Probabilistic graphical models for semi-supervised traffic classification

Traffic classification using machine learning continues to be an active research area. The majority of work in this area uses off-the-shelf machine learning tools and treats them as black-box classifiers. This approach turns all the modelling complexity into a feature selection problem. In this paper, we build a problem-specific solution to the traffic classification problem by designing a custom probabilistic graphical model. Graphical models are a modular framework to design classifiers which incorporate domain-specific knowledge. More specifically, our solution introduces semi-supervised learning which means we learn from both labelled and unlabelled traffic flows. We show that our solution performs competitively compared to previous approaches while using less data and simpler features. Copyright © 2010 ACM.

Model development & application for inland navigational canals

Several agencies in the United Kingdom have interest in the water quality of old navigational canals that have fallen into disuse after the decline of commercial canal transportation. The interested agencies desired a model to predict the water quantity and quality of inland navigational canals in order to evaluate management options to address the issues in the natural streams to which they discharge. Inland navigational canals have unique drivers of their hydrology and water quality compared to either natural streams, irrigation canals, or larger navigational canals connected to seas or oceans. Water in an inland canal is typically sourced from a reservoir and artificially pumped to a summit reach; its movement downhill is controlled by the activity of boats and overflow weirs. Stagnant impoundments between locks, which might normally be expected to result in a decrease in the concentration of sediment-associated pollutants, actually have surprisingly high levels of sediment due to boat traffic. Algal growth in the stagnant reach can be high. This paper describes a canal model developed to simulate hydrology and water quality in inland navigational canals. This model was successfully applied to the Kennet and Avon Canal to predict hydrology, sediment generation and transport, and algal growth and transport. The model is responsive to external influences such as sunlight, temperature, nutrient concentrations, boat traffic, and runoff from the contributing catchment area.

Effects of longer heavy vehicles on traffic congestion

Two-lane, "microscopic" (vehicle-by-vehicle) simulations of motorway traffic are developed using existing models and validated using measured data from the M25 motorway. An energy consumption model is also built in, which takes the logged trajectories of simulated vehicles as drive-cycles. The simulations are used to investigate the effects on motorway congestion and fuel consumption if "longer and/or heavier vehicles" (LHVs) were to be permitted in the UK. Baseline scenarios are simulated with traffic composed of cars, light goods vehicles and standard heavy goods vehicles (HGVs). A proportion of conventional articulated HGVs is then replaced by a smaller number of LHVs carrying the same total payload mass and volume. Four LHV configurations are investigated: an 18.75 m, 46 t longer semi-trailer (LST); 25.25 m, 50 t and 60 t B-doubles and a 34 m, 82 t A-double. Metrics for congestion, freight fleet energy consumption and car energy consumption are defined for comparing the scenarios. Finally, variation of take-up level and LHV engine power for the LST and A-double are investigated. It is concluded that: (a) LHVs should reduce congestion particularly in dense traffic, however, a low mean proportion of freight traffic on UK roads and low take-up levels will limit this effect to be almost negligible; (b) LHVs can significantly improve the energy efficiency of freight fleets, giving up to a 23% reduction in fleet energy consumption at high take-up levels; (c) the small reduction in congestion caused by LHVs could improve the fuel consumption of other road users by up to 3% in dense traffic, however in free-flowing traffic an opposite effect occurs due to higher vehicle speeds and aerodynamic losses; and (d) underpowered LHVs have potential to generate severe congestion, however current manufacturers' recommendations appear suitable. © 2013 IMechE.