Pētniecība
Publikācijas
729 publikācijas
- 2022
Construction and benchmark of an autonomous tracked mobile robot system(atveras jaunā logā)
Ahluwalia Vaibhav, Arents Janis, Oraby Aly, Greitans Modris
Robotic Systems and Applications, Sēj. 2 (1), lpp. 15-28, JVE Journals
Kopsavilkums
Robots require a certain set of skills to perceive and analyse the environment and act accordingly. For tracked mobile robots getting good odometry data from sensory information is a challenging key prerequisite to perform in an unstructured dynamic environment, thus an essential issue in the tracked mobile robotics domain. In this article, we construct a ROS-based tracked mobile robot system taking the Jaguar V4 mobile robot as the base platform. On which several visual odometry solutions based on different cameras and methods (Intel RealSense T265, Zed camera, RTAB-Map RGBD) are integrated and benchmark comparison is performed. Analysis of new challenges faced by different methods while applied on a tracked vehicle as well as recommendations and conclusions are presented. Intel RealSense T265 solution proved to perform well in uncertain conditions which involves bounded vibrations and low lighting conditions with low latency, which result in good map generation. Further evaluations with a path planning algorithm and Intel RealSense T265 were conducted to test the effect of the robot’s motion profiles on odometry data accuracy.
- 2022
Improving Semantic Segmentation of Urban Scenes for Self-Driving Cars with Synthetic Images(atveras jaunā logā)
M. Ivanovs, K. Ozols, A. Dobrajs, R. Kadikis
Sensors, Sēj. 22 (6), lpp. 2252, MDPI
Kopsavilkums
Semantic segmentation of an incoming visual stream from cameras is an essential part of the perception system of self-driving cars. State-of-the-art results in semantic segmentation have been achieved with deep neural networks (DNNs), yet training them requires large datasets, which are difficult and costly to acquire and time-consuming to label. A viable alternative to training DNNs solely on real-world datasets is to augment them with synthetic images, which can be easily modified and generated in large numbers. In the present study, we aim at improving the accuracy of semantic segmentation of urban scenes by augmenting the Cityscapes real-world dataset with synthetic images generated with the open-source driving simulator CARLA (Car Learning to Act). Augmentation with synthetic images with a low degree of photorealism from the MICC-SRI (Media Integration and Communication Center–Semantic Road Inpainting) dataset does not result in the improvement of the accuracy of semantic segmentation, yet both MobileNetV2 and Xception DNNs used in the present study demonstrate a better accuracy after training on the custom-made CCM (Cityscapes-CARLA Mixed) dataset, which contains both real-world Cityscapes images and high-resolution synthetic images generated with CARLA, than after training only on the real-world Cityscapes images. However, the accuracy of semantic segmentation does not improve proportionally to the amount of the synthetic data used for augmentation, which indicates that augmentation with a larger amount of synthetic data is not always better.
Scopus procentile: 91 · SNIP: -2.445
- 2022
Bin-Picking Solution for Randomly Placed Automotive Connectors Based on Machine Learning Techniques(atveras jaunā logā)
Pedro Torres, Janis Arents, Hugo Marques, Paulo Marques
Electronics 2022, Sēj. 11 (2), lpp. 476, MDPI
Kopsavilkums
This paper presents the development of a bin-picking solution based on low-cost vision systems for the manipulation of automotive electrical connectors using machine learning techniques. The automotive sector has always been in a state of constant growth and change, which also implies constant challenges in the wire harnesses sector, and the emerging growth of electric cars is proof of this and represents a challenge for the industry. Traditionally, this sector is based on strong human work manufacturing and the need arises to make the digital transition, supported in the context of Industry 4.0, allowing the automation of processes and freeing operators for other activities with more added value. Depending on the car model and its feature packs, a connector can interface with a different number of wires, but the connector holes are the same. Holes not connected with wires need to be sealed, mainly to guarantee the tightness of the cable. Seals are inserted manually or, more recently, through robotic stations. Due to the huge variety of references and connector configurations, layout errors sometimes occur during seal insertion due to changed references or problems with the seal insertion machine. Consequently, faulty connectors are dumped into boxes, piling up different types of references. These connectors are not trash and need to be reused. This article proposes a bin-picking solution for classification, selection and separation, using a two-finger gripper, of these connectors for reuse in a new operation of removal and insertion of seals. Connectors are identified through a 3D vision system, consisting of an Intel RealSense camera for object depth information and the YOLOv5 algorithm for object classification. The advantage of this approach over other solutions is the ability to accurately detect and grasp small objects through a low-cost 3D camera even when the image resolution is low, benefiting from the power of machine learning algorithms.
