Narada Dilp Warakagoda

• Orescanin, Marko; Olson, Derek R.; Duvio, Dalton; Jennings, Haley; Geilhufe, Marc & Hansen, Roy Edgar [Show all 7 contributors for this article] (2026). Physically Motivated Distributions for Bayesian Deep Learning of Synthetic Aperture Sonar Imaging Artifacts. IEEE Geoscience and Remote Sensing Letters. ISSN 1545-598X. 23. doi: 10.1109/LGRS.2026.3715375. Full text in Research Archive Show summary

  We demonstrate that the choice of weight prior in Bayesian neural networks (BNNs) is a consequential and nontrivial design decision for synthetic aperture sonar (SAS) imaging artifact classification. Motivated by established statistical models of acoustic seafloor backscatter, we replace the standard univariate Gaussian prior on the first convolutional layer with physically motivated alternatives: a multivariate normal (MVN) prior that captures empirical pixel covariance, and asymmetric priors comprising the Gumbel model, which corresponds to a log-transformed Weibull distribution. All priors are implemented via local reparameterization with closed-form Kullback–Leibler (KL) divergence, preserving computational tractability. On a dataset of 54 geographically diverse SAS images comprising 3×105 300×300 pixel patches, the Gumbel prior yields 97.6% macro F1 and 0.7% expected calibration error (ECE) on ResNet20, versus 87.1% F1 and 9.7% ECE for the Gaussian baseline. On ResNet50, Gumbel achieves 99.2% accuracy, surpassing the deterministic model. Therefore, prior specification deserves domain-specific consideration in BNN deployments.

• Erdmann, Niklas; Bentsen, Lars Ødegaard; Stenbro, Roy; Riise, Heine Nygard; Warakagoda, Narada Dilp & Engelstad, Paal E. (2026). Multi-Horizon Time Series Forecasting of Non-Parametric CDFs with Deep Lattice Networks. Proceedings of the AAAI Conference on Artificial Intelligence. ISSN 2159-5399. 40(1), p. 255–264. Full text in Research Archive Show summary

  Probabilistic forecasting is not only a way to add more information to a prediction of the future, but it also builds on weaknesses in point prediction. Sudden changes in a time series can still be captured by a cumulative distribution function (CDF), while a point prediction is likely to miss it entirely. The modeling of CDFs within forecasts has historically been limited to parametric approaches, but due to recent advances, this no longer has to be the case. We aim to advance the fields of probabilistic forecasting and monotonic networks by connecting them and propose an approach that permits the forecasting of implicit, complete, and nonparametric CDFs. For this purpose, we propose an adaptation to deep lattice networks (DLN) for monotonically constrained simultaneous/implicit quantile regression in time series forecasting. Quantile regression usually produces quantile crossovers, which need to be prevented to achieve a legitimate CDF. By leveraging long short term memory units (LSTM) as the embedding layer, and spreading quantile inputs to all sub-lattices of a DLN with an extended output size, we can produce a multi-horizon forecast of an implicit CDF due to the monotonic constraintability of DLNs that prevent quantile crossovers. We compare and evaluate our approach's performance to relevant state of the art within the context of a highly relevant application of time series forecasting: Day-ahead, hourly forecasts of solar irradiance observations. Our experiments show that the adaptation of a DLN performs just as well or even better than an unconstrained approach. Further comparison of the adapted DLN against a scalable monotonic neural network shows that our approach performs better. With this adaptation of DLNs, we intend to create more interest and crossover investigations in techniques of monotonic neural networks and probabilistic forecasting.

