Unsupervised online learning of condition-invariant images for place recognition

This paper presents an online, unsupervised training algorithm enabling vision-based place recognition across a wide range of changing environmental conditions such as those caused by weather, seasons, and day-night cycles. The technique applies principal component analysis to distinguish between aspects of a location’s appearance that are condition-dependent and those that are condition-invariant. Removing the dimensions associated with environmental conditions produces condition-invariant images that can be used by appearance-based place recognition methods. This approach has a unique benefit – it requires training images from only one type of environmental condition, unlike existing data-driven methods that require training images with labelled frame correspondences from two or more environmental conditions. The method is applied to two benchmark variable condition datasets. Performance is equivalent or superior to the current state of the art despite the lesser training requirements, and is demonstrated to generalise to previously unseen locations.

The role of human factors in led outdoor activity accidents: literature review and exploratory analysis

While the exact rate of incidence is unknown (due to the paucity of exposure data), it is acknowledged that safety compromising accidents and incidents occur in the led outdoor activity domain, and that they represent an important issue. Despite this, compared to other safety critical domains, very little is currently known about the key causal factors involved in such accidents and incidents. This report presents the findings derived from a review of the literature, the aim of which was to identify the Human Factors-related issues involved in accidents and incidents occurring in this area. In addition, to demonstrate the utility of systems-based, theoretically underpinned accident analysis methodologies for identifying the systemic and human contribution to accidents and incidents occurring in the led outdoor activity domain, three case-study accidents were analysed using two such approaches. In conclusion, the review identified a range of causal factors cited in the literature; however, it was noted that the majority of the research undertaken to date lacks theoretical underpinning and focuses mainly on instructor or activity leader causal factors, as opposed to the wider system failures involved. The accident analysis presented highlighted the utility of systems-based, theoretically underpinned accident analysis methodologies for analysing and learning from accidents and incidents in the led outdoor activity sector. In closing, the need for further research in the area is articulated, in particular focussing on the development of standardised and universally accepted accident and incident reporting systems and databases, the development of data driven, theoretically underpinned causal factor taxonomies, and the development and application of systems-based accident analysis methodologies.

Analyzing the impact of learning inputs on near-to-far terrain traversability estimation

With the increasing need to adapt to new environments, data-driven approaches have been developed to estimate terrain traversability by learning the rover’s response on the terrain based on experience. Multiple learning inputs are often used to adequately describe the various aspects of terrain traversability. In a complex learning framework, it can be difficult to identify the relevance of each learning input to the resulting estimate. This paper addresses the suitability of each learning input by systematically analyzing the impact of each input on the estimate. Sensitivity Analysis (SA) methods provide a means to measure the contribution of each learning input to the estimate variability. Using a variance-based SA method, we characterize how the prediction changes as one or more of the input changes, and also quantify the prediction uncertainty as attributed from each of the inputs in the framework of dependent inputs. We propose an approach built on Analysis of Variance (ANOVA) decomposition to examine the prediction made in a near-to-far learning framework based on multi-task GP regression. We demonstrate the approach by analyzing the impact of driving speed and terrain geometry on the prediction of the rover’s attitude and chassis configuration in a Marsanalogue terrain using our prototype rover Mawson.

Influencers on quality of life as reported by people living with dementia in long-term care: A descriptive exploratory approach

Background Over half of the residents in long-term care have a diagnosis of dementia. Maintaining quality of life is important, as there is no cure for dementia. Quality of life may be used as a benchmark for caregiving, and can help to enhance respect for the person with dementia and to improve care provision. The purpose of this study was to describe quality of life as reported by people living with dementia in long-term care in terms of the influencers of, as well as the strategies needed, to improve quality of life. Methods A descriptive exploratory approach. A subsample of twelve residents across two Australian states from a national quantitative study on quality of life was interviewed. Data were analysed thematically from a realist perspective. The approach to the thematic analysis was inductive and data-driven. Results Three themes emerged in relation to influencers and strategies related to quality of life: (a) maintaining independence; (b) having something to do, and; (c) the importance of social interaction. Conclusions The findings highlight the importance of understanding individual resident needs and consideration of the complexity of living in large group living situations, in particular in regard to resident decision-making.

