Abstract

div"Scientific problems are solved by finding the optimal solution for a specific task. Some problems can be solved analytically while other problems are solved using data driven methods. The use of digital technologies to improve the transportation of people and goods, which is referred to as intelligent mobility, is one of the principal beneficiaries of data driven solutions. Autonomous vehicles are at the heart of the developments that propel Intelligent Mobility. Due to the high dimensionality and complexities involved in real-world environments, it needs to become commonplace for intelligent mobility to use data-driven solutions. As it is near impossible to program decision making logic for every eventuality manually. While recent developments of data-driven solutions such as deep learning facilitate machines to learn effectively from large datasets, the application of techniques within safety-critical systems such as driverless cars remain scarce."/div""div"Autonomous vehicles need to be able to make context-driven decisions autonomously in different environments in which they operate. The recent literature on driverless vehicle research is heavily focused only on road or highway environments but have discounted pedestrianized areas and indoor environments. These unstructured environments tend to have more clutter and change rapidly over time. Therefore, for intelligent mobility to make a significant impact on human life, it is vital to extend the application beyond the structured environments. To further advance intelligent mobility, researchers need to take cues from multiple sensor streams, and multiple machine learning algorithms so that decisions can be robust and reliable. Only then will machines indeed be able to operate in unstructured and dynamic environments safely. Towards addressing these limitations, this thesis investigates data driven solutions towards crucial building blocks in intelligent mobility. Specifically, the thesis investigates multimodal sensor data fusion, machine learning, multimodal deep representation learning and its application of intelligent mobility. This work demonstrates that mobile robots can use multimodal machine learning to derive driver policy and therefore make autonomous decisions."/div""div"To facilitate autonomous decisions necessary to derive safe driving algorithms, we present an algorithm for free space detection and human activity recognition. Driving these decision-making algorithms are specific datasets collected throughout this study. They include the Loughborough London Autonomous Vehicle dataset, and the Loughborough London Human Activity Recognition dataset. The datasets were collected using an autonomous platform design and developed in house as part of this research activity. The proposed framework for Free-Space Detection is based on an active learning paradigm that leverages the relative uncertainty of multimodal sensor data streams (ultrasound and camera). It utilizes an online learning methodology to continuously update the learnt model whenever the vehicle experiences new environments. The proposed Free Space Detection algorithm enables an autonomous vehicle to self-learn, evolve and adapt to new environments never encountered before. The results illustrate that online learning mechanism is superior to one-off training of deep neural networks that require large datasets to generalize to unfamiliar surroundings. "/div""div"The thesis takes the view that human should be at the centre of any technological development related to artificial intelligence. It is imperative within the spectrum of intelligent mobility where an autonomous vehicle should be aware of what humans are doing in its vicinity. Towards improving the robustness of human activity recognition, this thesis proposes a novel algorithm that classifies point-cloud data originated from Light Detection and Ranging sensors. The proposed algorithm leverages multimodality by using the camera data to identify humans and segment the region of interest in point cloud data. The corresponding 3-dimensional data was converted to a Fisher Vector Representation before being classified by a deep Convolutional Neural Network. The proposed algorithm classifies the indoor activities performed by a human subject with an average precision of 90.3%. When compared to an alternative point cloud classifier, PointNet[1], [2], the proposed framework out preformed on all classes. "/div""div"The developed autonomous testbed for data collection and algorithm validation, as well as the multimodal data-driven solutions for driverless cars, is the major contributions of this thesis. It is anticipated that these results and the testbed will have significant implications on the future of intelligent mobility by amplifying the developments of intelligent driverless vehicles."/div

Abstract

Intelligent transportation systems (ITS) are becoming more and more effective. Robust and accurate short-term traffic prediction plays a key role in modern ITS and demands continuous improvement. Benefiting from better data collection and storage strategies, a huge amount of traffic data is archived which can be used for this purpose especially by using machine learning. For the data preprocessing stage, despite the amount of data available, missing data records and their messy labels are two problems that prevent many prediction algorithms in ITS from working effectively and smoothly. For the prediction stage, though there are many prediction algorithms, higher accuracy and more automated procedures are needed. Considering both preprocessing and prediction studies, one widely used algorithm is k-nearest neighbours (kNN) which has shown high accuracy and efficiency. However, the general kNN is designed for matrix instead of time series which lacks the use of time series characteristics. Choosing the right parameter values for kNN is problematic due to dynamic traffic characteristics. This thesis analyses kNN based algorithms and improves the prediction accuracy with better parameter handling using time series characteristics. Specifically, for the data preprocessing stage, this work introduces gap-sensitive windowed kNN (GSW-kNN) imputation. Besides, a Mahalanobis distance-based algorithm is improved to support correcting and complementing label information. Later, several automated and dynamic procedures are proposed and different strategies for making use of data and parameters are also compared. Two real-world datasets are used to conduct experiments in different papers. The results show that GSW-kNN imputation is 34% on average more accurate than benchmarking methods, and it is still robust even if the missing ratio increases to 90%. The Mahalanobis distance-based models efficiently correct and complement label information which is then used to fairly compare performance of algorithms. The proposed dynamic procedure (DP) performs better than manually adjusted kNN and other benchmarking methods in terms of accuracy on average. What is better, weighted parameter tuples (WPT) gives more accurate results than any human tuned parameters which cannot be achieved manually in practice. The experiments indicate that the relations among parameters are compound and the flow-aware strategy performs better than the time-aware one. Thus, it is suggested to consider all parameter strategies simultaneously as ensemble strategies especially by including window in flow-aware strategies. In summary, this thesis improves the accuracy and automation level of short-term traffic prediction with proposed high-speed algorithms.