• Zieglmeier, Sebastian Georg; Badyn, Mathias Hudoba de; Warakagoda, Narada Dilp; Krogstad, Thomas Røbekk & Engelstad, Paal E. (2025). Semi-Data-Driven Model Predictive Control: A Physics-Informed Data-Driven Control Approach. IEEE Conference on Decision and Control. Proceedings. ISSN 0743-1546. p. 2110–2117. doi: 10.1109/cdc57313.2025.11312410. Full text in Research Archive Show summary

  Data-enabled predictive control (DeePC) has emerged as a powerful technique to control complex systems without the need for extensive modeling efforts. However, relying solely on offline collected data trajectories to represent the system dynamics introduces certain drawbacks. Therefore, we present a novel semi-data-driven model predictive control (SD-MPC) framework that combines (limited) model information with DeePC to address a range of these drawbacks, including sensitivity to noisy data and a lack of robustness. In this work, we focus on the performance of DeePC in operating regimes not captured by the offline collected data trajectories and demonstrate how incorporating an underlying parametric model can counteract this issue. SD-MPC exhibits equivalent closed-loop performance as DeePC for deterministic linear time-invariant systems. Simulations demonstrate the general control performance of the proposed SD-MPC for both a linear time-invariant system and a nonlinear system modeled as a linear parameter-varying system. These results provide numerical evidence of the enhanced robustness of SD-MPC over classical DeePC.

• Orescanin, Marko; Olson, Derek; Harrington, Brian; Geilhufe, Marc; Hansen, Roy Edgar & Duvio, Dalton [Show all 7 contributors for this article] (2025). Classification of Imaging Artifacts in Synthetic Aperture Sonar With Bayesian Deep Learning. IEEE Journal of Oceanic Engineering. ISSN 0364-9059. doi: 10.1109/JOE.2025.3538948. Full text in Research Archive
• Bentsen, Lars Ødegaard; Warakagoda, Narada Dilp; Stenbro, Roy & Engelstad, Paal (2024). Relative evaluation of probabilistic methods for spatio-temporal wind forecasting. Journal of Cleaner Production. ISSN 0959-6526. 434. doi: 10.1016/j.jclepro.2023.139944. Full text in Research Archive Show summary

  Short-term wind power forecasting has become a de facto tool to better facilitate the integration of such renewable energy resources into modern power grids. Instead of point predictors, which produce single-value predictions for the expected power, probabilistic forecasts predict probability distributions over the expected power output or associated confidence intervals. In this study, three different parametric and non-parametric methods for uncertainty modelling in wind power forecasting were studied, namely quantile regression (QR), variational inference and a maximum likelihood estimation (MLE) method. Johnson’s SU distribution was studied as a novel candidate for modelling wind power, which is a transformed normal distribution that exhibits both skew and heavy tails. This was one of the first studies to provide a thorough investigation of Johnson’s SU distribution for uncertainty modelling in a complex deep learning framework for wind forecasting. It was found that Johsnon’s SU likelihood and QR-based models significantly outperformed models using Gaussian likelihoods, based on a range of quantitative metrics to evaluate probability distributions and qualitative investigation of produced forecasts. Variational inference models using Johnson’s SU likelihoods performed remarkably well, with near-perfect calibration and higher precision than models using any of the other methods for uncertainty modelling, as evaluated through the pinball loss, Average Coverage Error (ACE) and Prediction Interval Coverage Percentage (PICP) metric. With the superior performance of Johnson’s SU likelihood models, the study mainly contributes to the literature by introducing another candidate distribution for probabilistic wind forecasting, which is analytical, unbounded and easy to integrate into modern deep learning frameworks.

• Buskenes, Jo Inge; Midelfart, Herman; Midtgaard, Øivind & Warakagoda, Narada Dilp (2024). Real-Time Sonar Image Simulation for Adaptive Template Matching in Automatic Target Recognition. IEEE Journal of Oceanic Engineering. ISSN 0364-9059. 49(4), p. 1488–1500. doi: 10.1109/JOE.2024.3381390. Full text in Research Archive
• Erdmann, Niklas; Bentsen, Lars Ødegaard; Stenbro, Roy; Riise, Heine Nygard; Warakagoda, Narada Dilp & Engelstad, Paal (2024). Deep and Probabilistic Solar Irradiance Forecast at the Arctic Circle. Conference record of the Photovoltaic Specialists Conference. ISSN 0160-8371. p. 1599–1606. doi: 10.1109/PVSC57443.2024.10748862. Full text in Research Archive
• Bentsen, Lars Ødegaard; Warakagoda, Narada Dilp; Stenbro, Roy & Engelstad, Paal E. (2023). A Unified Graph Formulation for Spatio-Temporal Wind Forecasting. Energies. 16(20). doi: 10.3390/en16207179. Full text in Research Archive Show summary

