Training deep models with limited annotations poses a significant challenge when applied to diverse practical domains. Employing semi-supervised learning alongside the self-supervised model offers the potential to enhance label efficiency. However, this approach faces a bottleneck in reducing the need for labels. We observed that the semi-supervised model disrupts valuable information from self-supervised learning when only limited labels are available. To address this issue, this paper proposes a simple yet effective framework, active self-semi-supervised learning (AS3L). AS3L bootstraps semi-supervised models with prior pseudo-labels (PPL). These PPLs are obtained by label propagation over self-supervised features. Based on the observations the accuracy of PPL is not only affected by the quality of features but also by the selection of the labeled samples. We develop active learning and label propagation strategies to obtain accurate PPL. Consequently, our framework can significantly improve the performance of models in the case of limited annotations while demonstrating fast convergence. On the image classification tasks across four datasets, our method outperforms the baseline by an average of 5.4%. Additionally, it achieves the same accuracy as the baseline method in about 1/3 of the training time.

Autonomous Underwater Vehicles (AUVs) conduct regular visual surveys of marine environments to characterise and monitor the composition and diversity of the benthos. The use of machine learning classifiers for this task is limited by the low numbers of annotations available and the many fine-grained classes involved. In addition to these challenges, there are domain shifts between image sets acquired during different AUV surveys due to changes in camera systems, imaging altitude, illumination and water column properties leading to a drop in classification performance for images from a different survey where some or all these elements may have changed. This paper proposes a framework to improve the performance of a benthic morphospecies classifier when used to classify images from a different survey compared to the training data. We adapt the SymmNet state-of-the-art Unsupervised Domain Adaptation method with an efficient bilinear pooling layer and image scaling to normalise spatial resolution, and show improved classification accuracy. We test our approach on two datasets with images from AUV surveys with different imaging payloads and locations. The results show that generic domain adaptation can be enhanced to produce a significant increase in accuracy for images from an AUV survey that differs from the training images.

We describe a novel semi-supervised learning method that reduces the labelling effort needed to train convolutional neural networks (CNNs) when processing georeferenced imagery. This allows deep learning CNNs to be trained on a per-dataset basis, which is useful in domains where there is limited learning transferability across datasets. The method identifies representative subsets of images from an unlabelled dataset based on the latent representation of a location guided autoencoder. We assess the method's sensitivities to design options using four different ground-truthed datasets of georeferenced environmental monitoring images, where these include various scenes in aerial and seafloor imagery. Efficiency gains are achieved for all the aerial and seafloor image datasets analysed in our experiments, demonstrating the benefit of the method across application domains. Compared to CNNs of the same architecture trained using conventional transfer and active learning, the method achieves equivalent accuracy with an order of magnitude fewer annotations, and 85 % of the accuracy of CNNs trained conventionally with approximately 10,000 human annotations using just 40 prioritised annotations. The biggest gains in efficiency are seen in datasets with unbalanced class distributions and rare classes that have a relatively small number of observations.

For Underwater Vehicle Manipulator Systems (UVMS), the ability to keep a fixed end effector pose is required for intervention tasks. Maintaining a static configuration in a dynamic underwater environment requires significant amounts of energy over time, limiting the operational time for battery powered systems. In this work we consider learning the periodic components of the dynamic flow in order to generate periodic trajectories which keep the end effector fixed, yet minimise the energy expenditure over time. We compare this proposed ‘go with the flow’ approach to the static configuration case for a fixed end effector pose, and show a significant reduction in energy use.

Autonomous Underwater Vehicles (AUVs) are increasingly being used to support scientific research and monitoring studies. One such application is in benthic habitat mapping where these vehicles collect seafloor imagery that complements broadscale bathymetric data collected using sonar. Using these two data sources, the relationship between remotely-sensed acoustic data and the sampled imagery can be learned, creating a habitat model. As the areas to be mapped are often very large and AUV systems collecting seafloor imagery can only sample from a small portion of the survey area, the information gathered should be maximised for each deployment. This paper illustrates how the habitat models themselves can be used to plan more efficient AUV surveys by identifying where to collect further samples in order to most improve the habitat model. A Bayesian neural network is used to predict visually-derived habitat classes when given broad-scale bathymetric data. This network can also estimate the uncertainty associated with a prediction, which can be deconstructed into its aleatoric (data) and epistemic (model) components. We demonstrate how these structured uncertainty estimates can be utilised to improve the model with fewer samples. Such adaptive approaches to benthic surveys have the potential to reduce costs by prioritizing further sampling efforts. We illustrate the effectiveness of the proposed approach using data collected by an AUV on offshore reefs in Tasmania, Australia.

