Safety assurance for automated driving systems that can adapt using machine learning: A qualitative interview study.

INTRODUCTION: Automated Driving Systems (ADSs) present significant unresolved challenges for traditional safety assurance frameworks. These frameworks did not envisage, nor readily support, automated driving without the active involvement of a human driver, or support safety-critical systems using Machine Learning (ML) to modify their driving functionality during in-service operation. METHOD: An in-depth qualitative interview study was conducted as part of a broader research project on safety assurance of ADSs that can adapt using ML. The objective was to capture and analyze feedback from leading global experts, from both regulatory and industry stakeholders, with the key objectives of identifying themes that could assist with the development of a safety assurance framework for ADSs, and providing a sense of the level of support and feasibility for various safety assurance concepts relevant to ADSs. RESULTS: Ten themes were identified from an analysis of the interview data. Several themes support a whole-of-life safety assurance approach for ADSs, with strong support for ADS developers to be required to produce a Safety Case, and for ADS operators to maintain a Safety Management Plan throughout an ADSs operational life. There was also strong support for in-service ML-enabled changes to be allowed within pre-approved system boundaries, although there were mixed views on whether human oversight of such changes should be required. Across all themes identified, there was support for progressing reform within current regulatory frameworks, without requiring wholesale changes to current frameworks. The feasibility of some themes was identified as presenting challenges, particularly with the ability for regulators to develop and maintain an appropriate level of knowledge, capability and capacity, and with the ability to effectively articulate and pre-approve boundaries within which in-service changes can occur without additional regulatory approval. CONCLUSIONS: Further research on the individual themes and findings would be beneficial to support more informed reform decisions.

Pedestrian Origin-Destination Estimation Based on Multi-Camera Person Re-Identification.

Pedestrian origin-destination (O-D) estimates that record traffic flows between origins and destinations, are essential for the management of pedestrian facilities including pedestrian flow simulation in the planning phase and crowd control in the operation phase. However, current O-D data collection techniques such as surveys, mobile sensing using GPS, Wi-Fi, and Bluetooth, and smart card data have the disadvantage that they are either time consuming and costly, or cannot provide complete O-D information for pedestrian facilities without entrances and exits or pedestrian flow inside the facilities. Due to the full coverage of CCTV cameras and the huge potential of image processing techniques, we address the challenges of pedestrian O-D estimation and propose an image-based O-D estimation framework. By identifying the same person in disjoint camera views, the O-D trajectory of each identity can be accurately generated. Then, state-of-the-art deep neural networks (DNNs) for person re-ID at different congestion levels were compared and improved. Finally, an O-D matrix based on trajectories was generated and the resident time was calculated, which provides recommendations for pedestrian facility improvement. The factors that affect the accuracy of the framework are discussed in this paper, which we believe could provide new insights and stimulate further research into the application of the Internet of cameras to intelligent transport infrastructure management.

Pedestrian flow characteristics through different angled bends: Exploring the spatial variation of velocity.

Common geometrical layouts could potentially be bottlenecks, particularly during emergency and high density situations. When pedestrians are interacting with such complex geometrical settings, the congestion effect might not be uniform over the bottleneck area. This study uses the trajectory data collected through a controlled laboratory experiment to explore the spatial variation of speeds when a group of people navigates through bends. Four turning angles, i.e., 45°, 90°, 135° and 180°, with a straight corridor and two speed levels, i.e., normal speed walking and slow running (jogging), were considered in these experiments. Results explained that the speeds are significantly different over the space within the bend for all angles (except 0°) under both speed levels. In particular, average walking speeds are significantly lower near the inner corner of the bend as compared to the outer corner. Further, such speed variations are magnified when the angle of the bend and desired speed increase. These outcomes indicate that even smaller turning angles, e.g., 45° could create bottlenecks near the inner corner of the bend, particularly when the walking speeds are high. The findings of this study could be useful in understanding the congestion and bottleneck effects associated with complex geometrical settings, and calibrating microscopic simulation tools to accurately reproduce such effects.

An outer approximation method for the road network design problem.

