Researchers have examined crowd behaviour in the past by employing a variety of methods including ethnographic studies, computer vision techniques and manual annotation-based data analysis. However, because of the resources to collect, process and analyse data, it remains difficult to obtain large data sets for study. Mobile phones offer easier means for data collection that is easy to analyse and can preserve the user's privacy. The aim of this thesis is to identify and model different qualities of social interactions inside crowds using mobile sensing technology. This Ph.D. research makes three main contributions centred around the mobile sensing and crowd sensing area. Firstly, an open-source licensed mobile sensing framework is developed, named SensingKit, that is capable of collecting mobile sensor data from iOS and Android devices, supporting most sensors available in modern smartphones. The framework has been evaluated in a case study that investigates the pedestrian gait synchronisation phenomenon. Secondly, a novel algorithm based on graph theory is proposed capable of detecting stationary social interactions within crowds. It uses sensor data available in a modern smartphone device, such as the Bluetooth Smart (BLE) sensor, as an indication of user proximity, and accelerometer sensor, as an indication of each user's motion state. Finally, a machine learning model is introduced that uses multi-modal mobile sensor data extracted from Bluetooth Smart, accelerometer and gyroscope sensors. The validation was performed using a relatively large dataset with 24 participants, where they were asked to socialise with each other for 45 minutes. By using supervised machine learning based on gradient-boosted trees, a performance increase of 26.7% was achieved over a proximity-based approach. Such model can be beneficial to the design and implementation of in-the-wild crowd behavioural analysis, design of influence strategies, and algorithms for crowd reconfiguration.

Large-scale spatial data such as air quality of a city, biomass content in a lake, Wi-Fi Received Signal Strengths (RSS, also referred as fingerprints) in indoor spaces often play vital roles to applications like indoor localization. However, it is extremely labor-intensive and time-consuming to collect those data manually. In this thesis, the main goal is to develop efficient means for large-scale spatial data collection. Robotic technologies nowadays offer an opportunity on mobile sensing, where data are collected by a robot traveling in target areas. However, since robots usually have a limited travel budget depending on battery capacity, one important problem is to schedule a data collection path to best utilize the budget. Inspired by existing literature, we consider to collect data along informative paths. The process to search the most informative path given a limited budget is known as the informative path planning (IPP) problem, which is NP-hard. Thus, we propose two heuristic approaches, namely a greedy algorithm and a genetic algorithm. Experiments on Wi-Fi RSS based localization show that data collected along informative paths tend to achieve lower errors than that are opportunistically collected. In practice, the budget of a mobile robot can vary due to insufficient charging or battery degradation. Although it is possible to apply the same path planning algorithm repetitively whenever the budget changes, it is more efficient and desirable to avoid solving the problem from scratch. This can be possible since informative paths for the same area share common characteristics. Based on this intuition, we propose and design a reinforcement learning based IPP solution, which is able to predict informative paths given any budget. In addition, it is common to have multiple robots to conduct sensing tasks cooperatively. Therefore, we also investigate the multi-robot IPP problem and present two solutions based on multi-agent reinforcement learning. Mobile crowdsourcing (MCS) offers another opportunity to lowering the cost of data collection. In MCS, data are collected by individual contributors, which is able to accumulate a large amount of data when there are sufficient participants. As an example, we consider the collection of a specific type of spatial data, namely Wi-Fi RSS, for indoor localization purpose. The process to collect RSS is also known as site survey in the localization community. Though MCS based site survey has been suggested a decade ago~\cite{park2010growing}, so far, there has not been any published large-scale fingerprint MCS campaign. The main issue is that it depends on user's participation, and users may be reluctant to make a contribution. To investigate user behavior in a real-world site survey, we design an indoor fingerprint MCS system and organize a data collection campaign in the McMaster University campus for five months. Although we focus on Wi-Fi fingerprints, the design choices and campaign experience are beneficial to the MCS of other types of spatial data as well. The contribution of this thesis is two-fold. For applications where robots are available for large-scale spatial sensing, efficient path planning solutions are investigated so as to maximize data utility. Meanwhile, for MCS based data acquisition, our real-world campaign experience and user behavior study reveal essential design factors that need to be considered and aspects for further improvements. / Thesis / Doctor of Philosophy (PhD) / A variety of applications such as environmental monitoring require to collect large-scale spatial data like air quality, temperature and humidity. However, it usually incurs dramatic costs like time to obtain those data, which is impeding the deployment of those applications. To reduce the data collection efforts, we consider two mobile sensing schemes, i.e, mobile robotic sensing and mobile crowdsourcing. For the former scheme, we investigate how to plan paths for mobile robots given limited travel budgets. For the latter scheme, we design a crowdsourcing platform and study user behavior through a real word data collection campaign. The proposed solutions in this thesis can benefit large-scale spatial data collection tasks.