When a motor-powered vessel travels past a fixed hydrophone in a multipath environment, a Lloyd's mirror constructive/destructive interference pattern is observed in the output spectrogram. The power cepstrum detects the periodic structure of the Lloyd's mirror pattern by generating a sequence of pulses (rahmonics) located at the fundamental quefrency (periodic time) and its multiples. This sequence is referred to here as the `rahmonic component' of the power cepstrum. The fundamental quefrency, which is the reciprocal of the frequency difference between adjacent interference fringes, equates to the multipath time delay. The other component of the power cepstrum is the non-rahmonic (extraneous) component, which combines with the rahmonic component to form the (total) power cepstrum. A data processing technique, termed `cepstrum subtraction', is described. This technique suppresses the extraneous component of the power cepstrum, leaving the rahmonic component that contains the desired multipath time delay information. This technique is applied to real acoustic recordings of motor-vessel transits in a shallow water environment, where the broadband noise radiated by the vessel arrives at the hydrophone via a direct ray path and a time-delayed multipath. The results show that cepstrum subtraction improves multipath time delay estimation by a factor of two for the at-sea experiment.

The propagation of sound in a shallow water environment is characterized by boundary reflections from the sea surface and sea floor. These reflections result in multiple (indirect) sound propagation paths, which can degrade the performance of passive sound source localization methods. This paper proposes the use of convolutional neural networks (CNNs) for the localization of sources of broadband acoustic radiated noise (such as motor vessels) in shallow water multipath environments. It is shown that CNNs operating on cepstrogram and generalized cross-correlogram inputs are able to estimate more reliably the instantaneous range and bearing of transiting motor vessels when the source localization performance of conventional passive ranging methods is degraded. The ensuing improvement in source localization performance is demonstrated using real data collected during an at-sea experiment.

Recovering information from contrast-limited, SNR-limited, color-attenuated images in a scattering media is of paramount importance for the autonomous functioning of robotic agents. The task is challenging due to the transient state of the medium, unknown medium parameters and in many cases the need for fully autonomous operation. This work presents a target-less, calibration-less method for restoring underwater light field images and requires no explicit model of the medium. The method adopts a light-field imaging approach to capture, model and compensate for backscatter in the scene leading to the recovery of high- fidelity images. The proposed method for backscatter compensation is validated against other state-of-the-art methods and is demonstrated to yield superior image quality.

Handling loop closures and intervehicle observations in cooperative robotic scenarios remains a challenging problem due to data consistency, bandwidth limitations and increased computation requirements. This paper develops a general cooperative localisation and single vehicle Visual SLAM framework that includes direct intervehicle observations and pose to pose loop closures on each vehicle with states shared as required. This fuses single vehicle SLAM with cooperative localisation and avoids data association of map data across limited communication networks. The base problem is developed as a factor graph with each vehicle solving local subgraphs that are split based on intervehicle observations. We modify the order of variable elimination in subgraphs through manipulation of the square-root of the Information matrix to extract updates that include the historic states involved in the loop closures and do not require transmission of other states not involved in the measurement or retransmission of previously shared states. We demonstrate the effect on localisation accuracy and bandwidth using data captured from a set of five robots observing each other and landmarks compared to both single vehicle SLAM, pure cooperative localisation and a centralised solution.

A cost effective approach to remote monitoring of protected areas such as marine reserves and restricted naval waters is to use passive sonar to detect, classify, localize, and track marine vessel activity (including small boats and autonomous underwater vehicles). Cepstral analysis of underwater acoustic data enables the time delay between the direct path arrival and the first multipath arrival to be measured, which in turn enables estimation of the instantaneous range of the source (a small boat). However, this conventional method is limited to ranges where the Lloyd's mirror effect (interference pattern formed between the direct and first multipath arrivals) is discernible. This paper proposes the use of convolutional neural networks (CNNs) for the joint detection and ranging of broadband acoustic noise sources such as marine vessels in conjunction with a data augmentation approach for improving network performance in varied signal-to-noise ratio (SNR) situations. Performance is compared with a conventional passive sonar ranging method for monitoring marine vessel activity using real data from a single hydrophone mounted above the sea floor. It is shown that CNNs operating on cepstrum data are able to detect the presence and estimate the range of transiting vessels at greater distances than the conventional method.

This paper considers the modelling of scalar fields exhibiting non-stationary noise in the context of Gaussian Process (GP) regression. We show how a Heteroscedastic GP produces more accurate predictions of the variance of a process of this type compared to the standard Homoscedastic model. We present a parametric model for the noise process and derive analytical solutions to the Log Marginal Likelihood of the data and its gradients with respect to Hyper Parameters of the kernel and the noise process. We compare our parametric model to one which estimates a full GP for the noise and show analogous predictive performance with a model which has greater computational efficiency and is less complex to implement. We also discuss active sample selection in this framework and show through the numerical simulation of an arrested bathymetric front in an estuary, the superiority of using Mutual Information to Fisher Information, Entropy or Random sampling in terms of errors in the first two moments of the predicted distributions.

