BLOG: "Smart airports need a real-time and privacy-friendly system that identifies, analyses, and forecasts crowd densities to control passenger flows"

Systems capable of detecting critical situations in which dangerously large amounts of people are present allow to take measures, thereby preventing potentially lethal accidents or very long waiting queues.

    Blog airport passenger safety

    Currently, emergency control centers like an Airport Operations Centre (APOC) receive and process a wide array of input feeds (e.g. camera images, access-gate counters, consumption sales, Wi-Fi access point data, professional opinions by police officers in the field etc). However, many of these existing approaches suffer from major accuracy, cost- and privacy related issues.

    Furthermore, the variety of input data can often lead to disagreements between the different stakeholders because of conflicting interpretations based on their varying domain-specific expertise. As a result, there exists a need for a real-time system which identifies, analyses, and forecasts crowd densities to control passenger flows to ensure that the total occupancy remains below the capacity limits.

    And that's where the CrowdScan solution comes in.

      Privacy non-intrusive capacity sensing

      The proposed approach does not make use of camera images and operates entirely in a device-free manner in which the influence of the physical presence of human individuals on radio frequency signals in the environment is used to derive crowd size information. Using this methodology, it is impossible to determine the identity of individual crowd members, which therefore results in a system that is inherently privacy conscious. The basic principle is as follows: a set of sensors are mounted at the location of interest at an average waist height. These sensors are transmitting radio waves in different directions through the crowd.

      The key principle of this methodology relies on the impact of the passengers on the radio waves when compared to an unoccupied environment. The relative disturbance of the signal is used as input to an analytical model that is able to accurately measure the number of people present within the environment. In addition to this, the methodology allows us to divide the environment into different subareas. To identify the motion and activity of the passengers, the flow and flux between areas can be analyzed.

      This analysis will add the labels low, medium, high for the flow and increasing, decreasing, stabilizing, or steady for the flux. This enables operators to track passenger flows and advise security personnel in the field.

      This makes the system highly unique and useful for many different applications in which access to capacity and flow estimation data is necessary. Examples include border and security control.

      Compared to existing solutions, this technology has four unique differentiators:

      1. Accuracy

      Based on experiments performed at different large-scale events, this technology was able to measure the full crowd with a median error of 5% compared to an access control system.

      2. Privacy

      Because the disturbance of radio signals is used and not a connection with any device, this technology is privacy by design and thus no potential GDPR issues. Furthermore, there is no possibility to track any individual in large and dense crowd since we cannot identify any.

      3. Real-time

      Because of the fast synchronization (every 50ms), this approach processes the raw data to meaningful data that is aggregated every 30 seconds. This increases the robustness and the stability of the system so that a control room can use this in real-time.

      4. Technology independent

      Because of the use of a sub-GHz wireless communication standard and the device free system, the technology is not dependent on any other technology like Bluetooth or Wi-Fi to count individuals. This ensures the functionality in crisis situations.

        Passenger throughput vs crowd density

        As part of an APOC, different parameters have to be brought together to manage passenger flows. A real-time risk assessment framework has to be in place to enable real-time passenger management so that full control of the environment is maintained. This is necessary to allocate security personnel to the correct areas of interest to both optimize and increase the predictability of passenger flows. Moreover, real-time risk assessment enables operators to warn and alert security personnel in time, with a positive effect on both safety and security in reference to overall customer well-being.

        As an example, throughput time at border control is a function of waiting time (time in queue) and service time (time to handle a passenger). Being a major factor in customer satisfaction, passenger throughput time is constantly monitored in order to identify potential bottlenecks and reduce variance in waiting time.

        Airports often monitor throughput time of individual passengers using Bluetooth. This method is accurate, but mainly detects changes in throughput time instead of predicting such upcoming variances. An increase in throughput time can be the result of an increase in waiting time or service time or a combination of both. This makes it difficult to take effective and corresponding measures in time. Adding the technology in airports will allow for real-time monitoring of the number of passengers waiting. Combining this data with the actual throughput time will allow for proactive and more cost efficient measures. If the arrival rate of passengers increases for any reason, the number of people in the waiting area will also increase. This will immediately be detected by the wireless sensor network which alerts the APOC before the throughput time increases. This will allow for a timely response. On the other hand, if the number of passengers waiting remain stable and throughput time increases, this will be an indication that rather service time is affected so this bottleneck can be addressed.

          Passive crowd density estimation in COVID-19

          Due to the global COVID-19 pandemic, different capacity limits have been mandated by the government for indoor and outdoor public areas and venues to reduce the number of infections. As a result, it is important for an APOC to have objective crowd estimates that can be used by operators to manage active flows.

          Based on the applied proposed methodology, a real-time dashboard is created that will show the evolution of all measurements live. Moreover, such a dashboard will trigger an alert when an anomaly is detected. The definition of an anomaly can be very broad. In the context of this application it is the detection of a real-time crowd estimation value which exceeds a certain predefined threshold value. This value can be configured during installation and updated at any moment in time. This dynamic approach allows operators to set triggers more conservatively so that they get alerted more in advance when something will happen in the near future.

            Call to action

            Airport operations need to be monitored and managed with no negative impact on end-to-end passenger flows. To do so, processes must be optimized with a minimum amount of resources and a maximum throughput. In the current setting, the passenger flow will be determined by uncertain passenger behaviour. This passenger conduct and actions will influence crowd density at certain critical locations at the airport. Both from a health and safety perspective and from an efficiency point of view, queue management needs to be extended to the broader domain of crowd management. The implementation of automatic accurate and real-time crowd estimation systems would provide APOC’s with the means to react in real-time with ‘crowd alerts’. This would increase the safety and will indirectly enhance passenger experience in a privacy conscious manner.

              Written by Ben Bellekens, CEO and co-founder, CrowdScan and Bart Seuntjens, General Manager, Brussels Airport Consulting

              Source: ACI (September, 16, 2020)