LiDAR is one of the most used sensors in many areas like robotics, self-driving cars, and advanced driving assistance systems due to providing an accurate point cloud of the surroundings. However, to cope with challenges in perceiving the environment around a vehicle, LiDAR data is often combined with data from other sensors. Thermal cameras can provide complementary information that can be beneficial, especially for detecting pedestrians and seeing at nighttime and in fog, dust, etc. In this paper, we propose an algorithm for the extrinsic calibration of a thermal camera and a LiDAR sensor in a vehicle. First, one or more thermal image-point cloud pairs of our designed calibration target are collected. Then line and plane equations of the target’s edges and plane in both data modalities are found. Finally, the algorithm uses lines and plane correspondences to cross-calibrate the sensors. The proposed method obtains good results with one or more poses. We also show that it works well with sparse LiDAR data. Several experiments are presented to illustrate the effectiveness of the method.

Academic Research Computing Clouds are widely deployed worldwide by research institutions to support researcher's computations. While the nature of the hosted use-cases is diverse, literature and institutions' published guides point to highly parallel HPC workloads and state-heavy workloads as two popular use-cases in research-computing clouds. Additionally, our prior investigation has uncovered unique patterns in the users' VM-provisioning behaviors in four of Canada's research-computing clouds. These patterns, i.e., usage-trends, were generated by examining nearly 1 million VMs created by researchers over four and half years. The usage trends mimicked behaviors of provisioning highly parallel HPC workloads and state-heavy portals. In this paper, we exploit the knowledge of these usage trends to guide VM placement decisions in research-computing IaaS clouds. We propose a delayed-provisioning algorithm that postpones resource allocations for incoming VM requests in anticipation of running VMs termination, minimizing the number of active PMs and PM underutilization. We examine the performance of the proposed algorithm compared to other placement algorithms using a real-life validation dataset of nearly 850 thousand VMs. The results show significant improvements ranging from 7% to 33% reduction in the number of active PMs and up to 61% reduction in the number of hourly underutilized PMs.

Finding experts that can help address critical elements or problems in a project is a challenging task. This is especially true in global software development where there is often a need to identify developers with specific skill sets and expertise. It is also essential to identify developers that can help move the project forward. To address this, we propose an expert recommendation system to help identify and classify individuals with the expertise and skills that could collaborate on a project. To achieve this goal, we model a popular project on GitHub as a contribution network. Our approach analyzes both syntactic and semantic aspects of the data by exploring the network with machine learning algorithms and considering an ontology that can be used to extract topics from the project's key terms. Our approach looks to recommend individuals with the necessary skills, have been strong contributors in the past, and have a positive interest trend.

Smartphones present unique data collection opportunities for research. Ecological momentary assessment (EMA) studies have been conducted using smartphones in many areas, as participant responses can be collected at the same time as device sensor data. We wanted to study undergraduate mental health using smartphones while investigating data from as many smartphone sensors as possible. The app used for this had to be accessible from the standard app stores to use existing participant phones. The software also had to have an easy setup process at the server-side due to limited resources. As a result, we developed the Ecological Momentary Assessment eXtensions platform. We describe the platform, two studies that used Version 2 of it, and summarize data collection results. We then outline a pathway to a unified EMAX3 app. To the best of our knowledge, EMAX is the only platform developed and used for sensor-based EMA research at a Canadian university.

The analysis of social structures is important in many contexts, especially in Global Software Development, where various developers with diverse skills and knowledge are involved. In this sense, searching for essential members is a valuable task since they are fundamental to the network's evolution. With the main goal of identifying and monitoring these individuals, we propose a temporal analysis approach that explores syntactic and semantic network aspects. We describe experiments based on popular projects on Github that consider a network modeled over consecutive time periods. We also propose an ontology to represent the domain knowledge and explore the network by investigating its semantic context. We provide evidence that our approach can detect individuals considered essential based on their role in the network.