  With the rapid adoption of wind energy globally, there is a need for accurate short-term forecasting systems to improve the reliability and integration of such energy resources on a large scale. While most spatio-temporal forecasting systems comprise distinct components to learn spatial and temporal dependencies separately, this paper argues for an approach to learning spatio-temporal information jointly. Many time series forecasting systems also require aligned input information and do not naturally facilitate irregular data. Research is therefore required to investigate methodologies for forecasting in the presence of missing or corrupt measurements. To help combat some of these challenges, this paper studied a unified graph formulation. With the unified formulation, a graph neural network (GNN) was used to extract spatial and temporal dependencies simultaneously, in a single update, while also naturally facilitating missing data. To evaluate the proposed unified approach, the study considered hour-ahead wind speed forecasting in the North Sea under different amounts of missing data. The framework was compared against traditional spatio-temporal architectures that used GNNs together with temporal long short-term memory (LSTM) and Transformer or Autoformer networks, along with the imputation of missing values. The proposed framework outperformed the traditional architectures, with absolute errors of around 0.73–0.90 m per second, when subject to 0–80% of missing input data. The unified graph approach was also better at predicting large changes in wind speed, with an additional 10-percentage-point improvement over the second-best model. Overall, this paper investigated a novel methodology for spatio-temporal wind speed forecasting and showed how the proposed unified graph formulation achieved competitive results compared to more traditional GNN-based architectures.

• Orescanin, Marco; Harrington, Brian; Olson, Derek R.; Geilhufe, Marc; Hansen, Roy Edgar & Warakagoda, Narada Dilp (2023). A Study on the Effect of Commonly Used Data Augmentation Techniques on Sonar Image Artifact Detection Using Deep Neural Networks. IEEE International Geoscience and Remote Sensing Symposium proceedings. ISSN 2153-6996. doi: 10.1109/IGARSS52108.2023.10282626. Full text in Research Archive Show summary

  This paper presents an empirical study that evaluates the impact of different types of augmentations on the performance of Deep Learning (DL) models for detecting imaging artifacts in Synthetic Aperture Sonar (SAS) imagery. Despite the popularity of using DL in the SAS community, the impact of augmentations that violate the geometry and physics of SAS has not been fully explored. To address this gap, we developed a unique dataset for detecting imaging artifacts in SAS imagery with DL and trained a Bayesian neural network with a ResNet architecture using widely used augmentations in DL for computer vision, as well as common augmentations used in the SAS literature. The study shows that augmentations that violate the geometry and imaging physics of SAS can negatively impact supervised classification, but can sometimes improve performance. Overall, the study provides important insights into the impact of different types of augmentations on the performance of DL models in SAS applications.

• Orescanin, Marco; Harrington, Brian; Olson, Derek R.; Geilhufe, Marc; Hansen, Roy Edgar & Warakagoda, Narada Dilp (2023). Uncertainty quantification with deep learning through variational inference with applications to synthetic aperture sonar. Underwater Acoustics Conference & Exhibition (UACE). Full text in Research Archive Show summary