Best investment in the road infrastructure or the network design is perceived as a fundamental and benchmark problem in transportation. Given a set of candidate road projects with associated costs, finding the best subset with respect to a limited budget is known as a bilevel Discrete Network Design Problem (DNDP) of NP-hard computationally complexity. We engage with the complexity with a hybrid exact-heuristic methodology based on a two-stage relaxation as follows: (i) the bilevel feature is relaxed to a single-level problem by taking the network performance function of the upper level into the user equilibrium traffic assignment problem (UE-TAP) in the lower level as a constraint. It results in a mixed-integer nonlinear programming (MINLP) problem which is then solved using the Outer Approximation (OA) algorithm (ii) we further relax the multi-commodity UE-TAP to a single-commodity MILP problem, that is, the multiple OD pairs are aggregated to a single OD pair. This methodology has two main advantages: (i) the method is proven to be highly efficient to solve the DNDP for a large-sized network of Winnipeg, Canada. The results suggest that within a limited number of iterations (as termination criterion), global optimum solutions are quickly reached in most of the cases; otherwise, good solutions (close to global optimum solutions) are found in early iterations. Comparative analysis of the networks of Gao and Sioux-Falls shows that for such a non-exact method the global optimum solutions are found in fewer iterations than those found in some analytically exact algorithms in the literature. (ii) Integration of the objective function among the constraints provides a commensurate capability to tackle the multi-objective (or multi-criteria) DNDP as well.

Collective movements of pedestrians: How we can learn from simple experiments with non-human (ant) crowds.

INTRODUCTION: Understanding collective behavior of moving organisms and how interactions between individuals govern their collective motion has triggered a growing number of studies. Similarities have been observed between the scale-free behavioral aspects of various systems (i.e. groups of fish, ants, and mammals). Investigation of such connections between the collective motion of non-human organisms and that of humans however, has been relatively scarce. The problem demands for particular attention in the context of emergency escape motion for which innovative experimentation with panicking ants has been recently employed as a relatively inexpensive and non-invasive approach. However, little empirical evidence has been provided as to the relevance and reliability of this approach as a model of human behaviour. METHODS: This study explores pioneer experiments of emergency escape to tackle this question and to connect two forms of experimental observations that investigate the collective movement at macroscopic level. A large number of experiments with human and panicking ants are conducted representing the escape behavior of these systems in crowded spaces. The experiments share similar architectural structures in which two streams of crowd flow merge with one another. Measures such as discharge flow rates and the probability distribution of passage headways are extracted and compared between the two systems. FINDINGS: Our findings displayed an unexpected degree of similarity between the collective patterns emerged from both observation types, particularly based on aggregate measures. Experiments with ants and humans commonly indicated how significantly the efficiency of motion and the rate of discharge depend on the architectural design of the movement environment. PRACTICAL APPLICATIONS: Our findings contribute to the accumulation of evidence needed to identify the boarders of applicability of experimentation with crowds of non-human entities as models of human collective motion as well as the level of measurements (i.e. macroscopic or microscopic) and the type of contexts at which reliable inferences can be drawn. This particularly has implications in the context of experimenting evacuation behaviour for which recruiting human subjects may face ethical restrictions. The findings, at minimum, offer promise as to the potential benefit of piloting such experiments with non-human crowds, thereby forming better-informed hypotheses.

A Doppler Radar System for Sensing Physiological Parameters in Walking and Standing Positions.

Doppler radar can be implemented for sensing physiological parameters wirelessly at a distance. Detecting respiration rate, an important human body parameter, is essential in a range of applications like emergency and military healthcare environments, and Doppler radar records actual chest motion. One challenge in using Doppler radar is being able to monitor several patients simultaneously and in different situations like standing, walking, or lying. This paper presents a complete transmitter-receiver Doppler radar system, which uses a 4 GHz continuous wave radar signal transmission and receiving system, to extract base-band data from a phase-shifted signal. This work reports experimental evaluations of the system for one and two subjects in various standing and walking positions. It provides a detailed signal analysis of various breathing rates of these two subjects simultaneously. These results will be useful in future medical monitoring applications.

How Simple Hypothetical-Choice Experiments Can Be Utilized to Learn Humans' Navigational Escape Decisions in Emergencies.

How humans resolve non-trivial tradeoffs in their navigational choices between the social interactions (e.g., the presence and movements of others) and the physical factors (e.g., spatial distances, route visibility) when escaping from threats in crowded confined spaces? The answer to this question has major implications for the planning of evacuations and the safety of mass gatherings as well as the design of built environments. Due to the challenges of collecting behavioral data from naturally-occurring evacuation settings, laboratory-based virtual-evacuation experiments have been practiced in a number of studies. This class of experiments faces the traditional question of contextual bias and generalizability: How reliably can we infer humans' behavior from decisions made in hypothetical settings? Here, we address these questions by making a novel link between two different forms of empirical observations. We conduct hypothetical emergency exit-choice experiments framed as simple pictures, and then mimic those hypothetical scenarios in more realistic fashions through staging mock evacuation trials with actual crowds. Econometric choice models are estimated based on the observations made in both experimental contexts. The models are contrasted with each other from a number of perspectives including their predictions as well as the sign, magnitude, statistical significance, person-to-person variations (reflecting individuals' perception/preference differences) and the scale (reflecting context-dependent decision randomness) of their inferred parameters. Results reveal a surprising degree of resemblance between the models derived from the two contexts. Most strikingly, they produce fairly similar prediction probabilities whose differences average less than 10%. There is also unexpected consensus between the inferences derived from both experimental sources on many aspects of people's behavior notably in terms of the perception of social interactions. Results show that we could have elicited peoples' escape strategies with fair precision without observing them in action (i.e., simply by using only hypothetical-choice data as an inexpensive, practical and non-invasive experimental technique in this context). As a broader application, this offers promising evidence as to the potential applicability of the hypothetical-decision experiments to other decision contexts (at least for non-financial decisions) when field or real-world data is prohibitively unavailable. As a practical application, the behavioral insights inferred from our observations (reflected in the estimated parameters) can improve how accurately we predict the movement patterns of human crowds in emergency scenarios arisen in complex spaces. Fully-generic-in-parameters, our proposed models can even be directly introduced to a broad range of crowd simulation software to replicate navigation decision making of evacuees.

