Infrastructure Engineering - Theses

Permanent URI for this collection

http://hdl.handle.net/11343/348

Search Results

Now showing 1 - 10 of 476

Automated Error Correction in Volunteered Geographic Information Map Databases

Chittor Sundaram, Rajesh ( 2023-06)

Data provides indisputable evidence as opposed to assumptions or abstract observations. While data facilitates informed decision making, improved data quality leads to making better decisions. While there is no dearth in the availability of data, the quality of data leaves a lot to be desired. This thesis addresses Data Quality Augmentation in spatial data sets through Missing Value Imputation (MVI). It focuses on improving data quality in Volunteered Geographic Information (VGI) data sets, which often suffer from issues of poor data quality and thereby hindering their widespread adoption. Poor data quality, often characterized as Dirty Data, results in significant costs and decision-making challenges. These issues include incorrect or missing values, outliers, non-standard representations, and duplicate values, affecting the usability and reliability of such data. MVI is a popular strategy to address data quality issues, primarily applied and discussed in the realm of numerical data sets. The hypothesis of this research is that ‘Spatial data characteristics can facilitate Missing Value Imputation in VGI datasets’. The hypothesis centers on harnessing spatial data characteristics for performing MVI in VGI data sets. The thesis begins by identifying data quality pain points in VGI data through the analysis of the FIXME tags in OpenStreetMap (OSM) data as a case study. This approach leverages an intrinsic data quality assessment framework built using ISO-19157 quality indicators and Topic Models to expose critical data quality insights from the unstructured text corpus of over 1.5 million FIXME records. These insights provide focus areas for targeted data cleansing activities. By understanding OSM data quality issues, the framework lays the foundation for automated error correction algorithms. One of the key findings is the incompleteness of address attributes in OSM. The research introduces the Membership Imputation Algorithm (MIA) next, which imputes missing ‘Nominal Attribute Values’ by taking the case study of OSM Address Attribute ‘Street Name’, thereby enhancing data completeness and quality in spatial databases. MIA effectively utilizes the spatio-temporal characteristics of spatial data, achieving high imputation accuracy, particularly in challenging spatial contexts. The algorithm’s applicability across various geographical regions makes it a suitable solution for handling missing data in OSM. The thesis further contributes an Ordinal Imputation Framework (OIF) comprising of a Street Numbering Pattern Classifier and the Ordinal Imputation Algorithm. OIF focuses on imputing missing ‘Ordinal Attribute Values’ by taking the case study of OSM address attribute ‘House Number’. By harnessing spatial separation measures in spatial data sets, OIF successfully demonstrates the potential of spatial data characteristics to drive ordinal data imputation and contributing to the overarching goal of enhancing data quality in OSM data sets, and thereby making a significant contribution to improving VGI data completeness and quality. In summary, the research findings presented in this thesis enhance VGI usability and facilitates a better adoption of these data sets to benefit a wide range of end user applications and informed decision-making.
Relative Permeability Upscaling in the Presence of Meso-Scale Heterogeneity

ELSAIEED, Abdallah Abdrabelnaby Youssef ( 2023-06)