Autonomous underwater vehicles (AUVs) are widely used to perform information gathering missions in unseen environments. Given the sheer size of the ocean environment, and the time and energy constraints of an AUV, it is important to consider the potential utility of candidate missions when performing survey planning. In this paper, we utilise a multimodal learning approach to capture the relationship between in-situ visual observations, and shipborne bathymetry (ocean depth) data that are freely available a priori. We then derive information-theoretic measures under this model that predict the amount of visual information gain at an unobserved location based on the bathymetric features. Unlike previous approaches, these measures consider the value of additional visual features, rather than just the habitat labels obtained. Experimental results with a toy dataset and real marine data demonstrate that the approach can be used to predict the true utility of unexplored areas.

We demonstrate that the redundant information in light field imagery allows volumetric focus, an improvement of signal quality that maintains focus over a controllable range of depths. To do this, we derive the frequency-domain region of support of the light field, finding it to be the 4D hyperfan at the intersection of a dual fan and a hypercone, and design a filter with correspondingly shaped passband. Drawing examples from the Stanford Light Field Archive and images captured using a commercially available lenslet-based plenoptic camera, we demonstrate that the hyperfan outperforms competing methods including planar focus, fan-shaped antialiasing, and nonlinear image and video denoising techniques. We show the hyperfan preserves depth of field, making it a single-step all-in-focus denoising filter suitable for general-purpose light field rendering. We include results for different noise types and levels, through murky water and particulate matter, in real-world scenarios, and evaluated using a variety of metrics. We show that the hyperfan's performance scales with aperture count, and demonstrate the inclusion of aliased components for high-quality rendering.

This paper presents a system for crowdsourcing saliency interest points for robotically gathered 3D maps rendered on smartphones and tablets. An app was created that is capable of interactively rendering 3D reconstructions gathered with an Autonomous Underwater Vehicle. Through hundreds of thousands of logged user interactions with the models we attempt to data-mine salient interest points. To this end we propose two models for calculating saliency from human interaction with the data. The first uses the view frustum of the camera to track the amount of time points are on screen. The second treats the camera's path as a time series and uses a Hidden Markov model to learn the classification of salient and non-salient points. To provide a comparison to existing techniques, several traditional visual saliency approaches are applied to orthographic views of the models' photo-texturing. The results of all approaches are validated with human attention ground truth gathered using a remote gaze-tracking system that recorded the locations of the person's attention while exploring the models.

Autonomous Underwater Vehicles (AUVs) gather large volumes of visual imagery, which can help monitor marine ecosystems and plan future surveys. One key task in marine ecology is benthic habitat mapping, the classification of large regions of the ocean floor into broad habitat categories. Since visual data only covers a small fraction of the ocean floor, traditional habitat mapping is performed using shipborne acoustic multi-beam data, with visual data as ground truth. However, given the high resolution and rich textural cues in visual data, an ideal approach should explicitly utilise visual features in the classification process. To this end, we propose a multimodal model which utilises visual data and shipborne multi-beam bathymetry to perform both classification and sampling tasks. Our algorithm learns the relationship between both modalities, but is also effective when visual data is missing. Our results suggest that by performing multimodal learning, classification performance is improved in scenarios where visual data is unavailable, such as the habitat mapping scenario. We also demonstrate empirically that the model is able to perform generative tasks, producing plausible samples from the underlying data-generating distribution.

This paper introduces a distributed and decentralised method to solve the cooperative localisation problem utilising the factor graph framework. This method enables vehicles to create compact packets of their own sensor information between their involvement in intervehicle measurements. The packets are limited in size, with this size dependent on the size of the state space alone. The number of packets generated is also limited at two packets (one per vehicle) created per intervehicle measurement. The packets and measurements can be shared and propagated to all vehicles not just direct neighbours or to vehicles involved in the measurements. Vehicles maintain a local solver that incorporates all local sensor information and motion updates in full nonlinear form and includes the fixed linearised packets from other vehicles. This local solver is able to update local state variables and relinearise local and intervehicle factors but has to hold remote state variables and packet data at a fixed linearisation point. We show the estimated solution is of a similar quality to a state of the art centralised relinearising estimator iSAM2 and superior to a fixed linearisation filtering solution via comparing RMS error in position estimation in simulation. Consistency of the method is also shown via the NEES metric. Communication requirements for each of the competing methods are shown, with compaction of packets being more useful the larger the ratio between intervehicle and local measurement intervals. The technique is validated using multi-vehicle simulation and real datasets.

