Research interests and past projects

Measurement of congestion and risk in pedestrian crowds

Congestion is an important property of transportation systems and also a problem affecting millions of people on their daily lives. Congestion is typically seen in highways and rods when cars have to stop because of the high level of traffic preventing them from keep moving.

However, congestion is also seen in pedestrian crowds and people having to ride train during rush hours in busy urban centers are probably experiencing it on a daily basis. The problem is particularly known in rapidly developing Asian cities (China, Indonesia, India just to list few countries), but European and American cities are also affected by this problem.

Measuring congestion in pedestrian crowds is not a simple task and "congestion" is not a property which can be directly measured like time, distance or speed. In the frame of my research I developed a method which allows to assess the "smoothness" of pedestrian flows and thus determine the levels of congestion in crowds of people.

The method has been so far tested in several environments allowing to recognize lane formation in bidirectional streams, distinguish homogeneous (all people walk at the same speed) and heterogeneous (different speeds) crowds and determine the most risky area in evacuation.

The same method also allows to estimate how risky is the way people move. This can be useful during mass events or to monitor crowds in transportation hubs and help security personnel taking appropriate measures when dangerous levels of risk are reached.

Click on the image to view a related movie.

My research on this topic is contained in the following publications:

  • Claudio Feliciani and Katsuhiro Nishinari. "Measurement of congestion and intrinsic risk in pedestrian crowds." Transportation research part C: emerging technologies 91 (2018): 124-155 (link) (PDF).
  • Akihiro Fujita, Claudio Feliciani, Daichi Yanagisawa and Katsuhiro Nishinari. "Traffic flow in a crowd of pedestrians walking at different speeds ." Physical Review E 99.6 (2019): 062307 (link).
  • Claudio Feliciani and Katsuhiro Nishinari. "Investigation of pedestrian evacuation scenarios through congestion level and crowd danger." International Conference on Pedestrian and Evacuation Dynamics, Lund (Sweden), 2018 (in press).

Bidirectional pedestrian flow

The bidirectional flow is very commonly observed in pedestrian traffic. An example is the way crowds of people move in a corridor, a crosswalk or in sidewalks/walkways. In particular, bidirectional flows have been studied by researcher involved with pedestrian traffic because of the self-organization mechanism appearing when people organize themselves in lanes.

One of the most interesting and fascinating aspect in lane formation is that people are able to organize themselves without having a leader or some previous knowledge on the the way they should behave in group. Lanes simply appear as a collective process in which each person try to avoid collision with the counter-flow (you can see a video here of the lane formation process).

However, lanes do not always appear and sometimes people coming from both directions get stuck forming a dangerous dense crowd which has been the marginal cause for tragedies occurring in the past (like the Love Parade of Duisburg in 2010).

In particular, researchers have been trying to found out if bidirectional flows are more dangerous when the number of people from both directions is equal or when there is a large difference between them.

In my research I thoughtfully investigate bidirectional flows and the process of lane formation and was able to link the past literature on the subject.

We concluded that lane formation is more difficult when the number of people between both directions is equal because people need to consider both the crowd coming in the opposite direction and the people on their left and right side. However, lanes are more stable when groups from both direction are similar in number and thus tend to be more efficient on the long term. When there is a large difference between people number in both groups lanes are more difficult to form, but interactions are relatively limited thus resulting in small or no changes on the long term.

Click on the image to view a related movie.

My research on this topic is contained in the following publications:

  • Claudio Feliciani, Hisashi Murakami and Katsuhiro Nishinari. "A universal function for capacity of bidirectional pedestrian streams: Filling the gaps in the literature." PLoS one 13.12 (2018): e0208496 (link) (PDF).
  • Claudio Feliciani and Katsuhiro Nishinari. "Empirical analysis of the lane formation process in bidirectional pedestrian flow." Physical Review E 94.3 (2016): 032304 (link) (PDF).
  • Claudio Feliciani and Katsuhiro Nishinari. "Phenomenological description of deadlock formation in pedestrian bidirectional flow based on empirical observation." Journal of Statistical Mechanics: Theory and Experiment 2015.10 (2015): P10003 (link) (PDF).
  • Hisashi Murakami, Claudio Feliciani and Katsuhiro Nishinari. "Lévy walk process in self-organization of pedestrian crowds." Journal of the Royal Society Interface 16.153 (2019): 20180939 (link).

