Passenger Counting Research Articles

UAI-FI: using artificial intelligence for automatic passenger counting through Wi-Fi and GPS data

An important piece of information for planning public transportation is the number of passengers using the system. Several initiatives have started to explore the Wi-Fi packets generated by passengers’ smartphones as means to obtain this information. A sensing device located inside the bus can intercept and collect these packets. By applying filters, e.g., verifying if the signal strength is higher than a threshold, the sensor can infer passengers' presence/absence. However, such limits are set arbitrarily, leading to errors, for example, when close to bus stops. To address this issue, this article proposes a method (UAI-FI) based on an artificial intelligence technique (Support Vector Machine) to classify the origin of packets as inside or outside the bus. To validate UAI-FI, we applied and compared our approach to other methods in a bus line in Goiânia/Brazil. The results suggest that UAI-FI outperformed existing methods. Furthermore, it successfully classified the packet’s origin, obtaining 83.3% and 88.5% of the total number of passengers boarding and alighting the line. Despite the overall similarity, we highlight that UAI-FI’s counting curve presented a delay compared to the manual count indicating that the frequency that Wi-Fi packets are sent can cause the presence/absence of passengers to be perceived at different stops.

ECC: Passenger Counting in the Elevator Using Commodity WiFi

Elevators have become a kind of indispensable facility for everyday life, which bring people both convenience and safety hazards. Specifically in the household environment, an elevator’s lifespan is expected to be more than 20 years. An appropriate and regularly maintained counterweight is conducive to extending elevator life. This paper proposes a passenger counting approach in the elevator for regular counterweight adjustment based on commodity WiFi called ECC. Since the running time of the elevator between two adjacent floors is short, the major challenge of ECC is how to count passengers from the limited captured data. This paper first theoretically analyzes the relationship between the number of passengers and the variation of channel state information (CSI). Then ECC constructs a multi-dimensional feature by extracting the average of amplitude (AOA), time-varying spectrum (TVS), and percentage of non-zero elements (PEM) features from the limited data. Finally, the random forest (RF) classifier is used for passenger counting and the local optimization problem is solved by expanding the feature dataset through data segmentation. ECC is implemented by using off-the-shelf IEEE 802.11n devices, and its performance is evaluated via extensive experiments in typical real-world scenes. The estimated precision of ECC can reach more than 95%, and more than 97% of estimation errors are less than 2 persons, which demonstrates the superior effectiveness and generalizability of ECC.

Engineering the Neural Automatic Passenger Counter

Automatic passenger counting (APC) in public transportation has been approached with various machine learning and artificial intelligence methods since its introduction in the 1970s. It is mainly used for revenue sharing, which (in Germany alone) is in the billions annually and supply planning, which is essential for services of general interest. While equivalence testing is becoming more popular than difference detection (Student’s t-test), the former is much more difficult to pass to ensure low user risk. On the other hand, recent developments in artificial intelligence have led to algorithms that promise much higher counting quality (lower bias). However, gradient-based methods (including Deep Learning) typically run into local optima. In this work, we explore and exploit various aspects of machine learning to increase the reliability, performance, and counting quality of the Neural APC 3D depth video-based LSTM neural network. We perform a grid search with several fundamental parameters: the selection and size of the training set, which is similar to cross-validation, and the initial network weights and randomness during the training process. Using this experiment, we show how aggregation techniques such as ensemble quantiles can reduce bias, and we give an idea of the overall spread of the results. We utilize the test success chance, a simulative metric based on the empirical distribution. We also employ a post-training Monte Carlo quantization approach and introduce cumulative summation to turn counting into a stationary method and allow unbounded counts. All in all, our numerous additions provide a major quality increase to the NAPC.