  Deep learning (DL) has gained popularity in applications and research within the active sonar community both in academic and commercial settings due to the ability of such models to learn complex non-linear relationships between the input features and the labels, a data driven approach. This has led to significant improvements in automatic target recognition and seafloor texture understanding with Synthetic Aperture Sonar (SAS). Most of the DL models reported in literature are deterministic and do not provide estimates of uncertainty of their predictions limiting the utility for the downstream tasks such as ATR and change detection.\n\nIn this work we demonstrate the ability to quantify uncertainty in deep learning predictions by utilizing variational inference to develop Bayesian Neural Networks. Further, we explore decomposition of obtained uncertainty into aleatoric and epistemic components.We showcase the importance of this decomposition on classifier performance and interpretation of the results. We introduce and compare several state-of-the art methods in variational inference on the task of classifying imaging artifacts in SAS. We conduct this on a novel dataset developed for this classification task through introduction of physical perturbations in the image formation stage, namely : 1) sound speed error of 40 m/s, 2) navigation error through perturbation in yaw of 0.35° and 3) Gaussian noise over the imaging channels prior to pulse compression (lowering the average\nimage SNR to 5 dB). \n\nOverall, we demonstrate that our best model, a mean-field variational inference via flipout ResNet architecture, achieves 92% accuracy with calibrated uncertainty. By rejecting 10% of the data with highest uncertainty we achieve additional 4% improvement in accuracy.

• Bentsen, Lars Ødegaard; Warakagoda, Narada Dilp; Stenbro, Roy & Engelstad, Paal (2022). Wind Park Power Prediction: Attention-Based Graph Networks and Deep Learning to Capture Wake Losses. Journal of Physics: Conference Series (JPCS). ISSN 1742-6588. 2265. doi: 10.1088/1742-6596/2265/2/022035. Full text in Research Archive Show summary

  Abstract With the increased penetration of wind energy into the power grid, it has become increasingly important to be able to predict the expected power production for larger wind farms. Deep learning (DL) models can learn complex patterns in the data and have found wide success in predicting wake losses and expected power production. This paper proposes a modular framework for attention-based graph neural networks (GNN), where attention can be applied to any desired component of a graph block. The results show that the model significantly outperforms a multilayer perceptron (MLP) and a bidirectional LSTM (BLSTM) model, while delivering performance on-par with a vanilla GNN model. Moreover, we argue that the proposed graph attention architecture can easily adapt to different applications by offering flexibility into the desired attention operations to be used, which might depend on the specific application. Through analysis of the attention weights, it was showed that employing attention-based GNNs can provide insights into what the models learn. In particular, the attention networks seemed to realise turbine dependencies that aligned with some physical intuition about wake losses.

• Bentsen, Lars Ødegaard; Warakagoda, Narada Dilp; Stenbro, Roy & Engelstad, Paal (2022). Probabilistic Wind Park Power Prediction using Bayesian Deep Learning and Generative Adversarial Networks. Journal of Physics: Conference Series (JPCS). ISSN 1742-6588. 2362. doi: 10.1088/1742-6596/2362/1/012005. Full text in Research Archive Show summary

  The rapid depletion of fossil-based energy supplies, along with the growing reliance on renewable resources, has placed supreme importance on the predictability of renewables. Research focusing on wind park power modelling has mainly been concerned with point estimators, while most probabilistic studies have been reserved for forecasting. In this paper, a few different approaches to estimate probability distributions for individual turbine powers in a real off-shore wind farm were studied. Two variational Bayesian inference models were used, one employing a multilayered perceptron and another a graph neural network (GNN) architecture. Furthermore, generative adversarial networks (GAN) have recently been proposed as Bayesian models and was here investigated as a novel area of research. The results showed that the two Bayesian models outperformed the GAN model with regards to mean absolute errors (MAE), with the GNN architecture yielding the best results. The GAN on the other hand, seemed potentially better at generating diverse distributions. Standard deviations of the predicted distributions were found to have a positive correlation with MAEs, indicating that the models could correctly provide estimates on the confidence associated with particular predictions.