The ever increasing capabilities of mobile devices such as smartphones and their ubiquity in daily life has resulted in a large and interesting body of research into context awareness { the `awareness of a situation' { and how it could make people's lives easier. There are, however, diculties involved in realising and implementing context aware systems in the real world; particularly in a mobile environment. To address these diculties, this dissertation tackles the broad problem of designing and implementing mobile context aware systems in the eld. Spanning the elds of Articial Intelligence (AI) and Human Computer Interaction (HCI), the problem is broken down and scoped into two key areas: context sensing and interactive intelligence. Using a simple design model, the dissertation makes a series of contributions within each area in order to improve the knowledge of mobile context aware systems engineering. At the sensing level, we review mobile sensing capabilities and use a case study to show that the everyday calendar is a noisy `sensor' of context. We also show that its `signal', i.e. useful context, can be extracted using logical data fusion with context supplied by mobile devices. For interactive intelligence, there are two fundamental components: the intelligence, which is concerned with context inference and machine learning; and the interaction, which is concerned with user interaction. For the intelligence component, we use the case of semantic place awareness to address the problems of real time context inference and learning on mobile devices. We show that raw device motion { a common metric used in activity recognition research { is a poor indicator of transition between semantically meaningful places, but real time transition detection performance can be improved with the application of basic machine learning and time series processing techniques. We also develop a context inference and learning algorithm that incorporates user feedback into the inference process { a form of active machine learning. We compare various implementations of the algorithm for the semantic place awareness use case, and observe its performance using a simulation study of user feedback. For the interaction component, we study various approaches for eliciting user feedback in the eld. We deploy the mobile semantic place awareness system in the eld and show how dierent elicitation approaches aect user feedback behaviour. Moreover, we report on the user experience of interacting with the intelligent system and show how performance in the eld compares with the earlier simulation. We also analyse the resource usage of the system and report on the use of a simple SMS place awareness application that uses our system. The dissertation presents original research on key components for designing and implementing mobile context aware systems, and contributes new knowledge to the eld of mobile context awareness.

Loneliness and social isolation can have a serious impact on one’s mental health, leading to increased stress, lower self-esteem, panic attacks, and drug or alcohol addictions. Older adults and international students are disproportionately affected by loneliness. This thesis investigates Socialoscope, a smartphone app that passively detects loneliness in smartphone users based on the user’s day-to-day social interactions, communication and smartphone activity sensed by the smartphone’s built-in sensors. Statistical analysis is used to determine smartphone features most correlated with loneliness. A previously established relationship between loneliness and personality type is explored. The most correlated features are used to synthesize machine learning classifiers that infer loneliness levels from smartphone sensor features with an accuracy of 90%. These classifiers can be used to make the Socialoscope an intelligent loneliness sensing Android app. The results show that, of the five Big-Five Personality Traits, emotional stability and extraversion personality traits are strongly correlated with the sensor features such as number of messages, number of outgoing calls, number of late night browser searches, number of long incoming or outgoing calls and number of auto-joined trusted Wi-Fi SSIDs. Moreover, the classifier accuracy while classifying loneliness levels is significantly improved to 98% by taking these personality traits into consideration. Socialoscope can be integrated into the healthcare system as an early warning indicator of patients requiring intervention or utilized for personal self-reflection.