Abstract

The main goal of this thesis is to evaluate different machine learning models in order to classify buyers of an electric or a hybrid vehicle and to identify the characteristics of the buyers in the European Union. Machine learning algorithms and techniques were adopted to analyze the dataset and to create models that could predict, with a certain accuracy, the customer’s willingness to buy an EV. Identification of the characteristics of the buyers were based on the identified most important features from the machine learning models and statistical analysis. The research consisted of exploratory and explanatory methods (mixed method) with quantitative and qualitative techniques. Quantitative technique was applied to convert categorical values to ordinal and nominal numeric values, to establish cause and effect relationship between the variables by using statistical analysis and to apply machine learning methods on the dataset. The quantitative results were then analyzed by using quantitative and qualitative techniques in order to identify the characteristics of the buyers. The data analytics part relied on a publicly available large dataset from the EU containing transport and mobility related data. From the experiments with logistics regression, support vector machine, random forest, gradient boosting classifier and the artificial neural network it was found that ANN is the best model to identify who won’t buy an EV and gradient boosting classifier is the best model to identify who would like to buy and EV. ML based feature importance identification methods (MDI, permutation feature importance) were used to analyze the characteristics of the buyers. The major buyer’s characteristics found in this thesis are environmental concern, knowledge on car sharing, country of residence, education, control traffic, gender, incentive, education and location of residence. Authors have recommended green marketing as the potential enablers towards a faster and larger adoption of electrical vehicles in the market as environmental impact was found as the most significant behavior of the buyer. Finally, for the future researchers, the authors have recommended fine-tuning the algorithms extensively in order to achieve better accuracy and to collect primary data based on the most important features identified in this thesis.

Abstract

orts are central to stimulating the growth and development of economies around the world. With the trend of increasing demand for air travel and transportation of goods, there is mounting pressure on the existing airport infrastructure. This leads to an imbalance in the capacity and demand at airports that costs airlines, passengers and the economy at large billions of dollars. A cost-effective solution to this growing imbalance is to develop ways to utilize the existing infrastructure well. Information from historical days that are similar to a day of operations can be used to gain insights to support traffic management decisions on that day. Recent advances in machine learning and computing power have made it possible to mine and analyze sizable historical archives of different variables that characterize and influence airport operations. Finding similar historical days can help better understand the impact of different traffic management initiatives (TMIs) and identify areas of capacity underutilization. Reduction in airport capacity underutilization can lead to reduction in airport delays. Decision support tools that can identify similar days and the TMIs taken on these past similar days and their resulting outcomes can augment controller experience to guide decision-making on the reference day at an airport. This information can allow air traffic managers to make less conservative decisions and thus improve airport capacity and reduce delays. This dissertation develops similarity measures between days using airport capacity and demand data. We find that dimensionality reduction is feasible for capacity data, and base capacity similarity on the principal components. Dimensionality reduction cannot be efficiently performed on demand data; consequently demand similarity is based on original data in this case. We find that both capacity and demand data lack natural clusters and thus propose that similarity be viewed as a continuous measure. Finally we estimate measures of overall distance based on both capacity and demand similarity. The estimated distances are visualized using Metric Multidimensional Scaling plots and indicate that most days with significant air traffic management activity are similar to certain other days, validating the potential of this approach for decision support. Accurate demand and capacity estimates are necessary to generate meaningful similarity measures that can be used in decision-support tools. Predicting airport capacity accurately can also help make better tradeoffs between allowing more flights to operate at the airport and minimizing expensive airborne delay. We develop accurate demand estimates from the Aggregate Demand List (ADL), which contains fine-grained flight schedule data of all the flights operating at an airport. Capacity of an airport can be observed only at sufficiently large demand. However, if the throughput of an airport is limited by the demand, we can only conclude that the capacity is larger than or equal to the observed throughput. The inability to directly observe capacity makes capacity prediction a challenging and less explored problem. This dissertation applies machine-learning methods that incorporate observations censored by insufficient demand to develop an airport capacity prediction model. Specifically, we explore Kaplan Meier estimator, Cox Proportional Hazards model and Random Survival Forest model to predict airport capacity. These models predict a capacity distribution rather than a single capacity value for an hour of interest at an airport using its weather, fleet mix and scheduled demand data. The model results also indicate the influence of different variables on the capacity of the airport. Model performance is compared using several validation measures, including Integrated Brier Score (IBS), Concordance Index (C Index), R2 and RMSE of predicted throughput, that account for the presence of censored observations. The RSF model consistently outperforms the KM estimator and Cox model across all the validation measures. This dissertation also develops capacity based similarity metrics between two days using the predicted hourly capacity distributions. The evaluation of the estimated similarity metric is challenging owing to the lack of ground truth similarity measures. In this dissertation, we propose a framework to validate the estimated similarity metric between two days using predicted capacity CDFs, demand, TMI and operational outcomes data. The assumption for this framework is that days that are similar based on their capacity, demand and TMI features should be similar based on their operational outcomes. We use a Random Forest model to combine the capacity, demand and TMI based similarity metrics, supervised by operational outcomes similarity metric. This combined similarity matrix is evaluated by measuring its correlation with the operational outcomes similarity matrix on test data. The methodology developed in this dissertation to identify similar days can be extended to any airport around the world using their respective weather, demand, TMI and operational outcome data. This framework uses data that can be forecasted and thus can be used to guide decisions on a day-of-operations in order to guide decision making, as well as a post-operations setting to compare decisions and outcomes on similar historical days.