Regulating ride-sharing in the peer economy

The rise of the peer economy poses complex new regulatory challenges for policy-makers. The peer economy, typified by services like Uber and AirBnB, promises substantial productivity gains through the more efficient use of existing resources and a marked reduction in regulatory overheads. These services are rapidly disrupting existing established markets, but the regulatory trade-offs they present are difficult to evaluate. In this paper, we examine the peer economy through the context of ride-sharing and the ongoing struggle over regulatory legitimacy between the taxi industry and new entrants Uber and Lyft. We first sketch the outlines of ride-sharing as a complex regulatory problem, showing how questions of efficiency are necessarily bound up in questions about levels of service, controls over pricing, and different approaches to setting, upholding, and enforcing standards. We outline the need for data-driven policy to understand the way that algorithmic systems work and what effects these might have in the medium to long term on measures of service quality, safety, labour relations, and equality. Finally, we discuss how the competition for legitimacy is not primarily being fought on utilitarian grounds, but is instead carried out within the context of a heated ideological battle between different conceptions of the role of the state and private firms as regulators. We ultimately argue that the key to understanding these regulatory challenges is to develop better conceptual models of the governance of complex systems by private actors and the available methods the state has of influencing their actions. These struggles are not, as is often thought, struggles between regulated and unregulated systems. The key to understanding these regulatory challenges is to better understand the important regulatory work carried out by powerful, centralised private firms – both the incumbents of existing markets and the disruptive network operators in the peer-economy.

Regulating ride-sharing in the peer economy

The rise of the peer economy poses complex new regulatory challenges for policy-makers. The peer economy, typified by services like Uber and AirBnB, promises substantial productivity gains through the more efficient use of existing resources and a marked reduction in regulatory overheads. These services are rapidly disrupting existing established markets, but the regulatory trade-offs they present are difficult to evaluate. In this paper, we examine the peer economy through the context of ride-sharing and the ongoing struggle over regulatory legitimacy between the taxi industry and new entrants Uber and Lyft. We first sketch the outlines of ride-sharing as a complex regulatory problem, showing how questions of efficiency are necessarily bound up in questions about levels of service, controls over pricing, and different approaches to setting, upholding, and enforcing standards. We outline the need for data-driven policy to understand the way that algorithmic systems work and what effects these might have in the medium to long term on measures of service quality, safety, labour relations, and equality. Finally, we discuss how the competition for legitimacy is not primarily being fought on utilitarian grounds, but is instead carried out within the context of a heated ideological battle between different conceptions of the role of the state and private firms as regulators. We ultimately argue that the key to understanding these regulatory challenges is to develop better conceptual models of the governance of complex systems by private actors and the available methods the state has of influencing their actions. These struggles are not, as is often thought, struggles between regulated and unregulated systems. The key to understanding these regulatory challenges is to better understand the important regulatory work carried out by powerful, centralised private firms – both the incumbents of existing markets and the disruptive network operators in the peer-economy.

Ethical leadership in a time of increasing accountability

Australia, like many other countries, has embraced national testing as part of wider reforms and increased accountability in schooling. Results for standardised testing programs, such as NAPLAN, are widely published yet form only one part of accountability for educators. We argue that accountability also has moral, ethical and professional dimensions. In this paper we offer a discussion of background to our study of ethical leadership in a time of data driven or contractual accountability. Based on Starratt’s (1996) model, we define ethical leadership as a social, relational practice concerned with the moral purpose of education (Angus, 2006). Our central thesis is that given increasing accountabilities, school leaders need to consider approaches to ethical leadership to improve quality and equity in education and achieve equitable outcomes for all students. The paper concludes with key implications for school leaders.