Indeksēts Scopus / WoS · Scopus procentile: 60 · SNIP: 1.013
- 2022
Scenario simulation of a small-scale energy community management(atveras jaunā logā)
R. Lazdins and A. Mutule
2022 IEEE 63th International Scientific Conference on Power and Electrical Engineering of Riga Technical University (RTUCON), lpp. 1-5
- 2022
Synthetic Data of Randomly Piled, Similar Objects for Deep Learning-Based Object Detection(atveras jaunā logā)
Janis Arents, Bernd Lesser, Andis Bizuns, Roberts Kadikis, Elvijs Buls, Modris Greitans
ICIAP 2022, Sēj. 13232, lpp. 706–717, Springer
Kopsavilkums
Currently, the best object detection results are achieved by supervised deep learning methods, however, these methods depend on annotated training data. With the synthetic data generation approach, we intend to mimic the real data characteristics and diversify the dataset by a systematic rendering of highly realistic synthetic pictures. We systematically explore how different combinations and portions of real and synthetic datasets affect object detectors performance. The developed synthetic data generation framework shows promising results in deep learning-based object detection tasks and can supplement real data when the variety of real training data is insufficient. However, when synthetic data ratio increases over real data ratio, a decrease in average precision can be observed, which has the most affect on 0.75-0.95 IoU threshold range.
- 2022
Unintentional Force Drifts as Consequences of Indirect Force Control with Spatial Referent Coordinates(atveras jaunā logā)
Valters Abolins, Mark L. Latash
NEUROSCIENCE, Sēj. 481, lpp. 156-165, Elsevier
Kopsavilkums
We explored the phenomenon of unintentional force drifts in the absence of visual feedback. Based on the idea of direct force control with internal models and on the idea of indirect force control with referent coordinates to the involved muscle groups, contrasting predictions were drawn for changes in the drift magnitude when acting against external spring loads. Fifteen young subjects performed typical accurate force production tasks by pressing with the Index finger at 20% of maximal voluntary contraction (MVC) in isometric conditions and while acting against one of the three external springs with different stiffness. The visual feedback on the force was turned off after 5 s. At the end of each 20-s trial, the subjects relaxed and then tried to reproduce the final force level. The force drifts were significantly smaller in the spring conditions, particularly when acting against more compliant springs. The subjects were unaware of the force drifts and, during force matching, produced forces close to the initial force magnitude, which were not different across the conditions. There was a trend toward larger drifts during performance by the dominant hand. We view these observations as strong arguments in favor of the theory of control with spatial referent coordinates. In particular, force drifts were likely consequences of drifts of referent coordinates to both agonist and antagonist muscles. The lack of drift effects on both perception-to-report and perception-to-act fit the scheme of kinesthetic perception based on the interaction of efferent (referent coordinate) and afferent processes.
Indeksēts Scopus / WoS · Scopus procentile: 67 · SNIP: 0.967
- 2022
Unintentional force drifts across the human fingers: implications for the neural control of finger tasks(atveras jaunā logā)
Valters Abolins, Mark L. Latash
Experimental Brain Research, Sēj. 240 (3), lpp. 751-761, Springer Nature
Kopsavilkums
We explored the unintentional force drift across the four fingers of the dominant hand during accurate force production in isometric conditions caused by turning the visual feedback on force off. Our hypotheses were that the Index finger would show smallest drifts and best ability to eliminate the drifts with knowledge of performance in previous trials. Young healthy subjects produced force at 20% of the maximal force level by one finger at a time. There was no significant difference among the fingers in the root mean square error of force during performance with visual feedback. Turning visual feedback off caused force drift to lower magnitudes. The magnitude of force drift was the largest during tasks performed by the Index finger. After each block of twelve trials, the subjects were given feedback on the drift magnitude in that block and used it to correct performance in future trials. There was a total of six blocks. The magnitude of drift correction between consecutive blocks correlated with the magnitude of drift in the earlier block only after the second and fourth blocks. The Index finger failed to improve its performance more than other fingers and demonstrated significant residual drifts to lower force magnitudes in the sixth block of trials. These findings falsified both our hypotheses. Taken together with earlier studies showing advantage of the Index finger across a variety of tasks that require quick and accurate changes in performance, our results suggest that effector specialization along the stability-agility continuum is not limited to the phenomenon of cortical arm/hand dominance but can also be seen across fingers of a hand controlled by the same hemisphere, possibly reflecting the differences in the finger role in prehensile tasks.