During carbon geo-sequestration, the brine-CO2 distribution in heterogeneous rocks varies with viscous, gravitational, and capillary forces. There is mounting evidence that flow-rate dependent saturation functions are needed to model differences between capillary limit (CL) and viscous limit (VL) multiphase flow. This thesis explores functional forms that constrain these variations across scales. Specifically, four topics are investigated: 1) development of an upscaling framework for estimating equivalent REV-scale parameters at field conditions; 2) formalizing new anisotropic rate-dependent functions; 3) modelling the composite influence of nested heterogeneity from the core plug- to REV scale, and 4) numeric simulation of CO2 plume migration in aquifers as influenced by model resolution. The thesis studies geological heterogeneity at two length scales, and in a sequential manner such that the behaviour contained on the small-scale can be considered on the larger scale. At the small cm-scale, the thesis examines relative permeability upscaling for two heterogeneity features motivated by observations from the fluvio-deltaic Parraatte formation at the CO2CRC’s Otway International Test Centre (OITC), Australia. The composite rocktypes are grouped into two sets: 1) non-communicating layered media, and 2) communicating media characterized by weak to moderate capillary heterogeneity. At the larger scale, my research targets the upscaling of relative permeability to the geo-statistical permeability REV scale while considering the small-scale heterogeneity simultaneously. For the non-communicating layered media, the thesis introduces a new semi-analytical solution for estimating dynamic relative permeability. The approach exploits my new exact solution of the Dykstra-Parsons’ problem with a shift in the residual water saturation reflecting the maximum micro-displacement efficiency attained practically. This analysis confirms the promotion of instabilities at the inter-layer scale by flow barriers. Weak to moderate capillary heterogeneity in communicating porous media is modeled through a new workflow developed to estimate effective saturation functions. For the first time, the concept of variable-oriented pressure field is applied to match flow conditions in the field. The results show that the effective parameters are not only rate-dependent as known from 1D models, but rather, directionally rate-dependent. The analysis highlights the importance of accurate description of the internal structure of small to medium scale geologic heterogeneity. The tensorial-rate-dependent reflecting sub-seismic scale heterogeneity are upscaled to the tens of meter grid-block scale. The thesis employs the novel rate-dependent, directional functions to field data-based numeric-simulation to estimate upscaled parameters at the permeability REVs windows sampled from OITC. This upscaling is conducted by simulating the passage of CO2-water flow along a high-resolution (0.05 m 1 m) transect between two wells. The analysis delivers tensor-type rate-dependent relative permeability curves accounting for buoyancy-driven flow. My thesis also presents an analysis of the impact of vertical grid resolution on CO2 plume migration. The results show that insufficient resolution suppresses the ability of the model to capture the true mobility of the plume, leading to an artificially high sweep efficiency.
Development and evaluation of an event-based model incorporating updating for ensemble flood forecasting

Bahramian, Katayoon ( 2023-06)

Flood warnings play a crucial role in ensuring the safety of vulnerable communities by providing timely information about potential flood risks. However, when it comes to actively managing floods, such as the operation of flood gates to control water flows, it becomes imperative to quantify the uncertainty associated with the timing and magnitude of the forecasts. This allows decision-makers to assess various management strategies, particularly when they face significant disparities in the outcomes resulting from overestimation and underestimation. Meteorological inputs and the pre-storm wetness of the catchment are two significant sources of uncertainty that need to be considered when providing flood forecasts as they have a large influence on the accuracy and reliability of the forecasts. The methods available to consider the propagation of uncertainty due to antecedent catchment wetness depend heavily on the nature of the selected hydrological model. A wide user-base of practitioners utilise event-based hydrological models for flood forecasting due to their operational benefits, which include simplicity, low computational burden, and compatibility with other systems. In contrast to continuous hydrological models that explicitly account for changes in catchment wetness before significant rainfall events, event-based models lack the capability to directly estimate the antecedent condition of the catchment. As a result, these models rely on external specifications of event-based loss parameters to account for this deficiency. The research provides an in-depth examination of the performance of an event-based flood forecasting model implemented within a Monte Carlo simulation framework. It evaluates how uncertainties associated with rainfall depth, spatio-temporal distribution, and initial catchment wetness conditions affect the performance of the forecasts by assessing the relative contribution of major sources of uncertainty on the reliability and accuracy of the flood forecasts. The findings suggest that by including available independent estimates of catchment wetness, uncertainties caused by loss parameters can be reduced. Additionally, the study reveals that the average forecast rainfall depth has a more significant influence on uncertainty than forecast losses and spatio-temporal rainfall patterns. These results have important implications as they address the challenges associated with utilizing event-based models for flood forecasting.
Online Interpersonal Conflict Management in Open Peer Production Community

Choe, Youjin ( 2023-06)

