TADA

Researchers: Liao Ming

Introduction

TADA stands for Traffic Analytics with Data Accretion, a brand new tool/technique that allows contextualization of sensor data from physical sensors (e.g., GPS, vehicle sensors, traffic sensors) and from personal observations using physiology sensors.

Further description:

  • Contemporary traffic lights cause uneven flow of traffic in urban centers (e.g., the city of Edmonton); 3.6 billion vehicle-hours of delay; 67.5 billion in lost productivity – cost of traffic congestion to the USA economy in year 2000. In addition to human impact, this is a major impact on the environment – 21.6 billion liters of wasted fuel.
  • Contemporary mathematical and computational models are good at predicting the flow density of traffic; however, they are also impractical in the context of automating the flow of traffic.
  • We propose a novel technology that integrates mathematical and causal traffic models with the goal of optimizing the flow of traffic, at real-time. The technology will be able to use historical data as well as real-time data.
  • The proposed technology can be embedded in each traffic light controller along a major artery in an urban center.       The traffic lights will communicate and cooperate with each other, in a semi-automatic fashion, with the goal of smoother flow of traffic, hence saving the liquid gold.
  • Requirements:
    • A pilot city partner (e.g., The City of Edmonton)
    • A technology manufacturing partner firm to reconfigure the controllers of a few traffic lights along a major roadway (e.g., Gateway Boulevard in Edmonton); the firm will develop embedded controllers that dynamically synchronise/regulate the timing of the lights depending on traffic data
    • A technology video analysis partner firm to estimate the number of vehicles crossing a traffic light
    • A technology data analysis firm to compute wireless data
    • Funding to engage graduate researchers in this research and to develop software & hardware technologies by the industry 

Traffic is a major factor in all walks of life, consuming space, time, and energy in planetary scales. Literature reports a number of studies, systems, and policies that identify the impact of traffic on the quality of life and the environment in many dimensions such as urban traffic control, traffic of hazardous material, traffic pollution, and traffic and the human psyche. The GDP typically devoted to transportation by developed countries is in between 5 and 12 percent [Hazelton 2010].

Traffic modelling is a key research area in urban planning and environmental sciences. In 2000, road traffic congestion in USA alone caused 3.6 billon vehicle-hours of delay, 21.6 billion liters of wasted fuel, and US$67.5 billion in lost productivity. Yearly estimates on economic, health, and environmental cost of traffic congestion in New Zealand is in excess of NZ$1 billion [Hazelton, 2010]. Understandably, these statistics were derived almost exclusively from urban traffic data.

Traditionally, traffic modelling has concentrated on simulating traffic behaviour. The science of traffic analysis, modelling, and optimization aims to estimate traffic load, to detect and prevent traffic congestion, and to optimize the flow of traffic. Optimization of traffic flow not only reduces drivers’ stress levels, but also reduces air pollution [Angleno, 1999] and controls fuel consumption with respect to the environment and the economy. This proposal directly addresses the later – to optimise vehicular gasoline consumption in urban centers by regulating the flow of traffic using smart traffic lights.

Classical traffic models are mostly based on the treatment of vehicles on the road, their statistical distribution, or their density and average velocity as a function of space and time. Most models employ techniques ranging from cellular automata, particle-hopping, car-following, gas-kinetics, through to fluid dynamics present a passive approach to traffic optimization. That is, traffic data is collated apriori and the models are validated posthoc. In a compelling argument for the need to change the manual adjustments to traffic signals, Thorpe [1997] showed, using limited simulation models, that the best traffic signal performance could be achieved using Reinforcement Learning.

Modern traffic control is carried out in three incrementally informed methods. The first method, the least informed of all, simply employs humans at traffic junctions to manually regulate traffic signals. The second method employs traffic signal lights with static states1, where the states are fine-tuned manually based on information obtained from abstract traffic models [Huang et al., 2005; Wei et al., 2005]. For example, Thorpe [1997] reports the re-timing of a major artery in Denver, CO, USA, from 90 seconds to 100 seconds, in the heavy-flow direction, to yield 87% reduction in times vehicles stopped at light. The third method employs traffic lights that respond to real-time data obtained from devices such as road loops, video cameras, and other traffic detectors [Olsson, 1996].