• Bentsen, Lars Ødegaard; Warakagoda, Narada Dilp; Stenbro, Roy & Engelstad, Paal (2022). Spatio-temporal wind speed forecasting using graph networks and novel Transformer architectures. Applied Energy. ISSN 0306-2619. 333. doi: 10.1016/j.apenergy.2022.120565. Full text in Research Archive Show summary

  To improve the security and reliability of wind energy production, short-term forecasting has become of utmost importance. This study focuses on multi-step spatio-temporal wind speed forecasting for the Norwegian continental shelf. In particular, the study considers 14 offshore measurement stations and aims to leverage spatial dependencies through the relative physical location of different stations to improve local wind forecasts and simultaneously output different forecasts for each of the 14 locations. Our multi-step forecasting models produce either 10-minute, 1- or 4-hour forecasts, with 10-minute resolution, meaning that the models produce more informative time series for predicted future trends. A graph neural network (GNN) architecture was used to extract spatial dependencies, with different update functions to learn temporal correlations. These update functions were implemented using different neural network architectures. One such architecture, the Transformer, has become increasingly popular for sequence modelling in recent years. Various alterations have been proposed to better facilitate time series forecasting, of which this study focused on the Informer, LogSparse Transformer and Autoformer. This is the first time the LogSparse Transformer and Autoformer have been applied to wind forecasting and the first time any of these or the Informer have been formulated in a spatio-temporal setting for wind forecasting. By comparing against spatio-temporal Long Short-Term Memory (LSTM) and Multi-Layer Perceptron (MLP) models, the study showed that the models using the altered Transformer architectures as update functions in GNNs were able to outperform these. Furthermore, we propose the Fast Fourier Transformer (FFTransformer), which is a novel Transformer architecture based on signal decomposition and consists of two separate streams that analyse the trend and periodic components separately. The FFTransformer and Autoformer were found to achieve superior results for the 10-minute and 1-hour ahead forecasts, with the FFTransformer significantly outperforming all other models for the 4-hour ahead forecasts. Our code to implement the different models are made publicly available at: https://github.com/LarsBentsen/FFTransformer.

• Bryan, Oscar; Hansen, Roy Edgar; Haines, Tom S. F.; Warakagoda, Narada Dilp & Hunter, Alan Joseph (2022). Challenges of Labelling Unknown Seabed Munition Dumpsites from Acoustic and Optical Surveys: A Case Study at Skagerrak. Remote Sensing. doi: 10.3390/rs14112619. Full text in Research Archive Show summary

  The disposal of unexploded ordnance (UXOs) at sea is a global problem. The mapping and remediation of historic UXOs can be assisted by autonomous underwater vehicles (AUVs) carrying sensor payloads such as synthetic aperture sonar (SAS) and optical cameras. AUVs can image large areas of the seafloor in high resolution, motivating an automated approach to UXO detection. Modern methods commonly use supervised machine learning which requires labelled examples from which to learn. This work investigates the often-overlooked labelling process and resulting dataset using an example historic UXO dumpsite at Skagerrak. A counterintuitive finding of this work is that optical images cannot be relied on for ground truth as a significant number of UXOs visible in SAS images are not in optical images, presumed buried. Given the lack of ground truth, we use an ordinal labelling scheme to incorporate a measure of labeller uncertainty. We validate this labelling regime by quantifying label accuracy compared to optical labels with high confidence. Using this approach, we explore different taxonomies and conclude that grouping objects into shells, bombs, debris, and natural gave the best trade-off between accuracy and discrimination.