JP-DAP: An Intelligent Data Analytics Platform for Metro Rail Transport Systems

This paper deals with an intelligent data analytics platform - Jaison-Paul Data Analytics Platform (JP-DAP) - for metro rail transport systems. JP-DAP is intended to ensure smooth functioning, improved customer experience, ridership forecasting, and efficient administration of metro rail transportation systems by integrating and analysing its many data sources. It consists of a middleware which is built on the top of a Hadoop Distributed File System (HDFS) and Spark framework, along with a set of open-source software tools like Apache Hive, Pandas, Google TensorFlow and Spark ML-lib for real-time and legacy data processing. The benchmarking of JP-DAP was conducted using TestDFSIO and have found that it performs well according to industry standards. The specific use case for this project is Kochi Metro Rail Limited (KMRL). The analysis of Automated Fare Collection data from KMRL on JP-DAP framework have produced descriptive statistics visualisation of inflow and outflow analysis, travel patterns during weekdays and weekends, origin-destination matrix, etc.. Moreover JP-DAP framework is capable of producing short term passenger flow predictions using SVR machine learning algorithm with linear, radial basis function and polynomial kernels. Our experiments have shown that SVR linear kernel gives the most accurate results with the least errors in predicting the next day’s passenger count using the previous five weekdays data. The station usage (one-to-all) prediction using Long Short-Term Memory (LSTM) is also integrated to this framework. The visualisation as well as analytical outcomes of JP-DAP framework have also been made available to the external world using a rich set of REST APIs and are projected on to a web-dashboard.

Passenger Volumes and Dwell Times for Commuter Trains: A Case Study Using Automatic Passenger Count Data in Stockholm

This study investigates the relationship between passenger volumes and the frequency and size of dwell times for commuter trains. We use data collected by automatic passenger count systems in Stockholm and show that dwell times often take longer than scheduled. To understand the effect of passenger volumes on delays, we use the frequencies of delays for different passenger volumes and delay sizes. We find that these frequencies differ significantly from a frequency distribution independent of passenger volumes, indicating that passenger volumes have an effect on the frequency of dwell time delays. Neglecting passenger volumes underestimates the frequency of delays in most cases, especially for smaller delays. Although the frequency of dwell time delays increases as passenger volumes increase, the same is not necessarily true for their size. The relationship between passenger volumes and delay sizes is thus non-linear, and it depends on scheduled dwell times. We conclude that small increases in dwell times can result in sharp increases in on-time performance but that only increasing dwell times is not sufficient. Measures to speed up the boarding and alighting process are also necessary.

Frequency-based path flow estimator for transit origin-destination trip matrices incorporating automatic passenger count and automatic fare collection data

A frequency-based path flow estimator is proposed to estimate the transit passenger origin–destination (OD) matrix in an urban congested transit network. The proposed model not only considers the effect of congestion under an equilibrium framework, but also benefits from being formulated as a single-level model with the route-section-based network representation. Multiple transit data sources are incorporated, including automatic passenger count, automatic fare collection and automatic vehicle location data. A path-based diagonalization approach embedded with an iterative balancing scheme is developed to solve the model. Case studies are conducted to demonstrate the features and applicability of the proposed model and algorithm.

ИССЛЕДОВАНИЕ НЕРАВНОМЕРНОСТИ ПАССАЖИРОПОТОКА ВОКЗАЛА «ИРКУТСК-ПАССАЖИРСКИЙ»

The results of a study of the non-uniformity of the passenger traffic at the station during the day based on the passenger counting system are presented

Transit Quality of Service Assessment Using Smart Data

In this paper we assess the transit quality of service (QoS) from a user’s standpoint, using smart data. A number of bus lines with different characteristics, operating in the Metropolitan Area of Athens, were chosen as a case study. The data used were gathered by an Automatic Passenger Counting (APC) system. APC technologies provide exact temporalized passenger counting along the line for each service, thus assisting to better understand causalities of delays and avoid operational problems. By employing archived APC data from buses running on crosstown routes between 15 January 2019 and 15 April 2019 we conducted a statistical analysis to explore occupancies and assess QoS, including under a social distancing scenario. The passenger distribution along the stops, the bus’s occupancy level, the stops that are maximum occupancy points and their rate of occurrence and, lastly, the passenger’s average trip length during the day and the week are examined.

Cross-checking automated passenger counts for ridership analysis

Cross-checking automated passenger counts for ridership analysis

Passenger flow prediction in bus transportation system using deep learning.