Online interpersonal conflict has increased in OpenStreetMap (OSM) in recent years. Such conflict may negatively affect OSM data quality, member morale, and community operations. However, a systematic assessment of the conflict in OSM is lacking in the literature. Inspired by literature from human-computer interaction, organizational management, and linguistics, this thesis assesses online interpersonal conflict in OSM to improve OSM member experience in managing interpersonal conflict. This thesis suggests a holistic approach to interpersonal conflict analysis and management in OPPCs, using OSM as a case study. Findings from this thesis show that (1) interpersonal conflict in OSM occurs due to disagreements among sub-groups and uncooperative online behaviors during the collaboration process, (2) OSM members use multiple conflict management approaches which have different levels of effectiveness, and that (3) online discussion interface in OSM, where the most interpersonal conflict in OSM is manifested, could be improved to support better interpersonal conflict management. This is the first documented case of applying conflict management theory to interpersonal conflict management in an OPPC.
Improving Demolition Waste Management by Developing a Building Information Modelling (BIM)-based Sustainability Assessment Framework

Han, Dongchen ( 2023-09)

Construction and demolition waste (C&DW) arising from the rapid urbanisation process contributes to over 30% of global waste generation. Heterogeneous sustainability problems, including land degradation, natural resource depletion, and global warming, are intertwined with C&D activities and pertaining treatment procedures of the by-products. Compared to residual construction materials, demolition waste (DW) discharged from the building’s end-of-life (EoL) stage is greater in volume and more laborious to salvage. Extant literature suggests that decision-makers tend to prioritise the economic feasibility of C&DW management over its environmental and societal implications. Therefore, designing a comprehensive assessment framework for demolition waste management (DWM) is paramount for promoting sustainability-oriented DWM planning. The knowledge gap is identified as lacking a systematic sustainability assessment framework for assessing and prioritising the sustainability of different DWM schemes at the project level. Therefore, this study aims to develop a sustainability assessment framework to facilitate sustainability-oriented decision-making for the effective implementation of DWM. To this end, this study employed exploratory sequential mixed methods in a modified Delphi study to identify multifaceted indicators for appraising the sustainability performance of DWM schemes. The sustainability indicators derived from extant literature were subject to two rounds of validation, and their weights were obtained using the Analytic Hierarchy Process (AHP) method. A total of eight indicators were selected as the evaluation criteria of DWM schemes, including Global Warming Potential (GWP), energy efficiency (EE), land use (LU), acidification potential (AP), abiotic depletion potential (ADP), total cost (TC), landfill cost saving (LCS), and human toxicity (HT). The investigation uncovers the essential components of constructing a custom sustainability assessment framework based on the regional context and local industry practitioners’ preferences to prioritise sustainable DWM schemes. Furthermore, this study seeks to integrate this novel framework into the real-life DWM planning process for sustainability benchmarking. The lack of data interoperability often hinders the efficiency of data mapping and exchange between conventional design software and sustainability analysis tools. In order to enhance the applicability of sustainability-oriented DWM planning by improving data interoperability, this research leverages Building Information Modelling (BIM) as a collaborative data repository for managing the diverse information required for conducting the sustainability assessment. However, the BIM properties for storing essential data are yet to be fully accommodated into the BIM environment. Therefore, this study extended the BIM properties using an open data schema-Industry Foundation Classes (IFC). Moreover, to integrate this sustainability assessment framework into the actual DWM planning workflow, this study developed a Revit-based tool called ‘BIM-based visual DWM planning system’ that can perform inventory analysis, parallel DWM scenario comparison, sustainability score ranking, and recycling value visualisation within the BIM platform. In summary, this research first developed a methodological workflow for identifying and prioritising essential indicators of a DWM sustainability assessment framework based on local contexts. Second, by enriching IFC properties to integrate sustainability and DWM information into the BIM environment, the data mapping and exchange were streamlined for conducting the BIM-based inventory analysis. Thirdly, this research developed a BIM-based visual DWM planning tool to facilitate the interpretation and communication of the sustainability assessment results by adopting Multi-Criteria Decision-Aiding (MCDA) methods and Dynamo visual scripting. Finally, a case study project was used to demonstrate the applicability of the developed BIM-based prototype in the actual DWM planning process. Overall, this study contributes to sustainable development within the Architectural, Engineering, and Construction (AEC) industry by developing a framework to facilitate sustainability-oriented decision-making for DWM planning, thus enhancing the resource efficiency and cost-effectiveness of DWM.
Assessing the contribution of solar-induced fluorescence (SIF) to the estimation of nutrient status in almond orchards