An aging population has motivated research into Ambient Assisted Living (AAL) with the aim of supporting people to continue to live in their homes as they age or with chronic health problems. As part of this work, some researchers have focused on identifying and reporting daily activities of individuals at home in order to try to reduce the workload on caregivers and health professionals. Such an environment is usually monitored by non-wearable sensors that collect a vast amount of data. The use of this data requires computational methods that can process it in a reasonable time. This paper proposes an event-driven framework to detect unusual patterns in AAL environments. A Fog-Cloud paradigm and Lambda architecture are adopted as the framework to support computational solutions to deal with the volume of data, and machine learning techniques are used for recognition of daily activities. The framework was evaluated through a case study based on data collected in a real environment. Results point to the feasibility of the proposal.

Driving Assistance Systems increase safety and provide a more enjoyable driving experience. Among the objectives motivating these technologies rests the idea of predicting driver intent within the next few seconds, in order to avoid potentially dangerous manoeuvres. In this work, we develop a model of driver behaviour for turn manoeuvres that we then apply to anticipate the most likely turn manoeuvre a driver will effect a few seconds ahead of time. We demonstrate that cephalo-ocular behaviour such as variations in gaze direction and head pose play an important role in the prediction of driver-initiated manoeuvres. We tested our approach on a diverse driving data set recorded with an instrumented vehicle in the urban area of London, ON, Canada. Experiments show that our approach predicts turn manoeuvres 3.8 seconds before they occur with an accuracy over 80% in real-time.

Traffic sign detection and recognition systems are essential components of Advanced Driver Assistance Systems and self-driving vehicles. In this contribution we present a vision-based framework which detects and recognizes traffic signs inside the attentional visual field of drivers. This technique takes advantage of the driver 3D absolute gaze point obtained through the combined use of a front-view stereo imaging system and a non-contact 3D gaze tracker. We used a linear Support Vector Machine as a classifier and a Histogram of Oriented Gradient as features for detection. Recognition is performed by using Scale Invariant Feature Transforms and color information. Our technique detects and recognizes signs which are in the field of view of the driver and also provides indication when one or more signs have been missed by the driver.

Cloud computing continues to mature and more applications continue to be deployed in public clouds. Client applications deployed in the cloud should automatically scale up and down to match changing workload demands, though they must be careful to ensure that sufficient resources are provisioned to achieve performance objectives. The cloud provider, on the other hand, attempts to reduce costs by reducing power consumption by consolidating load onto fewer, highly utilized machines. In this work, we introduce an algorithm that integrates both application autoscaling and dynamic virtual machine (VM) allocation into a single algorithm in order to achieve the goals of both cloud provider and client. Further, we consider multi-VM applications, such as multi-tiered web-based applications, and extend the integrated algorithm to take the network topology into account when placing or migrating applications. The goal is to reduce VM-to-VM communication latency; our focus is on trying to contain applications within the same racks. We evaluate our work through simulation, showing that the integrated algorithm can achieve better application performance with a significant reduction in virtual machine live migrations, and the topology-aware extension successfully places applications within a single rack.

We present a robust and accurate technique for the cross-calibration of 3D remote gaze trackers with stereoscopic scene vision systems between which no common imaging area exists. We empirically demonstrate that a multi-depth calibration approach yields remarkably superior results for obtaining 3D Point-of-Gaze (PoG) when compared with traditional methods using monocular scene cameras and co-planar eye gaze calibration points.

This contribution consists of a frame-rate, vision-based, on-board method which detects vehicles within the attentional visual area of the driver. The method herein uses the 3D absolute gaze point of the driver obtained through the combined use of a front-view stereo imaging system and a non-contact 3D gaze tracker, alongside hypothesis-generation reducing techniques for vehicular detection such as horizon line detection. Trained AdaBoost classifiers are used in the detection process. This technique is the first of its kind in that it identifies vehicles the driver is most likely to be aware of at any moment while in the act of driving.