• Gunathilake, Miyuru Bandara; Panditharathne, Pushpika; Gunathilake, Anura Shantha & Warakagoda, Narada Dilp (2020). Application of a hec-hms model on event-based simulations in a tropical watershed. Engineering and Applied Science Research (EASR). ISSN 2539-6161. 47(4), p. 349–360. doi: 10.14456/easr.2020.38. Full text in Research Archive
• Vedeler, Alexandra Skau & Warakagoda, Narada Dilp (2020). Generative Adversarial Immitation Learning for Steering an Unmanned Surface Vehicle. Proceedings of the Northern Lights Deep Learning Workshop. 1. doi: 10.7557/18.5147. Full text in Research Archive Show summary

  The task of obstacle avoidance using maritime vessels, such as Unmanned Surface Vehicles (USV), has traditionally been solved using specialized modules that are designed and optimized separately. However, this approach requires a deep insight into the environment, the vessel, and their complex dynamics. We propose an alternative method using Imitation Learning (IL) through Deep Reinforcement Learning (RL) and Deep Inverse Reinforcement Learning (IRL) and present a system that learns an end-to-end steering model capable of mapping radar-like images directly to steering actions in an obstacle avoidance scenario. The USV used in the work is equipped with a Radar sensor and we studied the problem of generating a single action parameter, heading. We apply an IL algorithm known as generative adversarial imitation learning (GAIL) to develop an end-to-end steering model for a scenario where avoidance of an obstacle is the goal. The performance of the system was studied for different design choices and compared to that of a system that is based on pure RL. The IL system produces results that indicate it is able to grasp the concept of the task and that in many ways are on par with the RL system. We deem this to be promising for future use in tasks that are not as easily described by a reward function.

• Warakagoda, Narada Dilp; Dirdal, Johann Alexander & Johnsen, Erlend Faxvaag (2019). Fusion of LiDAR and Camera Images in End-to-end Deep Learning for Steering an Off-road Unmanned Ground Vehicle. In Shahbazian, Elisa & Llinas, James (Ed.), 22nd International Conference on Information Fusion. IEEE (Institute of Electrical and Electronics Engineers). ISSN 9780996452793. Full text in Research Archive
• Midtgaard, Øivind; Warakagoda, Narada Dilp; Davies, Gary; Connors, Warren A. & Geilhufe, Marc (2019). Multi-phase performance evaluation for modern minehunting systems. Proceedings of SPIE, the International Society for Optical Engineering. ISSN 0277-786X. doi: 10.1117/12.2518837. Full text in Research Archive Show summary

  Many NATO navies are in the process of replacing their dedicated minehunting vessels with systems of heterogeneous, unmanned modules. While traditional ship-based assets prosecute sonar contacts in sequence through to neutralisation, modern systems employ unmanned vehicles equipped with side-looking sonar to detect and classify minelike contacts in a full area segment before proceeding with contact identification and mine neutralisation. This shift in technology and procedure brings important operational advantages, but also introduces a need to modify the traditional minehunting performance evaluation based on the percentage clearance metric. Previous works have demonstrated that the achieved detection and classification performance of modern minehunting systems can be estimated from the collected sonar data (through-the-sensor) and reported as detailed geographical maps. This paper extends the map-based evaluation approach to the identification and neutralisation phases, and also includes the case where some of the contacts or mines intentionally are left unprosecuted, e.g. disposal of only the specific mines required for establishing a safe sailing route. Each map cell is assumed to be sufficiently small to contain at most one sonar contact and can thus be assigned a status based on the hunting results for that cell: minelike contact, identified mine, etc. To this end we derive Bayesian formulations of a new performance metric: the probability of a remaining mine in a given cell. Furthermore, we show that this metric provides consistent multi-phase performance evaluation and estimates of the mine impact risk for a follow-on ship transiting a specified route.

• Warakagoda, Narada Dilp & Midtgaard, Øivind (2018). Transfer-learning with deep neural networks for mine recognition in sonar images. Proceedings of the Institute of Acoustics. ISSN 1478-6095. 40, p. 115–122. Full text in Research Archive
• Husøy, T; Pettersen, M; Nilsson, B; Öberg, T.; Warakagoda, Narada Dilp & Lie, Arne (2011). Implementation of an underwater acoustic modem with network capability, OCEANS, 2011 IEEE - Spain : Santander 6-9 June 2011. IEEE (Institute of Electrical and Electronics Engineers). ISSN 9781457700866. doi: 10.1109/oceans-spain.2011.6003642. Full text in Research Archive Show summary

  This paper introduces the underwater acoustic modem as implemented within the UAN – Underwater Acoustic Network project. The low power modem has implemented turbo equalization algorithms in addition to variable spread rate direct sequence spread spectrum signaling. The network layer implemented on the modem support automatic network discovery, multi hop routing and support for mobile nodes, and when expanded with a single board computer via serial line it supports IP connectivity end-to-end. Experimental results from sea trials are presented.