Compiling Techniques for Coarse Grained Runtime Reconfigurable Architectures

In this paper we develop compilation techniques for the realization of applications described in a High Level Language (HLL) onto a Runtime Reconfigurable Architecture. The compiler determines Hyper Operations (HyperOps) that are subgraphs of a data flow graph (of an application) and comprise elementary operations that have strong producer-consumer relationship. These HyperOps are hosted on computation structures that are provisioned on demand at runtime. We also report compiler optimizations that collectively reduce the overheads of data-driven computations in runtime reconfigurable architectures. On an average, HyperOps offer a 44% reduction in total execution time and a 18% reduction in management overheads as compared to using basic blocks as coarse grained operations. We show that HyperOps formed using our compiler are suitable to support data flow software pipelining.

The accuracy of varietal selection using factor analytic models for multi-environment plant breeding trials

Modeling of cultivar x trial effects for multienvironment trials (METs) within a mixed model framework is now common practice in many plant breeding programs. The factor analytic (FA) model is a parsimonious form used to approximate the fully unstructured form of the genetic variance-covariance matrix in the model for MET data. In this study, we demonstrate that the FA model is generally the model of best fit across a range of data sets taken from early generation trials in a breeding program. In addition, we demonstrate the superiority of the FA model in achieving the most common aim of METs, namely the selection of superior genotypes. Selection is achieved using best linear unbiased predictions (BLUPs) of cultivar effects at each environment, considered either individually or as a weighted average across environments. In practice, empirical BLUPs (E-BLUPs) of cultivar effects must be used instead of BLUPs since variance parameters in the model must be estimated rather than assumed known. While the optimal properties of minimum mean squared error of prediction (MSEP) and maximum correlation between true and predicted effects possessed by BLUPs do not hold for E-BLUPs, a simulation study shows that E-BLUPs perform well in terms of MSEP.

Development of a model based on Bayesian networks to estimate the probability of sheep lice presence at shearing

This paper describes the development of a model, based on Bayesian networks, to estimate the likelihood that sheep flocks are infested with lice at shearing and to assist farm managers or advisers to assess whether or not to apply a lousicide treatment. The risk of lice comes from three main sources: (i) lice may have been present at the previous shearing and not eradicated; (ii) lice may have been introduced with purchased sheep; and (iii) lice may have entered with strays. A Bayesian network is used to assess the probability of each of these events independently and combine them for an overall assessment. Rubbing is a common indicator of lice but there are other causes too. If rubbing has been observed, an additional Bayesian network is used to assess the probability that lice are the cause. The presence or absence of rubbing and its possible cause are combined with these networks to improve the overall risk assessment.

Structure and dynamics of two-dimensional sheared granular flows

The structure and dynamics of the two-dimensional linear shear flow of inelastic disks at high area fractions are analyzed. The event-driven simulation technique is used in the hard-particle limit, where the particles interact through instantaneous collisions. The structure (relative arrangement of particles) is analyzed using the bond-orientational order parameter. It is found that the shear flow reduces the order in the system, and the order parameter in a shear flow is lower than that in a collection of elastic hard disks at equilibrium. The distribution of relative velocities between colliding particles is analyzed. The relative velocity distribution undergoes a transition from a Gaussian distribution for nearly elastic particles, to an exponential distribution at low coefficients of restitution. However, the single-particle distribution function is close to a Gaussian in the dense limit, indicating that correlations between colliding particles have a strong influence on the relative velocity distribution. This results in a much lower dissipation rate than that predicted using the molecular chaos assumption, where the velocities of colliding particles are considered to be uncorrelated.

Analysis and Modelling of the Effects of Water Stress on Maize Growth and Yield in Dryland Conditions.