Increasingly, applications are moving into the cloud, which is actually supported by large-scale data centres on the ground. These data centres are complex systems to manage and centralized solutions might not be able to meet the required scale nor make an efficient use of their networks. In this paper, we propose a hierarchical approach to dynamic resource management in data centres, where we leverage the topology of the data centre network to design the management hierarchy. We define a set of aggregate metrics at various levels in the hierarchy to convey system state information to higher management levels, and define managers' responsibilities and interactions. We evaluate our proposed approach through simulation. Experiments show that this management approach greatly reduces the flow of management data across the data centre network, thus reducing network overhead.

Vehicle tracking is an important issue in intelligent vehicle automation systems since it can be used to increase safety, convenience and efficiency in driving. Many of the methods for vehicle tracking use a vision-based system to recognize the neighboring vehicles and provide a real time map of nearby vehicles. In other methods, wireless communication between vehicles has been used to locate the vehicles within range and provide tracking information for driving assistance applications. In this paper we present a combination of both a vision-based system and a wireless based system to provide more accurate real-time information about neighboring vehicles. We assume that some of the vehicles are equipped with GPS receivers, a Dedicated Short Range Communication (DSRC) transceiver and one or more cameras mounted on the vehicle. This tracking method has been implemented and evaluated in urban, highway and intersection scenarios under different adoption rates. The results show that a combined approach can be more effective.

Computing today is increasingly moving into large-scale virtualized data centres, offering computing resources in the form of virtual machines (VMs) on a pay-per-usage basis. In order to minimize costs, VMs should be consolidated on as few physical machines (PMs) as possible, switching idle PMs into a power saving mode. It may be necessary to dynamically allocate and reallocate VMs to PMs in order to meet highly dynamic VM resource requirements. The problem of assigning VMs to PMs is known to be NP-Hard. Most solutions focus on a centralized approach, with a single management node making allocation decisions periodically. This approach suffers from poor scalability and the existence of a single point of failure. We present a fully distributed approach to dynamic VM management, and evaluate our approach using a simulation tool. Results indicate that the distributed approach can achieve similar performance to the centralized solution, while eliminating the single point of failure and reducing the network bandwidth required for management.

Computing is increasingly moving into large-scale data centres, providing resources on-demand for clients on a pay-per-usage basis. One form of such on-demand computing is an Infrastructure as a Service (IaaS) Cloud, which provides low level access to virtualized resources. Developing and evaluating data centre management techniques for such large-scale data centres presents a significant challenge. As such, most work turns to simulation tools as the test environment. We present a number of extensions and additional features to an existing simulation tool, DCSim. Our improvements to DCSim include work on the core of the simulator, improved event, communication and management mechanisms, and a more complete model of the structure of a data centre. We also present improved simulation configuration tools and output, including a unique visualization tool. We evaluate the usefulness of the simulator through a demonstration of its use in comparing dynamic VM management techniques.

The near future will see production vehicles equipped with intelligent Advanced Driving Systems using various sensor modalities. Meanwhile, other aspects such as Vehicle-to-Vehicle (V2V) communications have emerged as means of assisting drivers. We propose a framework (V2Eye) in which V2V augments and extends the range of visual sensors in ways that have yet to be explored.

Virtualization is essential to cloud computing as it can provide the means for flexible, reliable and manageable computing resources. At the same time, it raises new challenges for system management, including resource allocation for virtual machines (VMs), means for providing quality of service for applications, meeting organizational objectives, etc. This paper explores the use of policies to drive the autonomic management of virtualization systems. We describe an architecture for autonomically managing VMs and the applications running on those VMs based on policies, outline a prototype implementation and present present an analysis of its performance.

