Ben Waterson

The evolution of urban traffic control: changing policy and technology

Abstract The history of urban traffic control (UTC) throughout the past century has been a continued race to keep pace with ever more complex policy objectives and consistently increasing vehicle demand. Many benefits can be observed from an efficient UTC system, such as reduced congestion, increased economic efficiency and improved road safety and air quality. There have been significant advances in vehicle detection and communication technologies which have enabled a series of step changes in the capabilities of UTC systems, from early (fixed time) signal plans to modern integrated systems. A variety of UTC systems have been implemented throughout the world, each with individual strengths and weaknesses; this paper seeks to compare the leading commercial systems (and some less well known systems) to highlight the key characteristics and differences before assessing whether the current UTC systems are capable of meeting modern transport policy obligations and desires. This paper then moves on to consider current and future transport policy and the technological landscape in which UTC will need to operate over the coming decades, where technological advancements are expected to move UTC from an era of limited data availability to an era of data abundance.

Quantifying the potential savings in travel time resulting from parking guidance systems - a simulation case study

Parking Guidance and Information (PGI) signs are thought to enable a more efficient use of the available parking stock. Despite the installation of PGI systems in many cities and their operation for a number of years, there is a lack of reliable evidence of the size of the benefits that these systems can achieve. This paper describes the development of driver parking choice models (both during the journey and pre-trip) and the implementation of these models in the existing network traffic simulation model RGCONTRAM. Besides quantifying the effects of the PGI system on both the drivers seeking suitable parking spaces and the parking stock itself, this also enables quantification of the impact of parking choice on the wider network. Factors influencing PGI effectiveness are described and conclusions are drawn that illustrate the potential of PGI to induce the demand to spread more efficiently across the parking stock.

An international review of roundabout capacity modelling

Roundabouts are an increasingly common form of road junction worldwide, and their effective design requires a detailed analysis of maximum vehicle throughput capacities. In this paper, the worldwide state-of-the-art in roundabout capacity modelling is examined, covering the three main methodologies on which models are based: fully-empirical, gap acceptance and simulation. It is shown that due to their limitations, each of these methodologies on their own cannot completely explain the complex behavioural and physical processes involved in roundabout entries, hence all the models require strong semi-empirical or fully-empirical bases using data obtained from their countries of origin. Differences in driver behaviour and methodologies thus result in differences in predicted capacities by the various models, and although local calibration allows some transferability, it is often limited by the availability of data or an incomplete understanding of the relationships between model parameters and capacity.

An automated signalized junction controller that learns strategies from a human expert

An automated signalized junction control system that can learn strategies from a human expert has been developed. This system applies machine learning techniques based on logistic regression and neural networks to affect a classification of state space using evidence data generated when a human expert controls a simulated junction. The state space is constructed from a series of bids from agents, which monitor regions of the road network. This builds on earlier work which has developed the High Bid auctioning agent system to control signalized junctions using localization probe data. For reference the performance of the machine learning signal control strategies are compared to that of High Bid and the MOVA system, which uses inductive loop detectors. Performance is evaluated using simulation experiments on two networks. One is an isolated T-junction and the other is a two junction network modelled on the High Road area of Southampton, UK. The experimental results indicate that machine learning junction control strategies trained by a human expert can outperform High Bid and MOVA both in terms of minimizing average delay and maximizing equitability; where the variance of the distribution over journey times is taken as a quantitative measure of equitability. Further experimental tests indicate that the machine learning control strategies are robust to variation in the positioning accuracy of localization probes and to the fraction of vehicles equipped with probes.

An automated signalized junction controller that learns strategies by temporal difference reinforcement learning

This paper shows how temporal difference learning can be used to build a signalized junction controller that will learn its own strategies through experience. Simulation tests detailed here show that the learned strategies can have high performance. This work builds upon previous work where a neural network based junction controller that can learn strategies from a human expert was developed (Box and Waterson, 2012). In the simulations presented, vehicles are assumed to be broadcasting their position over WiFi giving the junction controller rich information. The vehicles position data are pre-processed to describe a simplified state. The state-space is classified into regions associated with junction control decisions using a neural network. This classification is the strategy and is parametrized by the weights of the neural network. The weights can be learned either through supervised learning with a human trainer or reinforcement learning by temporal difference (TD). Tests on a model of an isolated T junction show an average delay of 14.12s and 14.36s respectively for the human trained and TD trained networks. Tests on a model of a pair of closely spaced junctions show 17.44s and 20.82s respectively. Both methods of training produced strategies that were approximately equivalent in their equitable treatment of vehicles, defined here as the variance over the journey time distributions.