Plenoptic cameras are gaining attention for their unique light gathering and post-capture processing capabilities. We describe a decoding, calibration and rectification procedure for lenselet-based plenoptic cameras appropriate for a range of computer vision applications. We derive a novel physically based 4D intrinsic matrix relating each recorded pixel to its corresponding ray in 3D space. We further propose a radial distortion model and a practical objective function based on ray reprojection. Our 15-parameter camera model is of much lower dimensionality than camera array models, and more closely represents the physics of lenselet-based cameras. Results include calibration of a commercially available camera using three calibration grid sizes over five datasets. Typical RMS ray reprojection errors are 0.0628, 0.105 and 0.363 mm for 3.61, 7.22 and 35.1 mm calibration grids, respectively. Rectification examples include calibration targets and real-world imagery.

With the advent of cheap, high fidelity, digital imaging systems, the quantity and rate of generation of visual data can dramatically outpace a humans ability to label or annotate it. In these situations there is scope for the use of unsupervised approaches that can model these datasets and automatically summarise their content. To this end, we present a totally unsupervised, and annotation-less, model for scene understanding. This model can simultaneously cluster whole-image and segment descriptors, thereby forming an unsupervised model of scenes and objects. We show that this model outperforms other unsupervised models that can only cluster one source of information (image or segment) at once. We are able to compare unsupervised and supervised techniques using standard measures derived from confusion matrices and contingency tables. This shows that our unsupervised model is competitive with current supervised and weakly-supervised models for scene understanding on standard datasets. We also demonstrate our model operating on a dataset with more than 100,000 images collected by an autonomous underwater vehicle.

Maturing technology has allowed the reliable deployment of robots into large-scale environments for monitoring and exploration applications. Planning techniques which ignore the value of information gathered during transit are able to operate efficiently in these environments and generate trajectories between specified starting and ending locations. Including the value of information gathered during transit increases the complexity of the problem and often leads to algorithms which are unable to scale up to large environments. This paper presents a method for planning informative surveys in large-scale unexplored environments. The proposed methodology does not require a starting or ending location as a constraint. Instead, robot operators are required to specify a survey template, which satisfies both vehicle constraints and the scientific objectives of the deployment. This constraint converts the exploration problem into an experimental design problem where the objective is to choose a location for the specified survey trajectory. A functional representation of the survey utility is learnt using a Gaussian process. This model allows the utility of candidate survey placements to be queried in a continuous space and in arbitrary locations. The proposed exploration method is demonstrated and validated on marine data. The objective is to design a survey which allows the spatial distribution of habitats in a large marine environment to be estimated accurately. The results show that the proposed exploration method is able to model the hidden survey utility function successfully and recommend informative survey placements.

In this paper we detail a new calibration method for determining the spatial sensitivity of the field-of-view of a spectrometer as well as a method for determining the spatial co-registration of a co-mounted spectrometer and an RGB camera. The spatial sensitivity of the field-of-view of a spectrometer is not necessarily round nor even, and knowledge of this spatial sensitivity is essential for using the spectrometer in an imaging capacity. The combination of the spectrometer and RGB camera provides the capability to produce image data with both high spatial and spectral resolution, which is useful in classification and characterisation for remote sensing and mapping operations.

This paper presents recent developments in data processing of multi-year repeat survey imagery and precision automatic registration for monitoring long-term changes in benthic marine habitats such as coral reefs and kelp forests. Three different methods are presented and compared for precision alignment of imagery maps collected over a range of time-scales from 12 hours to two years between dives. The first method uses Scale Invariant Feature Transform (SIFT) features computed over imagery mosaics to compute the relative translational offset between repeat dives. The second method employs scan-optimisation using the bathymetry generated via structure-from-motion thus capturing more stable features in the environment, lending itself to larger timescale registration. The third method uses mutual information optimisation to register imagery maps, providing robustness to changes in the colour and brightness of objects in an underwater scene across multiple years. Results are presented from field data collected using an Autonomous Underwater Vehicle (AUV) in sites across the Australian coast between 2009 and 2011.

Modern robotic platforms, deployed for environmental monitoring and mapping, are able to rapidly accumulate large data sets. Whilst the data sets collected by these platforms are highly descriptive, they are often too large for human experts to analyse exhaustively. Although the large data sets could be analysed by humans in principle, the amount of labour and time required to process them is not cost effective. In this paper we focus on the classification task of learning the relationship between low resolution, remotely sensed data and categories derived from direct observations of the same phenomenon. To reduce the labour requirements of categorising the direct observations we forgo human supervision and rely on an unsupervised clustering model to segregate the observations into similar groups of data. Rather than using the discrete cluster labels to train a conventional classifier, we develop a new Gaussian process classifier capable of accepting probabilistic training targets. This allows the probabilistic information generated during clustering to be preserved during classification. We demonstrate the new model, in an environmental monitoring application, using data collected by an autonomous underwater vehicle.