The use of learning, in particular reinforcement learning, has been explored in the context of policy-driven autonomic management as a means of aiding decision making. In this context, the autonomic manager “learns” a model about what actions to take, for example, in certain situations. However, when the set of policies changes, the model is typically discarded or, if used, may yield misleading information. In contrast, this paper presents an approach for “re-using” past knowledge - by transforming a model learned from the use of one set of active policies to a new model when those policies change. This means that some of the “learned” knowledge can be utilized within the new environment. This is possible because our approach to modeling learning and adaptation is dependent only on the structure of the policies. Consequently, changes to policies can be mapped onto transformations specific to the model derived from the use of those policies. In this paper, we describe the model construction and policy modifications and elaborate, with a detailed case study, on how such changes could alter the currently learned model. Our analysis of the different kinds of policy modifications also suggest that, in most cases, most of the learned model can still be reused. This can significantly accelerate the learning process, essentially improving the overall quality of service, as the results presented in this paper demonstrate.

To more effectively use a network of high performance computing clusters, allocating multi-process jobs across multiple connected clusters, i.e., job co-allocation, offers the possibility of more efficient use of computer resources, reduced turn-around time and computations using numbers of processes larger than processors on any single cluster. Effective co-allocation, ultimately, depends on the inter-cluster communication cost. We previously introduced a scalable co-allocation strategy - maximum bandwidth adjacent cluster set (MBAS) strategy. It made use of two thresholds to control job co-allocation - one dealing with inter-cluster links and one controlling job partitioning. We subsequently introduced the adaptive threshold control system (ATCS), which used a fuzzy control approach to dynamically adjust these thresholds within MBAS. Results suggested that using ATCS during MBAS job co-allocation could achieve an overall performance improvement. However, these results only considered jobs that involved either master-slave or all-all communications among constituent processes. In this paper, we extend this analysis by also considering jobs that exhibit 2D-mesh communication patterns and evaluate ATCS further.

New commodity-off-shelf architectures are being adopted as a differential computational power to respond the increase demand to solve a large number of complex applications. These configurations can be considered as special grid environments, called as multi-clusters, where computational resources are scattered over a wide area network. In these grid environments a well known challenge to be circumvented is the heterogeneity of computational resources. Network protocols are especially important, because of the significance for the execution of distributed global applications. As a result, the goal of this paper is to plot performance tendency for TCP, XTP e SCTP protocols referring to throughput, latency and number of flows aspects.

In order to effectively manage todays complex systems, system administrators are turning to automated solutions. Policy-driven management offers significant benefits since the use of policies can make it more straight forward to define and modify systems behavior at run-time, through policy manipulation, rather than through re-engineering. The use of policies within autonomic computing allows system administrators to embed existing knowledge into policies and thereby drive autonomic management. Equally important, however, is a need for autonomic systems to adapt the use of these policies to deal with not only the inherent human error, but also the changes in the configuration of the managed environment and the unpredictability in workloads. This paper reports on the use of reinforcement learning methodologies to determine how to best use a set of enabled policies to meet different performance objectives. The work is presented in the context of an adaptive policy-driven autonomic management system.

The goal of distributed systems management is to provide reliable, secure and efficient utilization of the network, processors and devices that comprise those systems. The management system makes use of management agents to collect events and data from managed objects while policies provide information on how to modify the behaviour of a management system. Systems, as well as policies governing the behaviour of the system and its constituents, change dynamically. This paper explores the services needed to efficiently identify and deploy management entities and be able to respond to a change in the system being managed or in changes to the policies being applied. We show how these services can be built on existing management services found in commercial management systems

Packet drops are the primary means of signalling congestion in TCP. However, packet drops have a great impact on round-trip time (RTT) variation of TCP applications. A different and more proactive approach is used by SmoothTCP-q, where ICMP-SQ messages are sent directly to the sender every time a threshold is reached in the queue size of the router. In this mechanism, SmoothTCP-q tries to avoid packet drops since it is not necessary to overload the router queue to discover the network bandwidth as Standard TCP does. As a result, SmoothTCP-q generally presents better quality of service characteristics in terms of RTT variation and packet drops when compared to TCP. In this paper we describe SmoothTCP-q and illustrate its main characteristics related to this proactive property.