View all works in NVA

• Midtgaard, Øivind; Sæbø, Torstein Olsmo & Warakagoda, Narada Dilp (2023). Challenges of automated change detection in repeat-pass synthetic aperture sonar imagery. Full text in Research Archive
• Orescanin, Marco; Harrington, Brian; Olson, Derek R.; Geilhufe, Marc; Hansen, Roy Edgar & Warakagoda, Narada Dilp (2023). Uncertainty quantification with deep learning through variational inference with applications to synthetic aperture sonar. Full text in Research Archive
• Orescanin, Marko; Harrington, Brian; Olson, Derek R.; Geilhufe, Marc; Hansen, Roy Edgar & Warakagoda, Narada Dilp (2023). A Study on the Effect of Commonly Used Data Augmentation Techniques on Sonar Image Artifact Detection Using Deep Neural Networks. Full text in Research Archive
• Midtgaard, Øivind; Sæbø, Torstein Olsmo & Warakagoda, Narada Dilp (2023). Challenges of Automated Change Detection in Repeat-pass Synthetic Aperture Sonar Imagery. Full text in Research Archive Show summary

  Change detection extracts regions of interest as temporal differences by imaging a scene before and after changes may have occurred. We denote the initial image as the reference and the more recent as the current image. Automated tools for image comparison are a prerequisite for many change detection applications, because the large data rates and detailed, yet repetitive contents of the imagery make full manual analysis tedious and prone to errors. Although automated change detection (ACD) in repeat-pass imagery is a well-established technique for synthetic aperture radar (SAR) 1, it is far less mature for underwater sonar due to inherent properties of both traditional side-scan sonar (SSS) imaging and the complex ocean environment. These properties impede for example sensor trajectory control, data positioning accuracy and image resolution. The increasing use of autonomous underwater vehicles (AUVs) equipped with advanced synthetic aperture sonar (SAS) and aided inertial navigation systems (AINS) has partly remedied these limitations and facilitated the development of ACD methods for the sonar domain 2-5. Still, important challenges remain, in particular regarding scene stability, sensing consistency and image alignment between surveys.

• Schøyen, Vemund Sigmundson; Hagen, Per Espen; Midtgaard, Øivind & Warakagoda, Narada Dilp (2022). Automatic Detection of Seafloor Pipelines with Deep Learning. Hydro International. ISSN 1385-4569. 26(1), p. 10–13. Full text in Research Archive
• Holen, Martin; Ruud, Else-Line Malene; Warakagoda, Narada Dilp; Granmo, Ole-Christoffer; Engelstad, Paal E. & Knausgård, Kristian Muri (2022). Towards Using Reinforcement Learning for Autonomous Docking of Unmanned Surface Vehicles. doi: 10.1007/978-3-031-08223-8_38. Full text in Research Archive Show summary

  Providing full autonomy to Unmanned Surface Vehicles (USV) is a challenging goal to achieve. Autonomous docking is a subtask that is particularly difficult. The vessel has to distinguish between obstacles and the dock, and the obstacles can be either static or moving. This paper developed a simulator using Reinforcement Learning (RL) to approach the problem. We studied several scenarios for the task of docking a USV in a simulator environment. The scenarios were defined with different sensor inputs and start-stop procedures but a simple shared reward function. The results show that the system solved the task when the IMU (Inertial Measurement Unit) and GNSS (Global Navigation Satellite System) sensors were used to estimate the state, despite the simplicity of the reward function.