The forecasting of bus passenger flow is important to the bus transit system’s operation. Because of the complicated structure of the bus operation system, it’s difficult to explain how passengers travel along different routes. Due to the huge number of passengers at the bus stop, bus delays, and irregularity, people are experiencing difficulties of using buses nowadays. It is important to determine the passenger flow in each station, and the transportation department may utilize this information to schedule buses for each region. In Our proposed system we are using an approach called the deep learning method with long short-term memory, recurrent neural network, and greedy layer-wise algorithm are used to predict the Karnataka State Road Transport Corporation (KSRTC) passenger flow. In the dataset, some of the parameters are considered for prediction are bus id, bus type, source, destination, passenger count, slot number, and revenue These parameters are processed in a greedy layer-wise algorithm to make it has cluster data into regions after cluster data move to the long short-term memory model to remove redundant data in the obtained data and recurrent neural network it gives the prediction result based on the iteration factors of the data. These algorithms are more accurate in predicting bus passengers. This technique handles the problem of passenger flow forecasting in Karnataka State Road Transport Corporation Bus Rapid Transit (KSRTCBRT) transportation, and the framework provides resource planning and revenue estimation predictions for the KSRTCBRT.

NAPC: A Neural Algorithm for Automated Passenger Counting in Public Transport on a Privacy-Friendly Dataset

Real-time load information in public transport is of high importance for both passengers and service providers. Neural algorithms have shown a high performance on various object counting tasks and play a continually growing methodological role in developing automated passenger counting systems. However, the publication of public-space video footage is often contradicted by legal and ethical considerations to protect the passengers’ privacy. This work proposes an end-to-end Long Short-Term Memory network with a problem-adapted cost function that learned to count boarding and alighting passengers on a publicly available, comprehensive dataset of approx.13,000 manually annotated low-resolution 3D LiDAR video recordings (depth information only) from the doorways of a regional train. These depth recordings do not allow the identification of single individuals. For each door opening phase, the trained models predict the correct passenger count (ranging from 0 to 67) in approx.96% of boarding and alighting, respectively. Repeated training with different training and validation sets confirms the independence of this result from a specific test set.

Simple pre-post analysis overestimates the impacts of new public transit services on ridership: Evidence from a quasi-experimental study of new bus rapid transit in Columbus, Ohio, USA

This paper empirically demonstrates the value of quasi-experimental study designs to evaluate the direct impacts of new public transit services on ridership within its corridor. Using a new bus rapid transit (BRT) service, CMAX, in Columbus, Ohio, USA, as an example, we compare its impact on ridership based on a pre-post and quasi-experimental analysis framework. We conduct the pre-post analysis using a ridership space-time cube exploring a massive Automatic Passenger Counter (APC) database. Differences in total passenger counts before and after the BRT intervention indicate a 36% increase in ridership within its corridor. However, this patronage increase may not be attributable solely to the new public transit service. Potential confounding effects include systemwide ridership trends and a new unlimited transit pass program for downtown workers. To address these issues, we adopt a quasi-experimental study design with a difference-in-differences (DiD) identification strategy. We use propensity score matching (PSM) to match a counterfactual control group with the treatment group when implementing DiD model. After accounting for confounding effects, we find a less than 5% increase but not statistically significant impacts of CMAX on ridership. Our results support the argument that a simple pre-post analysis ignoring confounding effects can lead to a misleading evaluation of a new public transit service’s direct impact on ridership.

Methodology for public transport mode detection using telecom big data sets: case study in Croatia

Determining the number of passengers using public transport services is a challenging and time-consuming task that relies either on manual observations (e.g. manual counting passengers in vehicles or at stations) or the application of technical solutions (using data from automatic fare collection system (AFC) or automatic passenger counters (APC), which is characterized either by the provision of an incomplete picture (AFC) or by a solution which in practice is installed in a small number of vehicles if any (APC). The new approach which uses anonymized telecom-originated big data sets and data science principles can be used as a smart data driven approach for determining the use of public transport. Anonymized telecom big data sets represent “digital breadcrumbs” that people leave while moving through the city. When paired with additional data sets (e.g. public transport timetables, location of public transport stations, information on public transport lines, etc.), it can be used for modal split detection. In this paper, a new methodological approach is proposed that uses anonymized telecom big data sets and a statistical modelling approach to identify possible public transport trips among all other trips. This methodology has been tested in a case study in the City of Rijeka and validated using ground truth data obtained from traditional sources.

