ITS applications in public transport stations benefit from knowing the pedestrian flows inside and outside the building for the purpose of predicting waiting times or applying safety measures. Counting people inside the station is often easier than outside the building because the building is a closed system with well-defined cross-sections like doors or stairs instead of the open area outside. The question is, given a set of people counters inside an infrastructure, how to determine the pedestrian inflow into the building. The inflow may be higher than the people counts inside the station because bottlenecks limit the maximum rate of people entering the infrastructure. We present a real-time estimation model which uses people counts inside a metro station and extrapolates the expected inflow once the counting sensors saturate. The model is validated with a case study at a large subway station next to Viennas largest football stadium.