Macroscopic modeling of pedestrian and bicycle crashes: A cross-comparison of estimation methods.

The paper presents a cross-comparison of different estimation methods to model pedestrian and bicycle crashes. The study contributes to macro level safety studies by providing further methodological and empirical evidence on the various factors that influence the frequency of pedestrian and bicycle crashes at the planning level. Random parameter negative binomial (RPNB) models are estimated to explore the effects of various planning factors associated with total, serious injury and minor injury crashes while accounting for unobserved heterogeneity. Results of the RPNB models were compared with the results of a non-spatial negative binomial (NB) model and a Poisson-Gamma-CAR model. Key findings are, (1) the RPNB model performed best with the lowest mean absolute deviation, mean squared predicted error and Akaiki information criterion measures and (2) signs of estimated parameters are consistent if these variables are significant in models with the same response variables. We found that vehicle kilometers traveled (VKT), population, percentage of commuters cycling or walking to work, and percentage of households without motor vehicles have a significant and positive correlation with the number of pedestrian and bicycle crashes. Mixed land use is also found to have a positive association with the number of pedestrian and bicycle crashes. Results have planning and policy implications aimed at encouraging the use of sustainable modes of transportation while ensuring the safety of pedestrians and cyclist.

Application of a random effects negative binomial model to examine tram-involved crash frequency on route sections in Melbourne, Australia.

Safety is a key concern in the design, operation and development of light rail systems including trams or streetcars as they impose crash risks on road users in terms of crash frequency and severity. The aim of this study is to identify key traffic, transit and route factors that influence tram-involved crash frequencies along tram route sections in Melbourne. A random effects negative binomial (RENB) regression model was developed to analyze crash frequency data obtained from Yarra Trams, the tram operator in Melbourne. The RENB modelling approach can account for spatial and temporal variations within observation groups in panel count data structures by assuming that group specific effects are randomly distributed across locations. The results identify many significant factors effecting tram-involved crash frequency including tram service frequency (2.71), tram stop spacing (-0.42), tram route section length (0.31), tram signal priority (-0.25), general traffic volume (0.18), tram lane priority (-0.15) and ratio of platform tram stops (-0.09). Findings provide useful insights on route section level tram-involved crashes in an urban tram or streetcar operating environment. The method described represents a useful planning tool for transit agencies hoping to improve safety performance.

An empirical Bayes safety evaluation of tram/streetcar signal and lane priority measures in Melbourne.

OBJECTIVE: Streetcars/tram systems are growing worldwide, and many are given priority to increase speed and reliability performance in mixed traffic conditions. Research related to the road safety impact of tram priority is limited. This study explores the road safety impacts of tram priority measures including lane and intersection/signal priority measures. METHOD: A before-after crash study was conducted using the empirical Bayes (EB) method to provide more accurate crash impact estimates by accounting for wider crash trends and regression to the mean effects. Before-after crash data for 29 intersections with tram signal priority and 23 arterials with tram lane priority in Melbourne, Australia, were analyzed to evaluate the road safety impact of tram priority. RESULTS: The EB before-after analysis results indicated a statistically significant adjusted crash reduction rate of 16.4% after implementation of tram priority measures. Signal priority measures were found to reduce crashes by 13.9% and lane priority by 19.4%. A disaggregate level simple before-after analysis indicated reductions in total and serious crashes as well as vehicle-, pedestrian-, and motorcycle-involved crashes. In addition, reductions in on-path crashes, pedestrian-involved crashes, and collisions among vehicles moving in the same and opposite directions and all other specific crash types were found after tram priority implementation. CONCLUSIONS: Results suggest that streetcar/tram priority measures result in safety benefits for all road users, including vehicles, pedestrians, and cyclists. Policy implications and areas for future research are discussed.