Noninvasive Passenger Detection Comparison Using Thermal Imager and IP Cameras

Many modern vehicles today are equipped with an on-board e-call system that can send information about the number of passengers in the event of an accident. However, in case of fire or other major danger in a road tunnel, it is very important for rescue services to know not only the number of passengers in a given vehicle that has an accident and called help via e-call but how many people are in the tunnel in total. This paper deals with the issue of passenger detection and counting using the TPH3008-S Thermal camera and the VIVOTEK IP7361 IP Cameras noninvasively, i.e., the cameras are placed outside the vehicle. These cameras have their limitations; therefore, we investigated how to improve conditions and how to make detection better for future work. The main goal of this article is to summarize the achieved results and possibilities of improvement of the proposed system by adding other sensors and systems that would improve the final score of passenger detection. The experimental results demonstrate that our approach has to be modified and we have to add additional sensors or change methods to achieve more promising results. The results, findings and conclusions might be later used in tunnels and highways and also be applied in telematics and lead to better, safer road transport and improvement of existing tunnel systems sustainability by utilizing resources in a smarter way.

Clonal Hematopoiesis Is Driven By Aberrant Activation of TCL1A

Introduction: Clonal hematopoiesis of indeterminate potential (CHIP) may occur when a hematopoietic stem cell (HSC) acquires a fitness-increasing mutation resulting in its clonal expansion. A diverse set of driver genes, such as regulators of DNA methylation, splicing, and chromatin remodeling, have been associated with CHIP, but it remains largely unknown why HSCs bearing these mutations are positively selected. It has been challenging to identify the genetic and environmental factors mediating clonal expansion in humans, partially due to a lack of large cohorts with longitudinal blood sampling of participants. To circumvent this limitation, we developed a method to infer clonal expansion rate from single timepoint data called PACER (passenger-approximated clonal expansion rate).Methods: PACER is based on the principle that genomic passenger mutations can be used to infer the birth date of pre-malignant clones because these mutations accumulate fairly linearly over time. Thus, an individual with CHIP with a greater number of passenger mutations in the mutant clone is expected to have acquired the clone at a later age than someone with fewer passenger mutations. For two individuals of the same age and with clones of the same size, we expect the clone with more passengers to be more fit, as it expanded to the same size in less time. Typically, one would need to isolate single-cell colonies derived from HSCs in order to calculate the total passenger mutation burden. However, we hypothesized that this measure could also be approximated from whole genome sequencing of blood cell DNA, such as that used in large biobank projects. The expansion rate (PACER) is then estimated by adjusting the total passenger count for age and variant allele fraction in each individual. The ability of passengers to predict future clonal expansion was validated using longitudinal blood samples from 51 CHIP carriers in the Women's Health Initiative taken ~10 years apart (Figure 1). It also accurately predicted the known fitness effects due to different driver mutations in 5,551 CHIP carriers from the Trans-Omics for Precision Medicine (TOPMed) program (Figure 2).Results: Having validated the approach, we next hypothesized that we could identify germline variants influencing PACER, thus revealing genes and pathways mediating clonal expansion. The lead hit in a genome-wide association study (GWAS) of PACER was a common single nucleotide polymorphism (SNP) in the TCL1A promoter that was associated with slower clonal expansion in CHIP overall (Figure 3). TCL1A is an oncogene that is activated via translocation in T-cell prolymphocytic leukemia, but has no known role in CHIP or myeloid malignancies. A gene-level analysis indicated that the TCL1A SNP was associated with slower growth of clones bearing TET2 mutations, but had no effect on DNMT3A-mutant clone growth. We further found that those carrying two copies of the protective SNP had 40-80% reduced odds of having clones with driver mutations in TET2, ASXL1, SF3B1, SRSF2, and JAK2, but not DNMT3A. A concomitant decrease in incident myeloid malignancies was also seen in carriers of this protective SNP.Next, we interrogated how the protective SNP influenced TCL1A activity in HSCs. Normal human HSCs lacked open chromatin at the TCL1A promoter and TCL1A expression, but inducing frameshift mutations in TET2 via CRISPR editing led to accessibility of the promoter and gene/protein expression in HSCs (Figure 4). This effect was abrogated in HSCs from donors of the protective TCL1A SNP in a dose-dependent manner. Finally, we found that HSCs from donors homozygous for the protective SNP had markedly less expansion of phenotypic stem and progenitor cells in vitro after the introduction of TET2 mutations than TET2-edited HSCs from donors with two copies of the reference allele.Conclusions: In summary, we developed a novel method to infer the expansion rate of pre-malignant clones and performed the first ever GWAS for this trait. Our results indicate that the fitness advantage of several common driver genes in CHIP and hematological cancers is mediated through TCL1A activation, which may be a therapeutic target to treat these conditions. PACER is an approach that can be widely adopted to uncover genetic and environmental determinants of pre-malignant clonal expansion in blood and other tissues. [Display omitted] DisclosuresDesai: Bristol Myers Squibb: Consultancy; Kura Oncology: Consultancy; Agios: Consultancy; Astex: Research Funding; Takeda: Consultancy; Janssen R&D: Research Funding. Natarajan: Blackstone Life Sciences: Consultancy; Boston Scientific: Research Funding; Novartis: Consultancy, Research Funding; AstraZeneca: Consultancy, Research Funding; Apple: Consultancy, Research Funding; Amgen: Research Funding; Genentech: Consultancy; Foresite Labs: Consultancy. Jaiswal: Novartis: Consultancy, Honoraria; AVRO Bio: Consultancy, Honoraria; Genentech: Consultancy, Honoraria; Foresite Labs: Consultancy; Caylo: Current holder of stock options in a privately-held company.