Recent advances in low-power electronics, energy harvesting, and sensor technologies are poised to revolutionize mobile and embedded computing, by enabling networks of mobile sensor devices that are long-lived and self-managing. When realized, this new generation of perpetual systems will have a far-reaching and transformative impact, improving scientists’ ability to observe natural phenomena, and enabling many ubiquitous computing applications for which regular maintenance is not feasible. In spite of these benefits, perpetual systems face many programming and deployment challenges. Conditions at runtime are unknown and highly variable. Variations in harvested energy and energy consumption, as well as mobility-induced changes in network connectivity and bandwidth require systems that are able to adapt gracefully at run-time to meet different circumstances. However, when programmers muddle adaptation details with application logic, the resulting code is often difficult to both understand and maintain. Relying on system designers to correctly reason about energy fluctuations and effectively harness opportunities for cooperation among mobile nodes, is not a viable solution. This dissertation demonstrates that perpetual systems can be designed and deployed without sacrificing programming simplicity. We address the challenges of perpetual operation and energy-aware data delivery in the context of several applications, including in situ wildlife tracking and vehicular networks. Specifically, we focus on two specific systems. Eon, the first energy-aware programming language, allows programmers to simply express application specific energy policies and then delegate the complexities of energy-aware adaptation to the underlying system. Eon automatically manages application energy in order to indefinitely extend a device’s operating lifetime, requiring only simple annotations from the programmer. The second system, Tula, is a system that automatically balances the inherently dependent activities of data collection and data delivery, while also ensuring that devices have fair access to network resources. In our experiments, Tula performs within 75% of the optimal max-min fair rate allocation.

<p>Modern sensing apps require continuous and intense computation on data streams. Unfortunately, mobile devices are failing to keep pace despite advances in hardware capability. In contrast to powerful system-on-chips that rapidly evolve, battery capacities merely grow. This hinders the potential of long-running, compute-intensive sensing services such as image/audio processing, motion tracking and health monitoring, especially on small, wearable devices. </p><p>In this thesis, we present three pieces of work that target at improving the energy efficiency for mobile sensing. (1) In the first work, we study heterogeneous mobile processors that dynamically switch between high-performance and low-power cores according to tasks' performance requirements. We benchmark interactive mobile workloads and quantify the energy improvement of different microarchitectures. (2) Realizing that today's users often carry more than one mobile devices, in the second work, we extend the resource boundary of individual devices by prototyping a distributed framework that coordinates multiple devices. When devices share common sensing goals, the framework schedules sensing and computing tasks according to devices' heterogeneity, improving the performance and latency for compute-intensive sensing apps. (3) In the third work, we study the power breakdown of motion sensing apps on wearable devices and show that traditional offloading schemes cannot mitigate sensing’s high energy costs. We design a framework that allows the phone to take over sensing and computation by predicting the wearable's sensory data, when motions of the two devices are highly correlated. This allows the wearable to offload without communicating raw sensing data, resulting in little performance loss but significant energy savings.</p> / Dissertation