A method is presented for isolating moving distractors from a static background in imagery captured by a hovering or slowly moving Autonomous Underwater Vehicle (AUV). By reparameterizing a set of monocular images into a light field structure, it becomes possible to apply a linear fan filter and its inverse to extract the background and distractors, respectively. Results are shown for a hovering AUV imaging a region with non-trivial 3D structure and containing moving elements. The output is a distractor-free 3D light field model of the sea floor and a set of images of isolated distractors. We show that the technique is insensitive to parallax in the background elements, outperforming pixel differencing techniques.

This paper presents a vision-based underwater mapping system, which is demonstrated in an archaeological survey of the submerged ancient town of Pavlopetri. The snorkeler or diver operated system provides a low cost alternative to the use of an AUV or ROV in shallow waters. The system produces textured three-dimensional models, which contain significantly more information than traditional archaeological survey methods. The photo-realistic maps that are produced allow further archaeological research to be performed, without diving on a site during the restrictive time limitations of permits and field seasons. The hardware and software components of the mapping system and its method of operation are described, and initial results are presented and discussed.

Survey class Autonomous Underwater Vehicles (AUVs) rely on Doppler Velocity Logs (DVL) for precise navigation near the seafloor. In cases where the seafloor depth is greater than the DVL bottom lock range, transiting from the surface where GPS is available to the seafloor presents a localisation problem since both GPS and DVL are unavailable in the mid-water column. This is traditionally addressed by using acoustic positioning systems, which take extra time to deploy or require a tracking vessel. Such systems increase the costs of operating in deep waters and reduce the flexibility of AUV operations. This paper proposes an alternative approach to navigation in the mid-water column that exploits the stability of current profiles of water columns over short periods of time. Observation of these currents are possible with the ADCP (Acoustic Doppler Current Profiler) mode of the DVL. Results with real data from missions with the Sirius AUV show how the full integration of water column descent with the ADCP, seafloor view-based SLAM (Simultaneous Localisation And Mapping), and ascent to the sea surface with ADCP gives results similar to having continuous bottom lock and shows potential to act as an alternative to acoustic localisation.

This paper presents an efficient and featureless approach to Bathymetric Simultaneous Localization And Mapping (SLAM) that utilizes a Rao-Blackwellized Particle Filter (RBPF) and Gaussian Process (GP) Regression to provide loop closures in areas where little to no overlap with previously explored terrain is present. To significantly reduce the memory requirements of this approach (thereby allowing for the processing of large datasets) a novel map representation is also introduced that, instead of directly storing estimates of seabed depth, records the trajectory of each particle and synchronizes them to a common log of bathymetric observations. Upon detecting a loop closure each particle is then weighted by matching new observations to the current predictions generated from a local reconstruction of their map using GP Regression. Here the spatial correlation in the environment is fully exploited, allowing predictions of seabed depth to be generated in areas that may not have been directly observed previously. The particle resampling that is performed therefore not only enforces self-consistency in overlapping sections of the map but additionally enforces self-consistency between neighboring map borders. The results demonstrate how observations of seafloor structure with partial overlap can be used by bathymetric SLAM to improve map self consistency when compared to Dead Reckoning fused with Long-Baseline observations. In addition we show how mapping corrections can still be achieved even when no map overlap is present.

Three closed-form solutions are proposed for six degree of freedom (6-DOF) visual odometry for light field cameras. The first approach breaks the problem into geometrically driven sub-problems with solutions adaptable to specific applications, while the second generalizes methods from optical flow to yield a more direct approach. The third solution integrates elements into a remarkably simple equation of plenoptic flow which is directly solved to estimate the camera's motion. The proposed methods avoid feature extraction, operating instead on all measured pixels, and are therefore robust to noise. The solutions are closed-form, computationally efficient, and operate in constant time regardless of scene complexity, making them suitable for real-time robotics applications. Results are shown for a simulated underwater survey scenario, and real-world results demonstrate good performance for a three-camera array, outperforming a state-of-the-art stereo feature-tracking approach.

This paper demonstrates an implementation of pool-based active learning through uncertainty sampling using a Variational Dirichlet Process (VDP) model. The VDP is used for both pre-clustering and classification, and is extended to incorporate fixed labels from an oracle (human annotator). Three different uncertainty sampling techniques are explored - least confident sampling, margin sampling and entropy based sampling. Clustering with the VDP is done in a completely unsupervised manner, without the need to specify the number of clusters. This appears particularly useful in improving the results when there are only few labelled samples, or if the cost of labelling is high. Results are shown for a toy dataset and the performance is compared to similar implementations using an Expectation Maximisation model (EM) and a Naive Bayes classifier (NB). The VDP active learning framework is tested on a stereo image dataset obtained by an autonomous underwater vehicle that covers several linear kilometres and consists of thousands of stereo image pairs. Our results show that combining an active learning strategy with the VDP significantly reduces the number of labelled images required to achieve a desired level of accuracy.