Wang, Yue ( 2023-07)

Macro- and micro-nutrients are essential for plants to function efficiently, resist disease, and produce high yields and quality fruits. These nutrients are involved in various aspects of almond growth and development throughout the phenological cycle. High levels of nitrogen, phosphorus, and potassium are the most important inputs for almond production. Micro-nutrients, although needed at much lower levels, also play an important role in supporting growth, especially in key tissues. The most important aspect of fertilizer management is balancing the fertilizer program in order to maximize yields while minimizing environmental impacts. In precision agricultural management, a precise assessment of nutrient status is crucial to determine the optimal application of fertilizers. The traditional method of assessing nutrients is tissue testing in biochemical laboratories, but this is not cost- or time-effective for continuous monitoring over a large area. The use of remote sensing techniques has been explored in recent decades as a method of obtaining indicators for those nutrients, most notably nitrogen, in terms of their spatial orientation, efficiency, and rapidness. In remote sensing of leaf N assessment, empirical algorithms using chlorophyll a+b (Cab) sensitive vegetation indices, as well as radiative transfer model (RTM) inversion of plant pigments, are applied. In recent years, advances in leaf N estimation have relied on the assessment of leaf biochemistry and spectral characteristics linked to photosynthesis, such as solar-induced fluorescence (SIF), which has been demonstrated to be an indicator of stress caused by nutrient deficiencies in a wide range of crop species. As a result of the sensitive nature of SIF and the complexity of tree orchard canopy architecture, its performance and sensitivity to plant conditions need to be evaluated in tree-structured almond orchards. In spite of this, there is still a lack of understanding of proxies for other macro- and micro-nutrients and their interactions, an area that requires further investigation. This research investigates the response of spectral-based plant parameters to different nutrient elements in almond trees at both the leaf and canopy levels. It is intended that this study not only provides an improved assessment of N using a combination of robust proxies, but also it examines its evaluation at various spatial and spectral resolutions, from high-resolution airborne to coarser-resolution spaceborne platforms. The results from two years of data indicate that chlorophyll fluorescence can serve as a reliable proxy for the primary macro-nutrients (i.e., N, P, and K) across the two years, yielding r2 = 0.74 (p-values < 0.005) for both leaf steady-state measurements and canopy SIF with leaf N. Moreover, the biochemical constituents derived from radiative transfer modeling exhibited strong correlations with the primary macro-nutrients for both years, whereas vegetation indices exhibited generally inferior relationships with nutrients. Taking leaf N as an example, SIF and Cab derived from RTM inversion were found to be the most significant non-collinear indicators at both the airborne (0.4 m) and spaceborne (30 m) scales. An airborne-based model predicted field-measured leaf N with an r2 of 0.95 and RMSE of 0.05% over the course of two years. The newly developed spectrometer DESIS onboard the International Space Station (ISS) provided a model with an r2 of 0.83 and RMSE of 0.06% in 2021, while Sentinel-2 provided an inferior result (r2 = 0.72, RMSE = 0.08%). An emphasis has been placed in this research on the importance of Cab, SIF, and other plant pigments in determining the nutrient status of discontinuous tree-structured almond orchards. Moreover, this work provides a step forward towards achieving accurate and large-scale nutrient monitoring in precision agriculture.
Indoor LiDAR relocalization, drift-free odometry and building interior change detection using a 3D model

Zhao, Hang ( 2023-06)

