Difficulty Quickly and Accurately Analyzing What Causes Events and Airports to Be Crowded
Many people probably have felt tired from being in massive crowds at street events, malls on weekends, and airports. Event organizers and people in charge of malls take various countermeasures such as limiting entrance and guiding visitors with signboards, but these are, however, not too effective in relieving overcrowding. As such, visitors may not enjoy events despite coming all the way to one or spend hours shopping, resulting in a drop in visitor satisfaction.
Current research on approaches to relieving overcrowding involves building agent models that represent the attributes and behaviors of various visitors to events and malls, including age, gender, and purpose of visit, and creating crowd situations virtually. A wide variety of crowd situations are created by making thousands of such agents pick routes to their destinations while following signboards and referring to information on crowded areas. This research is known as "human behavior simulation."
In human behavior simulation, experts iterate the procedure of analyzing a large amount of data to identify the cause of overcrowding, forming a hypothesis for devising measures to prevent overcrowding, and evaluating it by the simulation. This approach causes problems such as requiring months before measures can be finalized and a tendency to fail to find effective measures due to causes being overlooked.
To solve such problems, Fujitsu has collaborated with the Faculty of Science and Engineering, Waseda University to develop a technology to automatically analyze the characteristics of agents related to overcrowding based on the results of human behavior simulations.
Automated Analysis of the Characteristics of Thousands of Agents
The most prominent characteristic of this technology is that it can analyze thousands of agents automatically. The analysis of agents must cover all the combinations of agent characteristics, including a number of attributes such as age, gender, and purpose of visit, as well as behavioral patterns, such as having a meal and looking at a signboard. Traditional analysis involves every combination of such characteristics, which resulted in an enormous number of combinations, making analysis time-consuming and prone to causes being overlooked.
On the other hand, the new technology focuses on what agents have in common, such as attributes, perceptions, and behaviors. Agents are characterized afresh through the process of grouping their characteristics according to common elements, clustering each group, and identifying which cluster each agent belongs to in each group. This helps reduce the number of combinations while maintaining useful information of data. Characterizing agents with fewer combinations helps identify what perceptions and behaviors relating to specific attributes are useful for relieving overcrowding when they are changed.
Figure 1: Characterizing agents with a small number of combinations by forming groups in terms of attributes, perceptions, and behaviors
Number of People Lining for Security Screenings Reduced to One-Sixth in an Airport Simulation
Fujitsu analysed a human behavior simulation simulating an actual airport based on behavioral data on people surveyed at airports in order to evaluate the effectiveness of the new technology. Regarding long security lines, the technology helped us identify four times as many causes as an expert did.
In addition, it revealed that the sudden growth of security lines is caused by passengers being kept at specific check-in counters. When schedules for check-in counters and security screenings were redesigned based on identified causes of overcrowding, it was confirmed that this redesign resulted in reductions in the number of people lining for security screenings to one-sixth and in personnel required for implementing measures to two-thirds, compared to schedules designed based on an expert's analysis. This technology also helped reduce analysis time to only a few minutes, compared to a period of months that a traditional method requires.
Figure 2: Airport simulation revealed various benefits
Toward a more convenient city environment and more comfortable life
In Japan, a country with a small land area and a population increasingly concentrating in Tokyo and other metropolitan areas, relieving crowding is an important subject. These results indicate that events, malls, airports, and other facilities may be less crowded and various problems, including the economic impact of disaster prevention and traffic controls, may be solved in the future.
Fujitsu and will move forward to conduct field trials using this technology, applying it to congestion at locations like event spaces, airports, and shopping malls, and evaluate its effectiveness, including with such measures as digital signage and the placement of commercial tenants. Fujitsu will also leverage its Human Centric AI Zinrai artificial intelligence and supercomputer technologies, and together with its Fujitsu Technical Computing Solution GREENAGES Citywide Surveillance software, which enables a real-time understanding of urban conditions, will aim to quickly realize a future predictive solution for congestion.