Autonomous underwater vehicles (AUVs) are commonly used to support oceanographic science by providing water-column mapping, seafloor bathymetric and photographic survey, and deep-sea exploration capabilities. In practice, the mapping activities carried out by AUVs consist of flying either pre-programmed tracklines (most propeller-driven AUVs), or else reporting data to human operators at regular intervals that permit retasking (typical for month-long underwater glider deployments). AUVs equipped with the ability to reason about scientific objectives in real time could significantly increase the value of individual deployments by enabling sampling efforts to be focused on targets or areas identified autonomously or semi-autonomously as scientifically interesting [1]. In this paper, we focus on AUV autonomy as it pertains to water-column sensing and argue that the classification of water-column sensor data represents an important enabling capability. We demonstrate practical, semi-supervised classification of water-column sensor data using a particular Bayesian, non-parametric clustering method, the Variational Dirichlet Process, combined with operator-supplied semantic labeling. The method is applied to the detection of a deep subsea hydrocarbon plume using data collected by the Woods Hole Oceanographic's Sentry AUV during an expedition to the Gulf of Mexico following the Deepwater Horizon blowout disaster.

This paper presents a novel architecture for a classification system based on the visual saliency of images. The work is motivated by the difficulty of reviewing large numbers of images as a human operator in the context of Autonomous Underwater Vehicle (AUV) surveys. We formulate a feature space in which an algorithm operates over color and texture to determine saliency and illustrate how this can be used to find interesting or unusual images within a large data set. The saliency classification based on these general image features allows for overlays highlighting interesting benthos or geologic structures on large scale 3D seafloor reconstructions, quickly providing spatial context to human observers. These results are validated using a set of human trials in which images are classified into salient and non-salient categories by a number of test subjects. The trials show good agreement both between subjects and between the human labels and the automated classification system. The results of the automated technique are also compared directly to a more traditional SVM classification system showing favorable results for our system for generalizing to new environments.

This paper describes an approach to achieving high resolution, repeated benthic surveying using an Autonomous Underwater Vehicle (AUV). A stereo based Simultaneous Localisation and Mapping (SLAM) technique is used to estimate the trajectory of the vehicle during multiple overlapping grid based surveys. The vehicle begins each dive on the surface and uses GPS to navigate to a designated start location. Once it reaches the designated location on the surface, the vehicle dives and executes a pre-programmed grid survey, collecting co-registered high resolution stereo images, multibeam sonar and water chemistry data. A suite of navigation instruments are used while the vehicle is underway to estimate its pose relative to the local navigation frame. Following recovery of the vehicle, the SLAM technique is used to refine the estimated vehicle trajectory and to find loop closures both within each survey and between successive missions to co-register the dives. Results are presented from recent deployments of the AUV Sirius at a site in North Eastern Tasmania. The objective of the deployments described in this work were to document the behaviour of barrens-forming sea sea urchins which have recently become resident in the area. The sea urchins can overgraze luxuriant kelp beds that once dominated these areas, leaving only rocky barrens habitat. The high resolution stereo images and resulting three dimensional surface models allow the nocturnal behaviour of the animals, which emerge to feed predominantly at night, to be described. Co-registered images and resulting habitat models collected during the day and at night are being analysed to describe the behaviour of the sea urchins in more detail.

Robotic agents that can explore and sample in a completely unsupervised fashion could greatly increase the amount of scientific data gathered in dangerous and inaccessible environments. Our application is imaging the benthos using an autonomous underwater vehicle with limited communication to surface craft. Robotic exploration of this nature demands in situ data analysis. To this end, this paper presents results of using a Gaussian Mixture Model (GMM), a Hidden Markov Model (HMM) filter, an Infinite Gaussian Mixture Model (IGMM) and a Variation Dirichlet Process model (VDP) for the classification of benthic habitats. All of the models are trained using unsupervised methods. Furthermore, the IGMM and VDP are trained without knowing the the number of classes in the dataset. It was found that the sequential information the HMM filter provides to the classification process adds lag to the habitat boundary estimates, reducing the classification accuracy. The VDP proved to be the most accurate classifier of the four tested, and also one of the fastest to train. We conclude that the VDP is a powerful model for entirely autonomous labelling of benthic datasets.