LiDAR localization and mapping systems are widely studied for practical applications such as building management, emergency response and unknown place exploration. Such systems usually use LiDAR odometry or simultaneous localization and mapping (SLAM) algorithms to continuously estimate the pose of the equipped devices such as the LiDAR scanner. These algorithms face several challenges: pose initialization, drift of the trajectory, and recovery from failure. To overcome these challenges, this thesis makes four important contributions. The first contribution of this thesis is MoLi-PoseNet, a novel LiDAR relocalization method using a surface-based 3D model. Unlike conventional LiDAR relocalization methods, where the LiDAR pose is estimated with respect to a map created by the same LiDAR, which may be inaccurate due to possible localization failures, the proposed approach takes advantage of a 3D model, which can be easily extracted from a BIM or created from floor plans. In MoLi-PoseNet, synthetic LiDAR scans are generated in the 3D model and then used to train a pose regression network. The trained network performs pose estimation on real LiDAR scans. Experimental evaluation of MoLi-PoseNet shows that MoLi-PoseNet can achieve meter-level relocalization accuracy in large indoor environments with low-level-of-detail models. The second contribution of this thesis is MoLO: a novel drift-free LiDAR odometry method. Current LiDAR odometry approaches suffer from drift in the estimated trajectory and these methods rely on loop closure and optimization to eliminate the drift. However, loops are not practical for the navigation tasks. In MoLO, the acquired LiDAR scans are registered with a 3D model to perform relocalization at a certain frequency and pose graph optimization is then implemented to optimize the trajectory. Experimental evaluation of MoLO shows that MoLO can eliminate drift and achieve real-time localization while providing an accuracy equivalent to loop closure optimization. The proposed MoLi-PoseNet and MoLO rely on the accuracy of the 3D model, but building changes are not always reflected in the 3D model. To address this problem, the third contribution of the thesis is a geometry-based change detection method based on entropy derived from LiDAR data and a 3D model. The geometry-based change detection method eliminates the need for feature extraction from LiDAR data and the training of a network. A sequence of real LiDAR scans is acquired with a static LiDAR scanner and the pose of the LiDAR scanner for each scan is then estimated. Synthetic LiDAR scans are generated with the pose of the LiDAR scanner using the 3D model. The real LiDAR scans and synthetic LiDAR scans are sliced horizontally with a certain angular interval and the entropy of all slices of LiDAR scans is then calculated. The differenced entropy between corresponding slices of real and synthetic LiDAR scans is calculated for the classification of the changes into four categories: unchanged, moving objects, structural change, and non-structural change. Experimental evaluation of the method shows that unchanged slices and slices containing moving objects can be accurately detected, achieving 100% accuracy while non-structural and structural changes are detected with an accuracy of 98.5% and 86.3% respectively. The fourth contribution of the thesis is a learning-based method based on LiDAR segmentation to perform real time change detection. Synthetic LiDAR scans are generated from a modified 3D model and real LiDAR scans are acquired from the real environment. A change detection network with two branches for synthetic LiDAR scans and real LiDAR scans respectively is built. The pairs of synthetic LiDAR scans and real LiDAR scans are used to train the network. The trained network can perform change detection in a comparable environment. Each point in real LiDAR scans is classified into one of the four change categories. Experimental evaluation of the method shows that the proposed approach can achieve 94% overall change classification accuracy with the SqueezeNet-based change detection network and the trained network is transferable to comparable indoor environments.
Evaluating the impact of the spectral configuration of airborne hyperspectral imaging sensors on the accurate estimation of solar-induced chlorophyll fluorescence (SIF)

Belwalkar, Anirudh ( 2023-08)