In contemporary models, traffic situations are represented by statistical or mathematical abstractions and traffic control is exerted by methods that utilize information gleaned from these abstractions. These methods employ sparse traffic data and/or abstract models of traffic dynamics. Approaching from a different angle, this proposal focuses primarily on loosely modelling the causality of traffic. The causal model then drives the state changes in traffic control. Such a causal model approaches a fully-informed solution; that is, the more we know at real-time about vehicles on the road the smoother the flow of traffic and the better the gasoline usage. Data is obtained from every vehicle that contributes to traffic and this data is used to contextualise traffic situations. Hence, the proposed method will be accurate enough to capture the exact nature of undesirable traffic outcomes (e.g., traffic jams, longer wait period, higher number of stops) as well as to model the causality of these undesirable traffic outcomes. It is also possible to direct the traffic to enact a desirable traffic outcome, such as one that clears a pathway for an ambulance or a VIP’s convoy. Further, the causal model is updated at real-time and hence the state changes are real-time responses to the dynamics of the causal model. In essence, we propose to develop a probabilistic model for traffic signal control to optimize traffic flow in real-time.

We propose to use Dynamic Bayesian Belief Network (DBBN) technique to model and simulate urban ground traffic behaviour and to show how the DBBN optimizes traffic flow in real-time by controlling states of traffic signals. GPS and GIS technologies enable real-time access to traffic data such as vehicle’s location, speed, and direction [Haupt, 1999]. This real-time data, in combination with data specific to road segments, is sufficient to model entities affecting the flow of the traffic in a DBBN.

Each traffic signal in the road network is associated with its own instance of DBBN. Such an island of traffic signal connected through road segments forms the topology of the DBBN. The island accepts traffic flow data as observation model (effect) and current states of the traffic signal controller as the transition model (cause), and computes probabilities of states of the traffic signal for optimal flow of traffic.

Adjacent islands coordinate to optimize flow of traffic across road segments, thus one would be able to dynamically route traffic in order to minimize the waiting time for a designated vehicle (such as an ambulance) or for a group of vehicles in between two traffic signals.

The proposed Bayesian technique has no more intrinsic ability to represent causation than a traditional mathematical model except that it is more explicit and is directly manipulable by end users (e.g., The traffic control section of the City of Edmonton). Also, Bayesian models use the same amount of information as that is available to contemporary mathematical and statistical models, hence model updates and validation metrics are quite comparable across all models if not the same.

In line with the goals of contemporary traffic control models, the proposed Bayesian traffic signal control has the:

  1. ability to model the causal elements of traffic situations
  2. ability to control traffic signals using probabilistic inferences
  3. ability to regulate traffic for vehicle-specific situations (e.g., ambulances, fire engines, VIP vehicle)
  4. ability to predict undesirable traffic situations and propose timely alternatives 

References:

M.Hazelton, “Statistical Methods in Transportation Research”, PowerPoint Presentation to Statistical Society of Australia Inc, WA Branch, Perth, 2005, http://www.massey.ac.nz/massey/about-massey/news/article.cfm?mnarticle=caught-up-in-traffic-25-04-2009

E.Angleino, “Traffic Induced Air Pollution in Milan City: a Modelling Study”, Urban Transport and the Environment for the 21st Century”, Rhodos, September 1999.

T.Haupt, “Planning and Analyzing Transit Networks: An Integrated Approach Regarding Requirements of Passengers and Operators”, the 2nd GIS in Transit Conference, Tampa, Florida, 1999.

T.L.Thorpe, “Vehicle Traffic Light Control Using SARSA”, Masters Thesis, Department of Computer Science, Colorado State University, April 1997.

Huang YS, Chung TH, Chen CT, “Modeling traffic signal control systems using timed colour Petri nets”, International conference on Systems, Man, and Cybernetics, Vol 1-4, pp: 1759-1764, 2005.

Wei J, Wang A, Du N, “Study of self-organizing control of traffic signals in an urban network based on cellular automata”, IEEE Transactions on Vehicular Technology, 54 (2), pp 744-748, 2005.

1 Traffic states refer to all possible combinations of light changes (red, green, amber), pedestrian controls (walk or no walk), turn signals (left, right, and so on.), and other control mechanisms

 

 

Leave a Reply