Simulation of high density crowds

Nowadays it is a common practice to use crowd simulation models when designing pedestrian infrastructures and planning mass events accommodating a large number of people. A considerable number of commercial software are available on the scope and many more are presented and developed every year.

However, most of the models used for pedestrian simulation are intended to recreate daily conditions and do not allow to consider very dense crowds. As a consequence, investigating crowd accidents or reproducing extreme conditions is not possible using conventional models.

Pedestrian densities are typically below 1 person/m² in most of public spaces. Above 2 persons/m² movements become difficult, but densities in the order of 6 persons/m² are routinely observed in packed trains during rush hours. Several researchers have reported that during crowds accidents densities were above 10 persons/m² (with some reporting up to 15 persons/m²).

Typical simulation models can only simulate crowds with a maximum density of about 5-6 persons/m². To allow simulating dense crowds I developed a simulation model which can be used to densities well above 10 persons/m² (the maximum density being 17 persons/m²), thus allowing to consider also extreme scenarios. The model has been validated using empirical data showing a good agreement in different conditions.

Click on the image to see a movie of a simulated bidirectional flow.

My research on this topic is contained in the following publications:

  • Claudio Feliciani and Katsuhiro Nishinari. "An improved Cellular Automata model to simulate the behavior of high density crowd and validation by experimental data." Physica A: Statistical Mechanics and its Applications 451 (2016): 135-148 (link) (PDF).
  • Claudio Feliciani and Katsuhiro Nishinari. "An Enhanced Cellular Automata Sub-mesh Model to Study High-Density Pedestrian Crowds." International Conference on Cellular Automata, Fes (Morocco), 2016 (link) (PDF).

Unsignalized crosswalks

Pedestrian and vehicular traffic is typically divided in areas accessible only by people and the ones accessible only by cars. People walk on the sidewalk and cars move in the road. Unsignalized crosswalks are one of the few places where cars and pedestrians interact. Unlike signalized crosswalks where traffic lights coordinate the motion of cars and people preventing collisions between them, in unsignalized crosswalks safety of pedestrians depend on the successful crossing negotiation occurring when people attempt to cross the road.

Accidents may easily occur because of misunderstandings or distractions and pedestrians are typically the victims. It is therefore important to understand the decision making process occurring when people attempt to cross the road and determine which variable are the most relevant in affecting volumes of traffic for both road users.

In the frame of this research (which has been carried out with the University of Milano-Bicocca) I used results from an on-field observation to develop a simulation model which reproduce the dynamics of pedestrians and drivers in unsignalized crosswalks.

The model has been validated using experimental data and we showed that it can be successfully employed to estimate waiting times caused by the presence of the crosswalk for both pedestrians and drivers. Such a model can be used, for example, to determine if a traffic light should be used to reduce waiting times or an unsignalized crosswalk is appropriate.

In a later study, I also investigated safety of pedestrians to determine which policy is more effective in reducing the gravity (and the number) of collisions between pedestrians and vehicles. Our research showed that while speed limit enforcement clearly help reducing pedestrian fatality, alternative "soft" solutions like the "shared space" may be also effective. We also showed how important is to address safety campaigns to drivers (while increasing pedestrians' awareness on risks).

In the current research, I am also investigating cognitive and environmental aspects related to unsignalized crosswalks using Virtual Reality and a self-developed driving simulator.

Click on the image above to see a movie of a simulated crosswalk.