Required or desired behavior of systems and applications can be expressed in terms of management policies. Such policies can in turn be used to express expected operational characteristics of these systems and possible management actions. We feel that policies can provide the kinds of directives best used for flexible autonomic management systems. The effective use of policies in autonomic management requires that the policies be captured and translated into actions within the autonomic system. In this paper we describe an architecture for an autonomic management and illustrate how policies are defined and mapped into its executable elements and how they are realized as actions. A prototype implementation of the autonomic system for managing an Apache Web server is described and the mechanisms for mapping policies presented.

The successful implementation of large-scale enterprise information systems is a central concern for the modern organization. In particular, it is important to assure management that these systems are functioning to their proper capacity and providing their indicated services. To this end, it is exceedingly important to obtain accurate measurements on the performance of the implemented systems without unduly burdening system administrators with additional management tasks. In this research, we examine the effectiveness of low administrative overhead approaches such as operational analysis and transaction-based approaches for determining the performance of implemented enterprise systems. In particular, this research examines where each approach is applicable through the use of a full-factorial experiment.

Companies have gradually abandoned internally developed enterprise applications in favor of commercial ones. These systems are often comprised of different, multi-vendor components, including, Web servers, databases and applications. For most organizations, performance problems are often addressed by acquiring larger servers or increasing network bandwidth. While the performance of an individual component or application can be and is analyzed, the overall performance of the integrated suite of applications remains largely ad hoc and determining the cause of problems becomes trial and error. We report on initial work to address these problems by studying a commercial enterprise application, understanding what information is important, how such information can be gathered and analyzed and what kinds of tools would be beneficial. We report on an analysis tool and some results of using that tool and on the potential of the approach and methodology for larger enterprise environments.

Operational profiles are a quantification of usage patterns for a software application. These profiles are used to measure software reliability by testing the software in a manner that represents actual use. The current definition of an operational profile states that it is the set of operations available in the application, and the operations probabilities of occurrence in customer usage scenarios. This definition is too limited. In most industrial applications, focusing on operations alone does not offer adequate representation of the use of software. The limited definition of operational profiles can restrict their applicability and hence software reliability analysis for many software development organizations. This paper describes a formal and practical extension of the current definition of operational profiles to increase their applicability.

The recent advancement in network technologies has made it possible for the convergence of voice and data networks. In particular, the differentiated services (DiffServ) model has helped in bridging the diverse performance requirements between voice and non real-time applications such as file transfer and e-mail. One key feature of the DiffServ model is its promise to bring scalable service discrimination to the Internet and intranets. With the heterogeneous nature of today's distributed systems, however, there is a demand for flexible management systems that can cope with not only changes in resource needs and the perceived quality of applications (which is often user-dependent and application-specific), but also changes in the configuration of the distributed environment. Policy-based management has been proposed as a means of providing dynamic change in the behavior of applications and systems at run-time rather than through reengineering. The paper proposes a policy-enabled DiffServ architecture to provide predictable and measurable quality of service (QoS) for voice over Internet Protocol (VoIP) applications, and illustrates its performance.

It is often difficult to ensure that every application in a computing environment receives the level of quality of service required by their users. In such cases, the demand for computing resources to do so simply exceeds the limited supply that is available. To ensure that at least some user applications meet quality of service requirements, service differentiation is one approach growing in popularity. In this approach, preferential service is given to selected applications, while others deemed less important suffer in comparison. Most work in this area bases service differentiation decisions on static information about the applications, such as the name and type of application, the owner of the application, execution time, and the host on which the application was executed. In this paper, we discuss a new approach to service differentiation that takes into consideration dynamic application usage information in service differentiation decisions. In doing so, we can make better or fairer service differentiation decisions that allow more users to enjoy higher levels of quality of service. This is accomplished by ensuring preferential service is given to applications whose users can actually benefit from the improved service, as opposed to those applications that are essentially ignored or in a state that renders them unusable.