Climate change has devastated agriculture and food production. In recent decades heatwaves and droughts have made it harder to meet global food demand. Understanding photosynthesis, plant adaptation, and how photosynthetic efficiency affects crop yields is essential for developing stress-resistant plants and increasing crop production. This climate crisis underlines the need for systems that evaluate photosynthetic efficiency to monitor plant health and increase efficiency. Solar-induced chlorophyll fluorescence (SIF) is a faint electromagnetic signal that can indicate plant stress and photosynthetic efficiency. Accurate SIF quantification requires sub-nanometer resolution sensors. However, sub-nanometer resolution imaging sensors onboard airborne platforms are expensive and difficult to operate, hindering their widespread operational use for plant phenotyping, stress detection, and precision agriculture applications. Consideration should therefore turn towards development of adequate airborne imaging sensors and approaches that use physically-based models to accurately interpret SIF from the sensor. This PhD thesis investigates whether commonly accessible narrow-band imaging sensors could potentially substitute for sub-nanometer imaging sensors in operational SIF retrieval for plant phenotyping, stress detection, and precision agriculture applications. A narrow-band imaging sensor and a sub-nanometer imaging sensor flown in tandem were compared for SIF. Physically-based models and machine learning were used to model the effect of spectral resolution (SR) on narrow-band far-red SIF (SIF760) estimates. Furthermore, an exploratory analysis was conducted to investigate the potential of solar Fraunhofer lines in the SIF emission region for estimating leaf nitrogen concentration across a field and detecting biotic stress in infected trees, using airborne sub-nanometer hyperspectral imagery. Airborne SIF760 retrievals from a narrow-band imaging sensor (5.8-nm FWHM) and a sub-nanometer imaging sensor (0.2-nm FWHM) were compared across two wheat and maize phenotyping trials grown under varied nitrogen fertiliser rates over the 2019–2021 growing seasons. The correlation between SIF760 values obtained from the two sensors was found to be significant (R2 = 0.77–0.9, p < 0.01). Notably, the narrow-band imager yielded higher estimates of SIF760 than the sub-nanometer imager did (root-mean-square error (RMSE) 3.28–4.69 mW/m2/nm/sr). The findings of this study suggest that narrow-band imaging sensors may accurately detect field-wide variations in relative SIF760, particularly when nitrogen fertilisation levels vary. The next part of this study focused on improving narrow-band-derived absolute SIF760 levels to reduce systematic bias. A Soil Canopy Observation, Photochemistry, and Energy fluxes (SCOPE) model with Support Vector Regression (SVR) scaling airborne narrow-band SIF760 values to 1-nm FWHM was used. As shown by the normalised RMSE values of 2.45–5.28% for the SCOPE simulated dataset and 4.5–16% for the airborne narrow-band hyperspectral dataset, the estimated SIF760 at 1-nm FWHM showed good agreement with the reference SIF760. This study suggests that the proposed SIF760 modelling approach can improve the understanding of relative SIF760 levels quantified by narrow-band hyperspectral imaging sensors in stress detection and plant physiological monitoring applications. An exploratory investigation of sub-nanometer imagery-derived Fraunhofer lines (FLs) concludes the thesis. The study found that including depths for two FLs near oxygen absorption features improved leaf nitrogen estimation (0.15 increment in R2 and 1% decrement in normalised RMSE). In addition, it was observed that biotic-induced stress caused due to Verticillium dahliae (Vd) infection was linked to FL activation in the red and far-red spectral regions. As biotic-induced stress level increased, the sensitivity of FLs in the discernment of and differentiation between symptomatic and asymptomatic trees also increased. These findings indicate the need for additional research on these specialised potential benefits of FLs, which would allow better understanding and more efficient management of the various factors that affect the physiological status of plants.
Integration and Mixed Reality Visualisation of Spatial Data and Models of Buildings

Radanovic, Marko ( 2023-07)

This thesis is inspired by two limitations related to the building information modelling of heritage and other existing buildings and the visualisation of these models. The first limitation is that, despite the great potential of Building Information Models (BIMs) for building life-cycle management, creating detailed BIMs of heritage or other existing buildings is a major challenge. To address this, a concept of multilayered documentation of buildings is proposed, which successfully integrates several types of spatial datasets into an interconnected multilayered model. The multilayered model offers high visual fidelity in the foreground while keeping the semantic and other information in the background within a visually simplified BIM, thus offering a major advantage over the existing building modelling approaches where there is always a trade-off between visual fidelity and semantic richness. The second limitation is that superimposing these models and other spatial data accurately onto the real building using Mixed Reality (MR) visualisation is not possible with current localisation methods, which inhibits large-scale MR applications such as facility management, navigation, or enhanced exhibitions. The present thesis addresses three key aspects of this limitation, which are (i) large-scale global localisation in MR for initial localisation within the indoor environment and initial alignment of the BIM to reality, (ii) large-scale continuous tracking in MR to maintain this alignment and (iii) the empirical evaluation of the localisation accuracy in terms of the alignment of the BIM to the real building. This thesis proposes a large-scale global MR localisation method that utilises recent advances in learning-based 3D-3D model registration to localise by automated registration of a low-density model created by the device to the existing model of the environment. A comprehensive evaluation of the approach within a real-world setting using a developed prototype demonstrates high reliability and localisation accuracy of 2.8 cm and 0.30 degrees. The approach offers considerable advantages over the existing global localisation methods dominated by image-based approaches, namely, it is shown to work in large-scale environments (300 sqm), it demonstrates excellent robustness to scene geometry changes, and is unaffected by changes in illumination. Next, an end-to-end trainable Convolutional Neural Network (CNN) is proposed to keep the initial alignment of the BIM and the real building. The network takes a pair of images, one real image and one synthetic BIM image, as input and regresses the 6 DoF relative camera pose difference between them directly and, unlike the existing model-based methods, without iterative and expensive re-rendering. The method is extensively tested and demonstrates a high localisation accuracy of 7.0 cm and 0.9 degrees, which is empirically shown to correct for drift and significantly improve the alignment of the BIM to the real building. Furthermore, the method shows high robustness to the domain shift between real and synthetic images and can be used in near-real time on the current generation of hardware. Finally, this thesis determines the relationship between localisation accuracy and the alignment of the BIM to the real building. A fully 3D geometrical approach is proposed, where real elements of different sizes are fitted to point clouds of their virtual counterparts to empirically determine their element retrieval success rate. The results of the experiment show a high correlation between the element retrieval success rate and the localisation accuracy. The localisation accuracy of 1.4 cm and 0.24 degrees is determined to enable retrieving objects as small as 4 cm in dimension with a 100% reliability, with the reliability dropping to 70, 50 and 0% for 3, 2 and 1 cm objects, respectively.
Information sensing, transmission and management between connected vehicles