<p>Environmental sensing is experiencing tremendous development due largely to the advancement of sensor technology and wireless technology/internet that connects them and enable data exchange. Environmental monitoring sensor systems range from satellites that continuously monitor earth surface to miniature wearable devices that track local environment and people's activities. However, transforming these data into knowledge of the underlying physical and/or chemical processes remains a big challenge given the spatial, temporal scale, and heterogeneity of the relevant natural phenomena. This research focuses on the development and application of novel sensing and inference techniques in air and water environments. The overall goal is to infer the state and dynamics of some key environmental variables by building various models: either a sensor system or numerical simulations that capture the physical processes.</p><p>This dissertation is divided into five chapters. Chapter 1 introduces the background and motivation of this research. Chapter 2 focuses on the evaluation of different models (physically-based versus empirical) and remote sensing data (multispectral versus hyperspectral) for suspended sediment concentration (SSC) retrieval in shallow water environments. The study site is the Venice lagoon (Italy), where we compare the estimated SSC from various models and datasets against in situ probe measurements. The results showed that the physically-based model provides more robust estimate of SSC compared against empirical models when evaluated using the cross-validation method (leave-one-out). Despite the finer spectral resolution and the choice of optimal combinations of bands, the hyperspectral data is less reliable for SSC retrieval comparing to multispectral data due to its limited amount of historical dataset, information redundancy, and cross-band correlation.</p><p>Chapter 3 introduces a multipollutant sensor/sampler system that developed for use on mobile applications including aerostats and unmanned aerial vehicles (UAVs). The system is particularly applicable to open area sources such as forest fires, due to its light weight (3.5 kg), compact size (6.75 L), and internal power supply. The sensor system, termed “Kolibri”, consists of low-cost sensors measuring CO2 and CO, and samplers for particulate matter and volatile organic compounds (VOCs). The Kolibri is controlled by a microcontroller, which can record and transfer data in real time using a radio module. Selection of the sensors was based on laboratory testing for accuracy, response delay and recovery, cross-sensitivity, and precision. The Kolibri was compared against rack-mounted continuous emission monitors (CEMs) and another mobile sampling instrument (the ``Flyer'') that had been used in over ten open area pollutant sampling events. Our results showed that the time series of CO, CO2, and PM2.5 concentrations measured by the Kolibri agreed well with those from the CEMs and the Flyer. The VOC emission factors obtained using the Kolibri are comparable to existing literature values. The Kolibri system can be applied to various open area sampling challenging situations such as fires, lagoons, flares, and landfills.</p><p>Chapter 4 evaluates the trade-off between sensor quality and quantity for fenceline monitoring of fugitive emissions. This research is motivated by the new air quality standard that requires continuous monitoring of hazardous air pollutants (HAPs) along the fenceline of oil and gas refineries. Recently, the emergence of low-cost sensors enables the implementation of spatially-dense sensor network that can potentially compensate for the low quality of individual sensors. To quantify sensor inaccuracy and uncertainty of describing gas concentration that is governed by turbulent air flow, a Bayesian approach is applied to probabilistically infer the leak source and strength. Our results show that a dense sensor network can partly compensate for low-sensitivity or high noise of individual sensors. However, the fenceline monitoring approach fails to make an accurate leak detection when sensor/wind bias exists even with a dense sensor network.</p><p>Chapter 5 explores the feasibility of applying a mobile sensing approach to estimate fugitive methane emissions in suburban and rural environments. We first compare the mobile approach against a stationary method (OTM33A) proposed by the US EPA using a series of controlled release tests. Analysis shows that the mobile sensing approach can reduce estimation bias and uncertainty compared against the OTM33A method. Then, we apply this mobile sensing approach to quantify fugitive emissions from several ammonia fertilizer plants in rural areas. Significant methane emission was identified from one plant while the other two shows relatively low emissions. Sensitivity analysis of several model parameters shows that the error term in the Bayesian inference is vital for the determination of model uncertainty while others are less influential. Overall, this mobile sensing approach shows promising results for future applications of quantifying fugitive methane emission in suburban and rural environments.</p> / Dissertation

Mobile sensing applications (MSAs) are an emerging class of applications that process continuous sensor data streams to make time-sensitive inferences. Representative application domains range from environmental monitoring, context-aware services to recognition of physical activities and social interactions. Example applications involve city air quality assessment, indoor localization, pedometer and speaker identification. The common application workflow is to read data streams from the sensors (e.g, accelerometers, microphone, GPS), extract statistical features, and then present the inferred high-level events to the user. MSAs in the healthcare domain especially draw a significant amount of attention in recent years because sensor-based data collection and assessment offer finer-granularity, timeliness, and higher accuracy in greater quantity than traditional, labor-intensive, data gathering mechanisms in use today, e.g., surveys methods. The higher fidelity and accuracy of the collected data expose new research opportunities, improve the reliability and accuracy of medical decisions, and empower users to manage personal health more effectively. Nonetheless, a critical challenge to practical deployment of MSAs in real-world is to effectively manage limited resources of mobile platforms to meet stringent quality of service (QoS) requirements in terms of processing throughput and delay while ensuring long term robustness. To address the challenge, we model MSAs in dataflows as a graph of processing elements that are connected by communication channels. The processing elements may execute in parallel as long as they have sufficient data to process. A key feature of the dataflow model is that it explicitly capture parallelism and data dependencies between processing elements. Based on the graph composition, we first proposed CSense, a stream-processing toolkit for robust and high-rate MSAs. In this work, CSense provide a simple language for developers to describe their sensing flow without the need to deal with system intricacy, such as memory allocation, concurrency control and power management. The results show up to 19X performance difference may be achieved automatically compared with a baseline using the default runtime concurrency and memory management. Following this direction, we saw the opportunities that MSAs can be significantly improved from the perspective of memory performance and energy efficiency in view of the iterative execution. Therefore, we next focus on optimizing the runtime memory management through compile time analysis. The contribution is a stream compiler that captures the whole program memory behavior to generate an efficient memory layout for runtime access. Experiments show that our memory optimizations reduce memory footprint by as much as 96% while matching or improving the performance of the StreamIt compiler with cache optimizations enabled. On the other hand, while there is a significant body of work that has focused on optimizing the throughput or latency of processing sensor streams, little to no attention has been given to energy efficiency. We proposed an accurate offline energy prediction model for MSAs that leverages the pipeline structure and iterative execution nature to search for the most energy saving batching configuration w.r.t. a deadline constraint. The developers are expected to visualize the energy delay trade-off in the parameter space without runtime profiling. The evaluation shows the worst-case prediction errors are about 7% and 15% for energy and latency respectively despite variable application workloads.