Safety impacts of platform tram stops on pedestrians in mixed traffic operation: A comparison group before-after crash study.

Tram stops in mixed traffic environments present a variety of safety, accessibility and transport efficiency challenges. In Melbourne, Australia the hundred year-old electric tram system is progressively being modernized to improve passenger accessibility. Platform stops, incorporating raised platforms for level entry into low floor trams, are being retro-fitted system-wide to replace older design stops. The aim of this study was to investigate the safety impacts of platform stops over older design stops (i.e. Melbourne safety zone tram stops) on pedestrians in the context of mixed traffic tram operation in Melbourne, using an advanced before-after crash analysis approach, the comparison group (CG) method. The CG method evaluates safety impacts by taking into account the general trends in safety and the unobserved factors at treatment and comparison sites that can alter the outcomes of a simple before-after analysis. The results showed that pedestrian-involved all injury crashes reduced by 43% after platform stop installation. This paper also explores a concern that the conventional CG method might underestimate safety impacts as a result of large differences in passenger stop use between treatment and comparison sites, suggesting differences in crash risk exposure. To adjust for this, a modified analysis explored crash rates (crash counts per 10,000 stop passengers) for each site. The adjusted results suggested greater reductions in pedestrian-involved crashes after platform stop installation: an 81% reduction in pedestrian-involved all injury crashes and 86% reduction in pedestrian-involved FSI crashes, both are significant at the 95% level. Overall, the results suggest that platform stops have considerable safety benefits for pedestrians. Implications for policy and areas for future research are explored.

Simulation of safety: a review of the state of the art in road safety simulation modelling.

Recent decades have seen considerable growth in computer capabilities, data collection technology and communication mediums. This growth has had considerable impact on our ability to replicate driver behaviour and understand the processes involved in failures in the traffic system. From time to time it is necessary to assess the level of development as a basis of determining how far we have come. This paper sets out to assess the state of the art in the use of computer models to simulate and assess the level of safety in existing and future traffic systems. It reviews developments in the area of road safety simulation models. In particular, it reviews computer models of driver and vehicle behaviour within a road context. It focuses on stochastic numerical models of traffic behaviour and how reliable these are in estimating levels of safety on the traffic network. Models of this type are commonly used in the assessment of traffic systems for capacity, delay and general performance. Adding safety to this assessment regime may allow more comprehensive assessment of future traffic systems. To date the models have focused primarily on vehicular traffic that is, cars and heavy vehicles. It has been shown that these models have potential in measuring the level of conflict on parts of the network and the measure of conflict correlated well with crash statistics. Interest in the prediction of crashes and crash severity is growing and new models are focusing on the continuum of general traffic conditions, conflict, severe conflict, crash and severe crashes. The paper also explores the general data types used to develop, calibrate and validate these models. Recent technological development in in-vehicle data collection, driver simulators and machine learning offers considerable potential for improving the behavioural base, rigour and application of road safety simulation models. The paper closes with some indication of areas of future development.

Factors affecting the probability of bus drivers being at-fault in bus-involved accidents.

Previous research has provided little insight into factors that influence the probability of bus drivers being at-fault in bus-involved accidents. In this study, an analysis was conducted on accident data compiled by a bus company that include an assessment on whether the bus driver was deemed by the company to hold primary responsibility for accident occurrence. Using a mixed logit modelling approach, roadway/environmental, vehicle and driver related variables that were identified to be influential were road type, speed limit, traffic/lighting conditions, bus priority, bus age/length and driver's age/gender/experience/historic at-fault accident record. Results were indicative of possible confined road-space issues that bus drivers face along routes with roadside traffic friction and point to the provision of exclusive right of way for buses as a possible way to address this. Results also suggest benefits in assigning routes comprising mainly divided roads as well as newer and shorter buses to less experienced drivers.

Bus accident analysis of routes with/without bus priority.

This paper summarises findings on road safety performance and bus-involved accidents in Melbourne along roads where bus priority measures had been applied. Results from an empirical analysis of the accident types revealed significant reduction in the proportion of accidents involving buses hitting stationary objects and vehicles, which suggests the effect of bus priority in addressing manoeuvrability issues for buses. A mixed-effects negative binomial (MENB) regression and back-propagation neural network (BPNN) modelling of bus accidents considering wider influences on accident rates at a route section level also revealed significant safety benefits when bus priority is provided. Sensitivity analyses done on the BPNN model showed general agreement in the predicted accident frequency between both models. The slightly better performance recorded by the MENB model results suggests merits in adopting a mixed effects modelling approach for accident count prediction in practice given its capability to account for unobserved location and time-specific factors. A major implication of this research is that bus priority in Melbourne's context acts to improve road safety and should be a major consideration for road management agencies when implementing bus priority and road schemes.