Remote automatic incident detection using inductive loops

This paper describes the remote automatic incident detection algorithm designed to detect abnormal periods of traffic congestion existing over single inductive loop detectors (typically 2 × 1·5 m). This algorithm identifies those detectors which show a critical increase in average loop-occupancy time per vehicle coinciding with a critical decrease in average time-gap between vehicles according to a set of rules previously defined by the operator. The rules define the maximum and minimum values of loop occupancy and time gap respectively for each detector, which when exceeded for a given duration, trigger a report of a potential traffic flow ‘abnormality’ for that time of day at that particular location on the network. Initial rules are developed by studying the 85th percentile values of loop occupancy returned by the urban traffic control system every 30 s. A real-time trial took place between 07:00 and 19:00 over 167 consecutive days involving 74 detectors situated along two sections of the A33 Bassett A...

Can locker box logistics enable more human centric medical supply chains

The fast flow of goods into hospitals is often stalled by the external–internal supply chain interface (i.e. the receipts department). This issue is particularly pertinent regarding the delivery of urgent items for specific patients or in the event of low inventory levels. An unattended electronic locker bank to which individual urgent items can be delivered and subsequently collected by the ‘user’ was proposed for Great Ormond Street Hospital in London, UK. The feasibility of this concept is quantified using a hill-climbing model operating with a significant database of consignment movements and qualitatively using staff interviews. Results indicate that a locker bank measuring 4 m length, 1.7 m height and 0.8 m depth, comprising 11 partitions, would be required to accommodate all urgent consignments for any given day. Staff perceptions of the concept were positive suggesting the locker would potentially improve the speed and quality of health care delivered to patients.

Modeling the dynamic cut-in situation

An instrumented vehicle study was performed on a motorway in the United Kingdom to examine the behavior of drivers faced with the cut-in of a vehicle lane-changing into the space between themselves and the preceding vehicle. Data concerning this activity are in short supply and may be used not only in formulating models of human response in driving but also in designing and optimizing driver assistance aids such as adaptive cruise control (ACC). The cut-in vehicle used was equipped with a rear-facing radar unit enabling it to monitor the degree and speed with which drivers attempted to restore their original headway. Cut-ins from both directions were examined—moving in from a slower lane (94 events) and from a faster lane (72 events). The criticality experienced by the follower vehicle ranged from moderately severe [time to collision (TTC) around 10 s and time gap around 0.35 s] to noncritical (lead car’s speed at cut-in greater than follower speed and time headway beyond steady-state values). Findings indicate that the “pullback” behavior, at least over the initial 5 to 10 s, can be described by a constant pullback speed (the rate of decrease of the initial speed), and causative models for this response have been derived using “instantaneous” variables (those that may be calculated on cut-in, such as relative speed and TTC) and longer-term “target” variables, such as desired headway, the former of which have been most effective in describing behavior. Finally, empirical responses have been compared with those that would be produced by ACC systems; it was found that a comparatively close match is produced for low values of relative speed.

The scope for joint household/commercial waste collections: a case study

Although commercial and household wastes are compositionally similar, common UK practice is for separate collections. This paper uses vehicle routing and scheduling software to predict the benefits of allowing household and commercial wastes to be collected together by a common vehicle fleet. This was compared in a case study in which collections were made from over 25,000 households on an alternate-weekly basis and from 577 commercial premises having one or more collections each week. Modelled joint collections reduced vehicle mileage by up to 9.8%, equating to an annual saving of around £36,800 and a carbon equivalent saving of 2688 kg per annum. The modelled benefits were greatest when a common starting time (6 a.m.) was adopted for the commercial and household collections. The modelled rounds were estimated to have sufficient time and vehicle capacity available to allow an additional 50% of commercial waste to be collected, equating to 35.8 tonnes per week.

Monitoring household waste recycling centres performance using mean bin weight analyses

This paper describes a modelling approach used to investigate the significance of key factors (vehicle type, compaction type, site design, temporal effects) in influencing the variability in observed nett amenity bin weights produced by household waste recycling centres (HWRCs). This new method can help to quickly identify sites that are producing significantly lighter bins, enabling detailed back-end analyses to be efficiently targeted and best practice in HWRC operation identified. Tested on weigh ticket data from nine HWRCs across West Sussex, UK, the model suggests that compaction technique, vehicle type, month and site design explained 76% of the variability in the observed nett amenity weights. For each factor, a weighting coefficient was calculated to generate a predicted nett weight for each bin transaction and three sites were subsequently identified as having similar characteristics but returned significantly different mean nett bin weights. Waste and site audits were then conducted at the three sites to try and determine the possible sources of the remaining variability. Significant differences were identified in the proportions of contained waste (bagged), wood, and dry recyclables entering the amenity waste stream, particularly at one site where significantly less contaminated waste and dry recyclables were observed.