Indeksēts Scopus / WoS · Scopus procentile: 36 · SNIP: 0.941
- 2022
Signal Loss in Body Coupled Communication: Guide for Accurate Measurements(atveras jaunā logā)
Juris Ormanis, Vladislavs Medvedevs, Valters Āboliņš, Gatis Gaigals, Atis Elsts
Workshop on Benchmarking Cyber-Physical Systems and Internet of Things (CPS-IoTBench), IEEE
Kopsavilkums
Body Coupled Communication (BCC) is a growing research area that promises to deliver more secure and more energy-efficient body-area wireless sensor networks. Existing research articles on BCC show high variance in their experimental results, indicating potential methodological problems in the published work. We perform analysis of several potential methodological errors in the measurements, for instance, impedance mismatch, cable length issues, and not separating the ground planes. Our experimental results quantify the impact of these issues. Finally, we discuss the best practices for measurements, as well as their limitations.
Indeksēts Scopus / WoS
- 2022
Sim2Real image translation to improve a synthetic dataset for a bin picking task(atveras jaunā logā)
Diana Duplevska, Maksims Ivanovs, Janis Arents, Roberts Kadikis
2022 IEEE EFTA, IEEE
Kopsavilkums
The use of synthetic data is a promising solution to the problem of the availability of real data needed for the development of robotic systems. However, the precision of the systems trained on synthetic data tends to decrease when they are deployed in the real-life scenarios, which happens due to the disparities between the artificial data and the real world. Therefore, efficient methods for Sim2Real translation are much needed for further progress in robotics. In our study, we use Generative Adversarial Networks (GANs) to generate more photorealistic data from the synthetic data created for training deep neural networks to handle a bin picking task. As a result, object detectors trained on the images improved with the GANs with the optimal approach to image translation demonstrate better performance than their counterparts trained on the original synthetic data.
- 2022
A Motion Capture and Imitation Learning Based Approach to Robot Control(atveras jaunā logā)
Racinskis, Peteris, Janis Arents, Modris Greitans
Applied Sciences, Sēj. 12 (12), lpp. 7186, MDPI
Kopsavilkums
Imitation learning is a discipline of machine learning primarily concerned with replicating observed behavior of agents known to perform well on a given task, collected in demonstration data sets. In this paper, we set out to introduce a pipeline for collecting demonstrations and training models that can produce motion plans for industrial robots. Object throwing is defined as the motivating use case. Multiple input data modalities are surveyed, and motion capture is selected as the most practicable. Two model architectures operating autoregressively are examined—feedforward and recurrent neural networks. Trained models execute throws on a real robot successfully, and a battery of quantitative evaluation metrics is proposed. Recurrent neural networks outperform feedforward ones in most respects, but this advantage is not universal or conclusive. The data collection, pre-processing and model training aspects of our proposed approach show promise, but further work is required in developing Cartesian motion planning tools before it is applicable in production applications.
Indeksēts Scopus / WoS · Q1/Q2 žurnāls · Scopus procentile: 74 · SNIP: 0.977
- 2022
Transmission Power Influence on WSN-Based Indoor Localization Efficiency(atveras jaunā logā)
Amr Elkenawy, Janis Judvaitis
MDPI Sensors, Sēj. 22 (11), lpp. 4154, MDPI
Kopsavilkums
Performing a good-quality indoor localization of a mobile target is a challenging task, which can be affected by many factors such as radio wave behavior, the nature of the experimental environment, and available infrastructure. (1) Background: An indoor localization experiment using an Internet of Things (IoT) wireless sensor network (WSN) testbed is performed, in order to study the influence of transmission power level on the quality of received signals, and consequently, the estimated target positional coordinates. (2) Methods: A received signal strength indicator (RSSI) range-based localization system using a geometrical constrained weighted least squares (WLS) estimator multilateration technique is selected to validate the influence of the transmission power level on the performance of the localization algorithm. (3) Results: Fair localization quality was obtained at the highest transmission power level instead of the proposed transmission power level. (4) Conclusion: Additional factors are discussed to fully represent the required operational transmission power for a better localization quality, along with suggested improvements of the infrastructure configuration as a future work.
Scopus procentile: 91 · SNIP: 1.42
- 2022
Transient phenomena analysis of peak detector for nanosecond pulse amplitude measurement.
Vladimir Aristov, Armands Ancans.
Automatic Control and Computer Sciences, lpp. 37-47, Allerton Press
Kopsavilkums
For the classical peak detector circuit, built on a modern element base and designed to measure the amplitude of pulses of nanosecond duration, mathematically, using the Laplace transform, an equivalent circuit is described, which additionally includes parasitic capacitors and inductances. Based on the results of the analysis of the characteristics of the equivalent circuit, recommendations are made to increase the measurement accuracy when it is possible to reduce the limit of the duration of the measured pulses.