• Schøyen, Vemund Sigmundson; Warakagoda, Narada Dilp & Midtgaard, Øivind (2021). Seafloor Pipeline Detection With Deep Learning. Full text in Research Archive
• Vedeler, Alexandra Skau & Warakagoda, Narada Dilp (2020). Generative Adversarial Immitation Learning for Steering an Unmanned Surface Vehicle. Full text in Research Archive
• Warakagoda, Narada Dilp & Midtgaard, Øivind (2020). Retrieval of similar targets in synthetic aperture sonar images with deep learning. Proceedings of Meetings on Acoustics (POMA). 40. doi: 10.1121/2.0001315. Full text in Research Archive
• Warakagoda, Narada Dilp & Midtgaard, Øivind (2020). Finding similar targets in synthetic aperture sonar images with deep learning. Full text in Research Archive
• Midtgaard, Øivind; Warakagoda, Narada Dilp; Davies, Gary; Connors, Warren A. & Geilhufe, Marc (2019). Multi-phase performance evaluation for modern minehunting systems. Full text in Research Archive
• Warakagoda, Narada Dilp & Midtgaard, Øivind (2018). Fine-tuning of imagenet-trained deep neural networks for seafloor mine recognition in sonar images. Full text in Research Archive
• Warakagoda, Narada Dilp & Johnsen, Magne Hallstein (2001). Speaker adaption in an ASR system based on nonlinear dynamics. Full text in Research Archive
• Warakagoda, Narada Dilp & Johnsen, Magne Hallstein (2001). Nonlinear dynamical system based acoustic modeling for ASR. Full text in Research Archive
• Warakagoda, Narada Dilp & Johnsen, Magne Hallstein (1999). Neural network based optimal feature extraction for ASR. Full text in Research Archive Show summary

  The procedure of calculating Mel Frequency based Cepstral Coefficient s (MFCC) is shown to resemble a three layer Multilayer Perceptron (ML P) like structure. Such an MLP is employed as a preprocessor in a hyb rid HMM-MLP system, and the possibility of optimizing the whole syste m as a single entity, with respect to a suitable criterion, is pointe d out. The system, together with the Maximal Mutual Information (MMI) criterion was tested on a speaker independent, five broad class, isl oated phoneme recognition task. The reported results of these prelimi nary experiments clearly indicate the advantage of optimizable prepro cessing.

• Warakagoda, Narada Dilp & Johnsen, Magne Hallstein (1999). Multilayer perceptron based optimal state spaced recognition for time series modelling. Full text in Research Archive Show summary

  Estimating the nonlinear dynamicial system which has generated a give n time series is considered. Traditionally this is achieved via two s eparate steps: first projecting the signal frames onto the state spac e and then estimating the predictive relationship between successive state vectors. In the method presented here, these two steps are perf ormed simultaneously, where both the projection and the prediction fu nctions are represented by MLPs. In addition, a measurement function, again represented by an MLP is used. This MLP maps each state vector to a sample of the given signal. The original estimation problem is then interpreted as a training (optimization) problem, and a backprop agation type training is proposed for solving it. The method proposed is tested both with an artificial time series produced by the Henon map and a speech signal.Performance is compared with results from pop ular traditional methods, which indicates a clear superiority for our new proposed method.

• Warakagoda, Narada Dilp & Johnsen, Magne Hallstein (1997). Dynamical system models for speech recognition. Full text in Research Archive Show summary

  An approach to Automatic Speech Recognition (ASR) based on the theory of Nonlinear Dynamical systems is presented. The speech signal at a given tim considered as a scalar measurement of a the state of an unk nown multidimensional nonlinear dynamical system. Thus speech models with a small numbers of parameters are possible. Further, the concept of synchronization is introduced in order to perform robust nonlinea r prediction, i.e. with capasity to handle the highly variable nature of speech.

View all works in NVA