With the increased availability and complexity of distributed systems comes a greater need for solutions to assist in the management of distributed system components. We address three main aspects of management: 1) the need to efficiently utilize resources in the system, 2) the need to satisfy user and management requirements, as well as the requirements on resource usage, and 3) the need for dynamic solutions, i.e., that can be reconfigured due to changes in requirements, resource availability and management policies. Despite the significant contributions made towards the development of management tools that monitor and control distributed systems, little has been done to address these points. In this paper we describe a model that captures the points above. The need for reconfiguration, our implementation and its performance are also described.

There is relatively little information on the process that should be taken to derive a software architecture. We propose a process for the development of software architectures for large systems and describe our experiences in the development of an architecture for distributed systems that helped in defining this process.

Management policies can be used to specify requirements about the desired behaviour of distributed systems. Violations of policies (faults) can then be detected, isolated, located and corrected using a policy-driven fault management system. Other work in this area to date has focused on network-level faults. We believe that in a distributed system it is more appropriate to focus on faults at the application level. Furthermore, this work has been largely domain-specific-a generic, structured approach to this problem is needed. Our work has focused on policy-driven fault management in distributed systems at the application level. In this paper, we define a generic architecture for policy-driven fault management and present a prototype system based on this architecture. We also discuss experience to date using and experimenting with our prototype system.

Optimism is a powerful technique for avoiding latency by increasing concurrency. By optimistically assuming the results of some computation, other computations can be executed in parallel, even when they depend on the assumed result. Optimistic techniques can be particularly beneficial to parallel and distributed systems because of the critical impact of inter-node communications latency. This paper describes how optimism can be used to enhance the performance of distributed programs by avoiding remote communications delay. We then present a new programming model that automates many of the difficulties of using optimistic techniques in a general programming environment, and describe a prototype implementation. Finally, we present performance measurements showing how optimism improved the performance of a test application in this environment.

The X.500 Directory Service provides a powerful mechanism for storing and retrieving information about objects in a distributed computing environment. This requires the functional components of the Directory Service to have knowledge of the structure and representation, or schema, of the information held within the directory. The management of the Directory Schema is a subject requiring further research and development. We identify the three major technical elements required for properly managing the Directory Schema. We then focus on one of these elements; propagation of the schema between functional components of the Directory Service. Three subproblems from within schema propagation are presented, along with several alternative solutions.>

The problem of synthesizing a procedure from example computations is examined. An algorithm for this task is presented, and its success is considered. To do this, a model of procedures and example computations is introduced, and the class of acceptable examples is defined. The synthesis algorithm is shown to be successful, with respect to the model of procedures and examples, from two perspectives. First, it is shown to be sound, that is, that the procedure synthesized from a set of examples produces the same result as the intended one on the inputs used to generate that set of examples. Second, it is shown to be complete, that is, that for any procedure in the class of procedures, there exists a finite set of examples such that the procedure synthesized behaves as the intended one on all inputs for which the intended one halts.

Beginning September, 1975, the Computer Science Department at the University of Western Ontario switched from FORTRAN to PASCAL as the primary language in the first year introductory course. The motivation for such a change was to introduce into the first year course concepts and issues which were becoming central to Computer Science. In particular, the instructors involved in the course wished to a) emphasize algorithm design, effect of choice of representation on design and the creation of well-structured programs, b) expose students to the use of a variety of control structures and data structures, c) introduce, at least informally, the notions of complexity and correctness.

In recent years there has been a growing interest and concern among professionals and educators in Computer Science about undergraduate and graduate programs. A student could possibly offer a new viewpoint into this area, being immediately involved in such a program. However, his view is restricted by the educational environment he has been exposed to, usually in a single undergraduate institution. The following presentation of an undergraduate student's views on a Computer Science education is based upon problems and solutions encountered in designing a possible undergraduate Computer Science curriculum for the University of Dayton. To be applicable, the curriculum needed to be constructed within the confines of current departmental and university constraints and also needed to consider faculty and their interests. This curriculum could be adapted to another university and to another faculty, providing that the faculty can cover the areas reflected in the curriculum.