This paper describes a recent study in which an autonomous underwater vehicle (AUV) with a high resolution stereo-imaging system was used to document nocturnal camouflage behaviour in cuttlefish at a well known spawning site in Whyalla, South Australia. The AUV's ability to fly at low altitude during day and night while closely following a desired survey pattern provided improved data collection compared to divers and previous work with a small ROV. Over the course of the week long expedition, the AUV Sirius was deployed on 38 dives at three sites in the survey area and collected tens of thousands of stereo images. Of these, nearly a thousand were seen to contain cuttlefish during post cruise analysis, with a large proportion showing evidence of camouflage. The distribution of images containing cuttlefish suggest that the animal concentrations were substantially higher closer in to shore in shallow waters, where the flat rocky substrate occurs; females lay their eggs on the underside of these rocks. Results demonstrate the strengths of using an AUV for surveying near-shore benthic habitats of ecological interest, with a particular emphasis on the ability to operate during both day and night time operations.

In this paper we propose an approach to SLAM suitable for bathymetric mapping by an autonomous underwater vehicle (AUV). AUVs typically do not have access to GPS while underway and the survey areas of interest are unlikely to contain features that can easily be identified and tracked using bathymetric sonar. We demonstrate how the uncertainty in the vehicle state can be modeled using a particle filter and an Extended Kalman Filter (EKF), where each particle maintains a 2D depth map to model the seafloor. Efficient methods for maintaining and resampling the joint maps and particles using Distributed Particle Mapping are then described. Our algorithm was tested using field data collected by an AUV equipped with multibeam sonar. The results achieved by Bathymetric distributed Particle SLAM (BPSLAM) demonstrate how observations of the seafloor structure improve the estimated trajectory and resulting map when compared to dead reckoning fused with USBL observations, the best navigation solution during the trials. Furthermore, the computational run time to deliver these results falls well below the total mission time, providing the potential for the algorithm to be implemented in real time.

This paper presents a simultaneous localization and mapping algorithm suitable for large-scale visual navigation. The estimation process is based on the viewpoint augmented navigation (VAN) framework using an extended information filter. Cholesky factorization modifications are used to maintain a factor of the VAN information matrix, enabling efficient recovery of state estimates and covariances. The algorithm is demonstrated using data acquired by an autonomous underwater vehicle performing a visual survey of sponge beds. Loop-closure observations produced by a stereo vision system are used to correct the estimated vehicle trajectory produced by dead reckoning sensors.

This paper gives an overview of component-based software engineering (CBSE), motivates its application to the field of mobile robotics, and proposes a particular component model. CBSE is an approach to system-building that aims to shift the emphasis from programming to composing systems from a mixture of off-the-shelf and custom-built software components. This paper argues that robotics is particularly well-suited for and in need of component-based ideas. Furthermore, now is the right time for their introduction. The paper introduces Orca - an open-source component-based software engineering framework proposed for mobile robotics with an associated repository of free, reusable components for building mobile robotic systems.

This paper presents an approach of integrating human operators into a sensor network formed by a heterogeneous team of unmanned air and ground vehicles. Several objectives of human-network interaction are identified. The main focus of this work is on human-to-network information flow, i.e. human operators are regarded as information sources. It is argued that operators should make raw observations which are converted into the sensor network's common representation by a probabilistic model. The concepts are discussed in the context of an outdoor sensor network under development. Human operators contribute geometric feature information in the form of range and bearing observations. Visual feature properties are specified via meaningful class labels. A sensor model, represented as a Bayesian network, translates label observations into the system's representation. The model is also used to classify features as observed by robotic sensors.

Localization of nodes within a sensing network is a fundamental requirement for many applications. This paper proposes a method by which sensors self-localize based on their uncertain observations of other nodes in the network, using both Monte Carlo and Kalman filtering techniques. The proposed methods are demonstrated in a laboratory environment where stationary camera nodes self-localized in real-time by observing Pioneer robots moving about within their field of view. The robots take observations of surveyed beacons in the environment and provide estimates of their poses to the rest of the network.

The paper presents a decentralized approach to the solution of the distributed information gathering problem. The main design objectives are scalability with the number of network components, maximum flexibility in implementation and deployment, and robustness to component and communication failure. The design approach emphasizes interactions between components rather than the definition of the components themselves. The architecture specifies a small set of interfaces sufficient to implement a wide range of information gathering systems. The results of two deployment scenarios on an indoor sensor network are presented.

This paper presents results of the application of the simultaneous localisation and mapping algorithm to data collected by an unmanned underwater vehicle operating on the Great Barrier Reef in Australia. By fusing information from the vehicle's on-board sonar and vision systems, it is possible to use the highly textured reef to provide estimates of the vehicle motion as well as to generate models of the gross structure of the underlying reefs. Terrain-aided navigation promises to revolutionise the ability of marine systems to track underwater bodies in many applications. This work represents a crucial step in the development of underwater technologies capable of long-term, reliable deployment. Results of the application of this technique to the tracking of the vehicle position are shown.

