1,720,993 research outputs found
Dataset for Modelling the dispersion of aircraft trajectories using Gaussian processes
No full textMatlab code and trajectory data supporting:
Eerland, Willem, Box, Simon and Sobester, Andras (2016) Modelling the dispersion of aircraft trajectories using Gaussian processes. Journal of Guidance Control and Dynamics.</span
Supervised learning from human performance at the computationally hard problem of optimal traffic signal control on a network of junctions
No full textOptimal switching of traffic lights on a network of junctions is a computationally intractable problem. In this research, road traffic networks containing signallized junctions are simulated. A computer game interface is used to enable a human ‘player’ to control the traffic light settings on the junctions within the simulation. A supervised learning approach, based on simple neural network classifiers can be used to capture human player's strategies in the game and thus develop a human-trained machine control (HuTMaC) system that approaches human levels of performance. Experiments conducted within the simulation compare the performance of HuTMaC to two well-established traffic-responsive control systems that are widely deployed in the developed world and also to a temporal difference learning-based control method. In all experiments, HuTMaC outperforms the other control methods in terms of average delay and variance over delay. The conclusion is that these results add weight to the suggestion that HuTMaC may be a viable alternative, or supplemental method, to approximate optimization for some practical engineering control problems where the optimal strategy is computationally intractable
Design and testing of a nanoparticle spectrometer
No full textThis thesis is concerned with a project to design and test a new Nanoparticle Spectrometer (NPS). The NPS is an instrument designed to make fast measurements of the size distribution and number concentration of aerosol samples containing particles in the size range 5–300nm. The intended application of the NPS is to take time dependant measurements of the aerosols emitted from internal combustion engines. The primary motivation for this work is ultimately the potentially detrimental effects on human health and the environment of combustion generated aerosols.In common with previous aerosol spectrometers, the Nanoparticle Spectrometer consists of a charger to give particles an electrostatic charge, a classifier, which separates the particles in an aerosol sample according to their electrical mobility (a function of size) and an array of counting devices that count the numbers of particles with different mobilities. The novelty of the NPS is the geometry of the instrument, which, it will be argued, has certain advantages.The behaviour of particles in the classifier has been modelled numerically and this model has been used to optimise the classifier geometry. Two charger designs were considered, and two analytical charger models developed and compared. The classifier model was combined with the selected charger model to create a simulation of the instrument operation, which predicts the NPS’ output signal for a given aerosol sample size distribution and number concentration.A prototype NPS was designed, built and tested experimentally. The objective of the experiments was to test the validity of the instrument model and compare the performance of the NPS to an established slow response particulate measuring instrument, the SMPS. The experiments showed good agreement between modelled and measured results, as well as close correlation between the NPS and the SMPS results across most of the instruments range.The experiments also revealed some areas in which the performance of the NPS could be improved; for instance, the modelling of diffusion in the classifier and of the fluid flow in the particle charger
Cambridge Rocketry Simulator V3.1
No full textThe Cambridge Rocketry Simulator can be used to simulate the flight of unguided rockets for both design and operational applications. The software consists of three parts; the first part is a GUI that enables the user to design a rocket. The second part is a verified and peer-reviewed physics model that simulates the rocket flight. This includes a Monte Carlo wrapper to model the uncertainty in the rocket's dynamics and the atmospheric conditions. The third part generates visualizations of the resulting trajectories, including nominal performance and uncertainty analysis, e.g. a splash-down region with confidence bounds. The project is available on SourceForge, and is written in Java (GUI), c++ (simulation core) and Python (visualization). While all parts can be executed from the GUI, the three components share information via XML, accommodating modifications, and re-use of individual components.</span
Quantifying the impact of probe vehicle localisation data errors on signalised junction control
No full textThis study investigates theoretical signal control algorithms based solely on probe vehicle data. Through development of a simulation system that can model urban signalised junction control using localisation probe data from all vehicles in the local area, improvements in junction operational efficiency that result from the improved input data are demonstrated for both isolated and coordinated junctions. Results from the isolated junction scenario show that the richness of the information contained within probe vehicle data means that control algorithms based just on positions and velocities of vehicles can produce 25% reductions in average delay compared to the current standard control algorithm MOVA. Results from the twin junction scenario confirm the importance of using high-level synchronisation to coordinate closely connected junctions, achieving reductions in average delays (compared to independent control approaches) of up to 40% through a process of weighting the probe vehicle data to reflect prior stage decisions of other parts of the junction. Critical to achieving these benefits, however, is the availability of high localisation accuracy probe data, with results indicating that the levels of accuracy necessary are representative of the typical performance of current in-vehicle global positioning system units, except when those vehicles are operating in urban canyon environments
Cambridge Rocketry Simulator V3.0
No full textThe third version of the Cambridge Rocketry Simulator consists of three parts to simulating a high power rocket flight. First the user is able to build a rocket from basic components using a GUI. Second, a verified and reviewed physics model simulates the rocket flight, this includes a Monte Carlo wrapper to quantify the uncertainty in dynamics and atmosphere. Finally, the visualization of the resulting trajectories includes nominal performance and uncertainty analysis, e.g. a splash-down region with confidence bounds. The project is available on SourceForge, and is written in Java (GUI), C++ (simulation core) and Python (visualization). While all parts can be executed from the GUI, the three components share information via XML, accommodating modifications, and re-use of individual components. The software can used to simulate the flight of unguided rockets, with both design and operational applications.</span
Machine learning in signalized junction control algorithms
No full textMachine learning techniques can be applied to develop signalized junction control algorithms that can learn control strategies from examples of good control and from experience. This paper discusses the conceptual differences between the conventional approach to signal control and the machine learning approach. An example is presented where a junction control agent was developed to learn strategies from a human expert. This learning junction agent uses localization probe data from vehicles and a system of bids to describe the state of the network. The junction agent learns from the human expert by employing a Neural Network to classify its bid space based on evidence of the human’s decision making.Simulation experiments are used to evaluate the performance of learning junction agent and these show that the agent can outperform the High Bid signal control system both in terms of delay and in terms of equitability. The paper concludes with a discussion on how the approach described above can be extended to allow the junction control agent to learn from observational data and experience using reinforcement learning.<br/
Investigating the effects of mixed driver reaction times in the transport network
No full textGiven the advances in autonomous vehicle technology, vehicles may soon be ableto drive as well as humans. Here, autonomous vehicles are approximated as drivers with reaction times smaller than those of human drivers, but with the same driving competency. Experiments were performed to simulate the effects of loading the transport network with increasingly high proportions of vehicles with faster than normal reaction times, on four different road models
- …