The thesis is to explore wireless location tracking algorithms based on geometric dilution of precision (GDOP) in the mobile environment. The GDOP can be used as an indication of positioning accuracy, affected by the geometric relationship between the target and sensing units. The smaller the GDOP is, the better positioning accuracy. By using the information of sensing units and time difference of arrival (TDOA) positioning method, we use extended Kalman filter as an estimator to track and predict the state of a moving target. From previous research, the lowest GDOP value, located at the center of a regular polygon, represents the best positioning accuracy in 2-D scenario with numerous sensing units. It is important to find the best locations for the sensing units. Simulated annealing algorithm was used in previous studies. However, it only finds a location at a time, and consumes computation load and time. Due to the above-mentioned reasons, we propose a location tracking system, which consists of a base traiver station and numerous mobile sensing units. By using the information of a base transceiver station and the predicted position of target, we can obtain the best locations for all the mobile sensing units with the calculation of rotation matrix. The locations can also be used as beacons for relocating mobile sensing units. It may take many cycles to move mobile sensing units to the best locations. We have to perform path planning for mobile sensing units. Due to the location change of the moving target, the routes need adjustment accordingly. If the predicted stay of a mobile sensing unit is inside the obstacle, we adjust the route of the mobile sensing unit to make it stay out of the obstacle. Therefore, we also propose a path planning scheme for mobile sensing units to avoid obstacles. Through simulations, the proposed method decreases the tracking time effectively, and find the best locations precisely. When mobile sensing units move toward the best locations, they successfully avoid obstacles and move toward the position with the minimum GDOP. Through the course, good positioning accuracy can be maintained.

長久以來，學者專家為進行各種與人們行為有關的實驗，使用了多種方法進行數據的收集，卻存在著一些缺點或是成本過高，不易大規模推行。近年來，隨著行動科技的高速發展，智慧型手機已經相當普及。由於智慧型手機內建眾多的感測器(Sensors)及裝置(Devices)，透過撰寫及執行特定的手機應用程式（APP），即可蒐集手機所偵測到的使用者行為相關之資訊，並加以分析處理。行動感測(Mobile Sensing)遂成為新型態的數據收集方式。但是，僅就特定之感測實驗所需數據項目，開發數據蒐集的行動應用程式，亦或是不管實驗需求，讓程式蒐集全部項目的感測數據，都不是理想的作法：不是欠缺彈性考量就是未能考慮手機使用者的隱私關切。 本研究實作一個通用於所有行動作業系統上之可設定式之行動感測服務平台，它可以協助研究者根據其需求自行設定感測實驗項目及其條件規範。之後，透過建立於伺服器端與客戶端之間特定的資料交換機制，進行實驗的發布及兩端的互動溝通。整體運作過程中，客戶端的參與者只需簡易地安裝一套行動應用程式，便能夠輕鬆參與進行各種感測實驗、貢獻實驗數據。最後，我們模擬幾項實驗，用以驗證平台之實際運作效能。 / Experimental data obtained in scientific research is extremely important. For a long time, people use a variety of methods for data collection, but most of them are either restricted or expensive. In recent years, with the rapid development of mobile technology, smart phones are becoming very popular. With many built-in sensors and devices, smart phones can be used as a new tool of data collection, hence the emergence of mobile sensing. By installing a mobile application (APP) on a user’s smart phone, researchers can collect those required sensor data from the user and analyze it for their study. This research presents a configurable service platform for mobile sensing which aims to reconcile the flexibility needed by researchers and privacy concerns of smart phone users. In particular, our platform allows researchers to use our GUI tool to easily set up an experiment by composing an experiment configuration file (ECF) which specifies the sensor types to collect and the filtering rules for data selection. Users of smart phones can join any experiments by installing a single piece of logger APP developed according to our ECF specification. Besides, users will be fully informed of the data to collect before agreeing to participate a specific experiment. In such a manner, we achieve a proper balance between flexibility and privacy. Finally, we conducted several experiments to validate the feasibility of our service platform with users of Android smart phones.