It is essential to provide experimental evidence and reliable predictions of the effects of water stress on crop production in the drier, less predictable environments. A field experiment undertaken in southeast Queensland, Australia with three water regimes (fully irrigated, rainfed and irrigated until late canopy expansion followed by rainfed) was used to compare effects of water stress on crop production in two maize (Zea mays L.) cultivars (Pioneer 34N43 and Pioneer 31H50). Water stress affected growth and yield more in Pioneer 34N43 than in Pioneer 31H50. A crop model APSIM-Maize, after having been calibrated for the two cultivars, was used to simulate maize growth and development under water stress. The predictions on leaf area index (LAI) dynamics, biomass growth and grain yield under rain fed and irrigated followed by rain fed treatments was reasonable, indicating that stress indices used by APSIM-Maize produced appropriate adjustments to crop growth and development in response to water stress. This study shows that Pioneer 31H50 is less sensitive to water stress and thus a preferred cultivar in dryland conditions, and that it is feasible to provide sound predictions and risk assessment for crop production in drier, more variable conditions using the APSIM-Maize model.

Refining the process of agent selection through understanding plant demography and plant response to herbivory

Understanding plant demography and plant response to herbivory is critical to the selection of effective weed biological control agents. We adopt the metaphor of 'filters' to suggest how agent prioritisation may be improved to narrow our choices down to those likely to be most effective in achieving the desired weed management outcome. Models can serve to capture our level of knowledge (or ignorance) about our study system and we illustrate how one type of modelling approach (matrix models) may be useful in identifying the weak link in a plant life cycle by using a hypothetical and an actual weed example (Parkinsonia aculeata). Once the vulnerable stage has been identified we propose that studying plant response to herbivory (simulated and/or actual) can help identify the guilds of herbivores to which a plant is most likely to succumb. Taking only potentially effective agents through the filter of host specificity may improve the chances of releasing safe and effective agents. The methods we outline may not always lead us definitively to the successful agent(s), but such an empirical, data-driven approach will make the basis for agent selection explicit and serve as testable hypotheses once agents are released.

Analysing the Impact of Learning Inputs - Application to Terrain Traversability Estimation

Data-driven approaches such as Gaussian Process (GP) regression have been used extensively in recent robotics literature to achieve estimation by learning from experience. To ensure satisfactory performance, in most cases, multiple learning inputs are required. Intuitively, adding new inputs can often contribute to better estimation accuracy, however, it may come at the cost of a new sensor, larger training dataset and/or more complex learning, some- times for limited benefits. Therefore, it is crucial to have a systematic procedure to determine the actual impact each input has on the estimation performance. To address this issue, in this paper we propose to analyse the impact of each input on the estimate using a variance-based sensitivity analysis method. We propose an approach built on Analysis of Variance (ANOVA) decomposition, which can characterise how the prediction changes as one or more of the input changes, and also quantify the prediction uncertainty as attributed from each of the inputs in the framework of dependent inputs. We apply the proposed approach to a terrain-traversability estimation method we proposed in prior work, which is based on multi-task GP regression, and we validate this implementation experimentally using a rover on a Mars-analogue terrain.

Development of a Multiplexed Bead-Based Suspension Array for the Detection and Discrimination of Pospiviroid Plant Pathogens

Efficient and reliable diagnostic tools for the routine indexing and certification of clean propagating material are essential for the management of pospiviroid diseases in horticultural crops. This study describes the development of a true multiplexed diagnostic method for the detection and identification of all nine currently recognized pospiviroid species in one assay using Luminex bead-based suspension array technology. In addition, a new data-driven, statistical method is presented for establishing thresholds for positivity for individual assays within multiplexed arrays. When applied to the multiplexed array data generated in this study, the new method was shown to have better control of false positives and false negative results than two other commonly used approaches for setting thresholds. The 11-plex Luminex MagPlex-TAG pospiviroid array described here has a unique hierarchical assay design, incorporating a near-universal assay in addition to nine species-specific assays, and a co-amplified plant internal control assay for quality assurance purposes. All assays of the multiplexed array were shown to be 100% specific, sensitive and reproducible. The multiplexed array described herein is robust, easy to use, displays unambiguous results and has strong potential for use in routine pospiviroid indexing to improve disease management strategies.