Transit pollution exposure monitoring using low-cost wearable sensors

• Wearable low-cost sensors are a viable solution for transport pollution monitoring. • Novel wearable sensor for monitoring personal exposure during commuting. • Our sensor is accurate (R = 0.78) and consistent (R > 0.98). • Our sensor captures variations caused by route, passenger count and sensor location. • Low-cost sensors support new applications and multiple stakeholders. Transit activities are a significant contributor to a person’s daily exposure to pollutants. Currently obtaining accurate information about the personal exposure of a commuter is challenging as existing solutions either have a coarse monitoring resolution that omits subtle variations in pollutant concentrations or are laborious and costly to use. We contribute by systematically analysing the feasibility of using wearable low-cost pollution sensors for capturing the total exposure of commuters. Through extensive experiments carried out in the Helsinki metropolitan region, we demonstrate that low-cost sensors can capture the overall exposure with sufficient accuracy, while at the same time providing insights into variations within transport modalities. We also demonstrate that wearable sensors can capture subtle variations caused by differing routes, passenger density, location within a carriage, and other factors. For example, we demonstrate that location within the vehicle carriage can result in up to 25 % increase in daily pollution exposure – a significant difference that existing solutions are unable to capture. Finally, we highlight the practical benefits of low-cost sensors as a pollution monitoring solution by introducing applications that are enabled by low-cost wearable sensors.

Impact of COVID-19 on Waiting Passenger Distribution on a Bus Rapid Transit Station Platform in Brisbane, Australia.

The operation of a bus rapid transit (BRT) station platform is a key factor that affects BRT system performance. As waiting passengers occupy more platform space than circulating passengers, evaluation of their distribution across the platform is important. Public transport systems have been affected by the global pandemic Coronavirus disease 2019 (COVID-19). This may have affected the waiting passenger distribution on BRT platform space. Therefore, this study aimed to identify the impact of COVID-19 on waiting passenger distribution on a platform during the peak period at an important station on the BRT system in Brisbane, Australia. Manual data collection was carried out before COVID-19 and during COVID-19. Waiting passenger counts in each case were evaluated separately to identify variation across the platform. The total waiting passenger count on the platform at a given time reduced significantly during COVID-19. To compare the two cases, data sets were normalized, and statistical analysis performed. The test results indicated that the distribution of waiting passengers during COVID-19 has significantly changed, bringing more waiting passengers into the platform center than the two ends, whereas before COVID-19, more waiting passengers were observed at the upstream half of the platform. There was also greater temporal variation across the whole platform during COVID-19. These findings were used to postulate the reasons behind the changes resulting from COVID-19, which affected platform operation.