My research on this topic is contained in the following publications:

  • Claudio Feliciani, Luca Crociani, Andrea Gorrini, Giuseppe Vizzari, Stefania Bandini and Katsuhiro Nishinari. "A simulation model for non-signalized pedestrian crosswalks based on evidence from on field observation." Intelligenza Artificiale 11.2 (2017): 117-138 (link) (PDF).
  • Claudio Feliciani, Andrea Gorrini, Luca Crociani, Giuseppe Vizzari, Katsuhiro Nishinari and Stefania Bandini. "Calibration and validation of a simulation model for predicting pedestrian fatalities at unsignalized crosswalks by means of statistical traffic data." Journal of Traffic and Transportation Engineering (English Edition) (in press).
  • Claudio Feliciani, Luca Crociani, Andrea Gorrini, Giuseppe Vizzari, Katsuhiro Nishinari and Stefania Bandini. "Assessment of Pedestrian Fatality Risk at Unsignalized Crosswalks by Means of Simulation." Conference on Traffic and Granular Flow, Washington D.C. (USA), 2017 (in press).
  • Stefania Bandini, Luca Crociani, Claudio Feliciani, Andrea Gorrini and Giuseppe Vizzari. "Collision Avoidance Dynamics Among Heterogeneous Agents: The Case of Pedestrian/Vehicle Interactions." Conference of the Italian Association for Artificial Intelligence, Bari (Italy), 2017 (link) (PDF).

Measurement of body orientation and pedestrian properties through commercial devices

Typically, pedestrians are detected using cameras or special sensors (recently distance sensors are increasingly used) which while being relatively expensive also partially violate people's privacy. Although solutions not making use of privacy-sensitive data are being developed, pedestrians spaces are more friendly when they lack of a large number of cameras and sensors.

Nowadays, most of the people walking in public spaces carry a smartphone or a tablet. These kind of devices contain a large number of sensors and are constantly connected to a communication network.

GPS is sometimes used to obtain people's position and estimate their speed, but this is also partially involving privacy-sensitive information. A more efficient and user-friendly solution is to make use of inertial sensors contained in electronic devices and estimate how fast people move by analyzing the motion of their body.

In the frame of this research, I assessed the feasibility of the use of inertial sensors from commercial electronic devices to estimate speed and density of pedestrian crowds. Our research showed that under controlled conditions both speed and density can be estimated with a sufficient accuracy. However, our results also showed that a large scale application may require several improvements and would need a relatively large number of active users to guarantee reliable results.

The software platform developed to gain inertial data from multiple devices on real-time was also used by colleagues to measure body orientation making use of the gyroscope sensor.

Click on the image above to see a movie of the experiment.

My research on this topic is contained in the following publications:

  • Claudio Feliciani and Katsuhiro Nishinari. "Pedestrians rotation measurement in bidirectional streams." International Conference on Pedestrian and Evacuation Dynamics, Hefei (China), 2016 (link) (PDF).
  • Claudio Feliciani and Katsuhiro Nishinari. "Estimation of pedestrian crowds' properties using commercial tablets and smartphones." Transportmetrica B: Transport Dynamics 7.1 (2019): 865-896 (link) (PDF).
  • Hiroki Yamamoto, Daichi Yanagisawa, Claudio Feliciani and Katsuhiro Nishinari. "Body-rotation behavior of pedestrians for collision avoidance in passing and cross flow." Transportation Research Part B: Methodological 122 (2019): 486-510 (link) (PDF).
  • Daichi Yanagisawa, Feliciani, Claudio and Katsuhiro Nishinari. "Unidirectional and bidirectional flow in a narrow corridor with body rotation. " International Conference on Pedestrian and Evacuation Dynamics, Lund (Sweden), 2018 (in press).

Behavior and recognition of dyads (pairs) in pedestrian crowds

All crowds are composed by people, but relationship between their members are an important aspect determining the overall behavior. Individuals behave in a different ways from couples and couples behave differently from families. Understanding how groups of people behave in different situations and what are the differences from individual behavior is necessary to improve models used in simulation.