Hu, Wenyan ( 2023-07)

The concept of connected vehicles is gaining momentum in the research communities thanks to the development of wireless communication technologies. Due to their ubiquity, mobility and connectivity, connected vehicles have the potential to not only revolutionise the transportation industry, but also be exploited for information sensing and transmission, as they are equipped with a wide range of sensors in addition to the wireless communication devices. Despite having the power supply and computing capacity to continuously host such sensors, the unpredictability and disorganisation of connected vehicles are presented as challenges for collecting and sharing information. Therefore, the aim of this research is to explore the role, usage and methodology of connected vehicles for information sensing, transmission and management. With regard to information sensing through connected vehicles in urban areas, this study investigates sensor network deployment, vehicle selection, and route assignment with the aim of collecting data that are appropriate for the task at hand. First, a framework is proposed to optimise the configuration of stationary sensors and opportunistic vehicular sensors in hybrid sensing for a given sensing task to improve sensing coverage. After knowing the number of vehicles needed for the sensing task, the next step is to select the proper vehicles. To optimise sensing coverage, a vehicle selection framework is proposed that integrates a model of forecasting fine-grained sensing coverage through coarse-grained information about candidate vehicles and the genetic algorithm. Sensing coverage mainly relies on the trajectories of these selected vehicles. Therefore, in order to further reduce the sensing overlap that exists between selected vehicles, an activity-based route assignment strategy integrates the sensing task requirements and the planned activities of the selected vehicles to assign routes that would not interfere with these activities. The implementation of these proposed approaches yields a win-win outcome for both the task initiators and the participants. Once a connected vehicle has sensed anomalous information, e.g., an incident, it can share the information to other connected vehicles via peer-to-peer networks without centralised infrastructures. In the context of transmitting information about ephemeral incidents in traffic, two event-driven models that transmit and manage transient information in vehicular networks are developed in this study. First, the time-geography framework is used to offer a decentralised transmission model that takes into account the transient nature of traffic incidents. Furthermore, a model for managing traffic incident information is put forth to improve traffic efficiency when traffic incidents occur by managing information regarding their range, timely updating outdated information in vehicular networks, and guiding traffic for affected vehicles. Experimental results show that these two proposed models can reduce not only invalid broadcasts but also incident-induced traffic congestion. Overall, the potential of connected vehicles for information sensing, transmission and management is explored in this thesis. The findings of this study demonstrate that the proposed approaches are capable of increasing sensing coverage so that the information sensed is appropriate for intended uses, and improving the efficiency of information sharing while requiring less broadcast and shortening the duration of invalid broadcast.

Infrastructure Engineering - Theses

Permanent URI for this collection

Filters

Date

Author

Subject

Type

Settings

Sort By

Results per page

Statistics

Citations

Search Results