Updating National Air Passenger Demand from Traffic Counts: The Case of a Secondary Airport in an Underdeveloped Region

Demand models allow to estimate the choices made by users on different alternatives. Demand models depend on the characteristic attributes of the users and transport networks, as well as on parameters. Their significance translates into the reliability of the model in reproducing users’ choices as demand values. Traffic counts are aggregated data that can be used to update demand values of O/D matrix and/or for re-calibrating parameters from sets of parameters obtained in different situations or at different times in the same scenario using a reverse assignment modal. This paper provides the use of passenger counts to update national air transport demand by calibrating a hierarchical logit model. The application focuses on estimating the demand values for a secondary airport of an underdeveloped European region with the calibration of the logsum parameter working between distribution and modal choice. The updated model can be used to test new conditions for the supply of a new service or to increase the frequency or to modify the ticket level by means of public service obligations. The results show that the introduction of public obligations in the secondary airport in an underdeveloped region is crucial for future sustainability. Considering the decline in the economic, social and environmental sustainability in the region, the airport is central to economic and social development at the same time as being important for environmental sustainability, as it limits the impacts on the territory related to the construction of large transport infrastructures.

Passenger counting in minivan-taxis using crowd-sourcing and hierarchical clustering

A novel method for estimating the passenger densities of minivan taxis popularly known as Trotros in Ghana is proposed. A smartphone is used to collect time, location and velocity data from groups of passengers travelling in parts of the Kumasi Metropolitan Assembly, Ghana. Passengers are clustered by four different combinations of their location, time and average acceleration data using the agglomerative hierarchical clustering algorithm. A classification method was then used to externally validate the results by comparing the cluster labels to an initial class labelling which had been assigned during data collection called the group code. The count of the group code represented the estimated number of passengers aboard the vehicle. Results from the various clustering combinations performed indicated that using the time and location variables only for classification gave the highest accuracy of about 89.2% as compared to the other combinations. The proposed method of counting passengers in moving vehicles is particularly useful in the Ghanaian context due to the fact that trotros do not have to be retro-fitted with expensive devices for data collection and thus can be implemented without financially burdening the privately-owned trotro industry. Also, counting passengers in trotros adds to the growing pool of trotro research data which is beneficial for improvements in the trotro industry and also for future research.

Estimating the route-level passenger demand profile from bus dwell times

This paper explores the feasibility of estimating the demand profile of a bus route consisting of boarding flows, alighting flows, passenger loads, and origin–destination (OD) flows using dwell times as the main data source. Dwell times can be obtained from Automatic Vehicle Location (AVL) data, or door opening and closing records. This data is often more obtainable than the direct demand information based on passenger counts or fare collection information. A flexible Bayesian framework allowing for various dwell time models is employed to infer the passenger demand profile conditional on the observed bus dwell times. We model passenger flows by stop-specific arrival rates and a conditional alighting probability matrix. A modified gravity model is used to further explain the alighting probability matrix and to reduce the number of unknown parameters. Hamiltonian Monte Carlo sampling is implemented to draw the posterior distribution for the parameters of interest. The methodology is validated by three lines of a bus system where alighting times are the main activity time. For these lines dwell time as well as passenger count data are available. The case study shows that accurate estimates of alighting flows and passenger loads can be derived by the proposed framework with dwell times only but that room remains for improving the accuracy of boarding flows and OD flows. Additional boarding counts of a few bus runs, which can be easily integrated into the observation block of the estimation framework, raise the estimation accuracy of boarding and OD flows to an acceptable level. We therefore conclude that the proposed methodology can help “data-poor” bus operators to overcome demand estimation problems for transit planning and operation.