In the frame of my research I collaborated with different institutions (University of Milano-Bicocca, Okayama University, ATR Kyoto and Kyoto University) to study in particular the behavior of dyads (groups composed by 2 persons).

Although my role in this topic has been mostly marginal and limited to assist the execution of experiments and/or provide advice on analytical methods, the different collaborations allowed me to learn and gain knowledge on group behavior and methods employed to analyze/simulate their behavior.

Click on the image to see a movie of the experiment.

My research on this topic is contained in the following publications:

  • Zeynep Yucel, Francesco Zanlungo, Claudio Feliciani, Adrien Gregorj and Takayuki Kanda. "Estimating social relation from trajectories. " International Conference on Pedestrian and Evacuation Dynamics, Lund (Sweden), 2018 (in press) (PDF).
  • Andrea Gorrini, Luca Crociani, Claudio Feliciani, Pengfei Zhao, Katsuhiro Nishinari and Stefania Bandini. "Social groups and pedestrian crowds: experiment on dyads in a counter flow scenario." International Conference on Pedestrian and Evacuation Dynamics, Hefei (China), 2016 (link) (PDF).
  • Luca Crociani, Andrea Gorrini, Claudio Feliciani, Giuseppe Vizzari, Katsuhiro Nishinari and Stefania Bandini. "Micro and Macro Pedestrian Dynamics in Counterflow: the Impact of Social Groups", Conference on Traffic and Granular Flow, Washington D.C. (USA), 2017 (in press).

Previous (past) research topics

Before starting my research on pedestrians crowds I have been involved in several research projects from different disciplines mostly related with fluid-dynamics.

As an undergraduate student I investigated the creation of turbulence in a jet-stream by performing laboratory experiments. High-speed jet streams were investigated using the PIV (Particle Image Velocimetry) to determine if such a technique could be employed to study turbulence.

As a graduate student in nuclear engineering I developed a device to mix gases and allow the measurement of radioactive Xe using gas mass microscopy. The device allowed to dilute radioactive gas with a high concentration of Xe with a precisely known dilution ratio and later allow its analysis using an highly sensitive special equipment.

While working as a R&D engineer in the frame of polymer science I developed a method to compute thermal conductivity of fiber-reinforced polymer using simulation. A patent has been submitted in relation with the proposed method. During my experience as an engineer I had the change to use different commercial software used in fluid dynamics (here is a movie of falling ice simulated using Flow3D).

During my education and research I also had the opportunity to work with several codes related with fluid-dynamics but ranging from various disciplines including reactive polymers and tsunami propagation (see a movie here).

While working for a small company in Switzerland (MAIN Gmbh) I also developed a software to simulate the dynamics of a twin roll casting plant and estimate the ROI (Return of Investment) depending on several variables such as electricity and gas cost and invested capital (screenshot).

My research on this topic is contained in the following publications:

  • Claudio Feliciani and Yoshihiro Takai. "Measurement and numerical prediction of fiber‐reinforced thermoplastics' thermal conductivity in injection molded parts." Journal of Applied Polymer Science 131.2 (2014) (link).
  • Claudio Feliciani, Yoshihiro Takai and Seiki Hiramatsu. "A method for predicting thermal conductivity in fiber-reinforced thermoplastic parts produced by injection molding.", 24th Annual Meeting of the Japan Society of Polymer Processing, Tokyo (Japan), 2013 (PDF).
  • Claudio Feliciani, Niko Kivel, Judith Kobler Waldis, Beat Wernli and Ines Gunther-Leopold. “Development of an isotope dilution technique for the quantitative analysis of fission gases in nuclear fuels”, 22. ICP-MS Anwendertreffen, Berlin (Germany), 2010 (PDF).
  • Vipluv Aga, Claudio Feliciani, Ndaona Chokani and Reza Abhari. "Turbulence Measurements of a High Reynolds Number Inclined Jet in Crossflow using PIV and FRAP", 60th Annual Meeting of the Division of Fluid Dynamics, American Phys. Soc., Salt Lake City (USA), 2007 (link) (PDF).