This work is concerned with mapping of indoor environment by a team of robots who share the information they gather to build a global representation of the environment in a consistent, efficient, and scalable way. The well known approach of Decentralized Data Fusion is applied to a widely used representation of Certainty Grid maps. The result is a scalable and intuitive algorithm for combining observations from multiple heterogeneous sensing platforms with built-in processing and communication capabilities. The platforms are assumed to have access to external localization. Results of realistic simulations are presented.

This paper presents a novel approach to the multi-vehicle simultaneous localisation and mapping (SLAM) problem that exploits the manner in which observations are fused into the global map of the environment to manage the computational complexity of the algorithm and improve the data association process. Rather than incorporating every observation directly into the global map of the environment, the constrained local submap filter (CLSF) relies on creating an independent, local submap of the features in the immediate vicinity of the vehicle. This local submap is then periodically fused into the global map of the environment. This representation is shown to reduce the computational complexity of maintaining the global map estimates as well as improving the data association process. This paper examines the prospect of applying the CLSF algorithm to the multi-vehicle SLAM problem.

This paper presents a novel approach to the simultaneous localisation and mapping algorithm that exploits the manner in which observations are fused into the global map of the environment to manage the computational complexity of the algorithm and improve the data association process. Rather than incorporating every observation directly into the global map of the environment, the constrained local submap filter relies on creating an independent, local submap of the features in the immediate vicinity of the vehicle. This local submap is then periodically fused into the global map of the environment using appropriately formulated constraints between the common feature estimates. This approach is shown to be effective in reducing the computational complexity of maintaining the global map estimates as well as improving the data association process.

Exploration involving mapping and concurrent localization in an unknown environment is a pervasive task in mobile robotics. In general, the accuracy of the mapping process depends directly on the accuracy of the localization process. This paper address the problem of maximizing the accuracy of the map building process during exploration by adaptively selecting control actions that maximize localisation accuracy. The map building and exploration task is modeled using an Occupancy Grid (OG) with concurrent localisation performed using a feature-based Simultaneous Localisation And Mapping (SLAM) algorithm. Adaptive sensing aims at maximizing the map information by simultaneously maximizing the expected Shannon information gain (Mutual Information) on the OG map and minimizing the uncertainty of the vehicle pose and map feature uncertainty in the SLAM process. The resulting map building system is demonstrated in an indoor environment using data from a laser scanner mounted on a mobile platform.

Integrated exploration strategy advocated in this paper refers to a tight coupling between the tasks of localization, mapping, and motion control and the effect of this. coupling on the overall effectiveness of an exploration strategy. Our approach to exploration calls for a balanced evaluation of alternative motion actions from the point of view of information gain, localization quality, and navigation cost. To provide a uniform basis of comparison of localization quality between different locations, a "localizability" metric is introduced. It is based on the estimate of the lowest vehicle pose covariance attainable from a given location.

We present a system for behavior-based control of an autonomous underwater vehicle for the purpose of inspection of coral reefs, a task currently performed by divers holding a video camera while following a rope. Using sonar and vision-based approaches, behaviors have been developed for guiding the robot along its intended course, for maintaining a constant height above the sea floor, and for avoiding obstacles. A task-level controller selects which behaviors should be active according to user-defined plans and in response to system failures. Behavior arbitration has been implemented using both fuzzy logic and utility fusion. Initial experiments have been conducted in a natural coastal inlet, and the system is to be soon demonstrated in the coral reef environment.

We present results of the application of a simultaneous localisation and map building (SLAM) algorithm to estimate the motion of a submersible vehicle. Scans obtained from an on-board sonar are processed to extract stable point features in the environment. These point features are then used to build up a map of the environment while simultaneously providing estimates of the vehicle location. Results are shown from deployment in a swimming pool at the University of Sydney as well as from field trials in a natural environment along Sydney's coast. This work represents the first instance of a deployable underwater implementation of the SLAM algorithm.

This article describes work in progress at the Canadian Space Agency on the design and implementation of haptic devices. Haptic devices are a special class of robotic mechanisms for which structural transparency is a foremost design criterion. Also notable is the fact that often, three or four degrees of freedom, rather than six as in general robotic tasks, are sufficient for many haptic applications. Furthermore, in order to make these devices readily available to many users, it is necessary their kinematic models be sufficiently simple such that they can be controlled by inexpensive means. Various three and four-DOF mechanisms, some of which recently developed by the Canadian Space Agency, are discussed herein in terms of their suitability for haptic applications. For five or six-DOF applications, the concept of the virtual handle is introduced, reducing the problems and complexity usually associated with mechanisms with a high number of degrees of freedom.