Scopus procentile: 32 · SNIP: 0.657
- 2022
Denoising Diffusion for Sampling SAT Solutions(atveras jaunā logā)
Karlis Freivalds, Sergejs Kozlovičs
NeurIPS 2022 Workshop on Score-Based Methods
Kopsavilkums
Generating diverse solutions to the Boolean Satisfiability Problem (SAT) is a hard computational problem with practical applications for testing and functional verification of software and hardware designs. We explore the way to generate such solutions using Denoising Diffusion coupled with a Graph Neural Network to implement the denoising function. We find that the obtained accuracy is similar to the currently best purely neural method and the produced SAT solutions are highly diverse, even if the system is trained with non-random solutions from a standard solver.
- 2022
Unsupervised Training for Neural TSP Solver(atveras jaunā logā)
Elīza Gaile, Andis Draguns, Emīls Ozoliņš, Kārlis Freivalds
The 16th Learning and Intelligent Optimization Conference, Sēj. 13621, lpp. 334 - 346, Springer Nature
Kopsavilkums
There has been a growing number of machine learning methods for approximately solving the travelling salesman problem. However, these methods often require solved instances for training or use complex reinforcement learning approaches that need a large amount of tuning. To avoid these problems, we introduce a novel unsupervised learning approach. We use a relaxation of an integer linear program for TSP to construct a loss function that does not require correct instance labels. With variable discretization, its minimum coincides with the optimal or near-optimal solution. Furthermore, this loss function is differentiable and thus can be used to train neural networks directly. We use our loss function with a Graph Neural Network and design controlled experiments on both Euclidean and asymmetric TSP. Our approach has the advantage over supervised learning of not requiring large labelled datasets. In addition, the performance of our approach surpasses reinforcement learning for asymmetric TSP and is comparable to reinforcement learning for Euclidean instances. Our approach is also more stable and easier to train than reinforcement learning.
Scopus procentile: 50 · SNIP: 0.534
- 2022
Electrode Comparison for Heart Rate Detection via Bioimpedance Measurements(atveras jaunā logā)
Didzis Lapsa, Rims Janeliukštis, Atis Elsts
Workshop on Microwave Theory and Techniques in Wireless Communication (MTTW'22), IEEE
Kopsavilkums
Compact wrist-worn devices that can continuously measure human bio-parameters such as heart rate offers valuable health information and facilitate healthcare outside clinical settings. Bioimpedance is a promising alternative sensing modality for vital signs monitoring in wearable devices but requires close proximity to the subject’s skin. The current study evaluates the suitability of different types of electrodes for on-wrist heart-rate measurements via bioimpedance. We compare copper, gold, and stainless steel electrodes with smooth and bumpy surfaces. To this end, we develop a signal-to-heart-rate data processing pipeline and a SNR-based metric that algorithmically evaluates the signal’s quality. The results show that while there is a clear trend in the data that supports the selection of stainless steel over other metals and bumpy over smooth electrode surfaces, most of the variance in the signal quality comes from factors other than the electrode type.
- 2022
Forest Stand Volume Estimation by Species from Sentinel-2 and LiDAR Data Using Regression Models(atveras jaunā logā)
Juris Siņica-Siņavskis, Gunta Grūbe
2022 18th Biennial Baltic Electronics Conference (BEC), lpp. 1-5, IEEE
Kopsavilkums
The paper presents a novel forest stand volume (FSV) estimation approach based on remote sensing (RS) data when the forest inventory data used as reference are limited. The proposed approach consists of several steps, such as filtering of existing inventory data, identifying individual tree tops from the Canopy Height Model (CHM), classifying dominant tree species from Sentinel-2 data, and creating the polynomial regression model for stand volume estimation based on training data. The study area was located in the Zemgale region of Southeast Latvia, where the dominant tree species are Scots pine (Pinus sylvestris L.), Norway spruce (Picea abies (L.) Karst.), birch (Betula pendula Roth, Betula pubescens Ehrh.) and black alder (Alnus glutinosa (L.) Gaertn.). The FSV (m3/ha) for each dominant species was estimated, and the accuracy against the harvester data was evaluated by calculating the root mean square error (RMSE). Additionally, a cross-validation was performed using sparse and partially imprecise inventory data, and the RMSE errors were less than 20% for pine, 22% for spruce, 28% for birch, and 23% for black alder. In general, the developed approach can be used with species for which there is a sufficient number of inventory compartments in the analysis region where these species dominate. The proposed approach can be used in automatic workflows estimating forest inventory parameters from RS data
- 2022
Mathematical Model and Synthetic Data Generation for Infra-Red Sensors(atveras jaunā logā)
Laura Leja, Vitālijs Purlans, Rihards Novickis, Andrejs Cvetkovs, Kaspars Ozols.
Sensors, Sēj. 22 (23), lpp. 9458, MDPI
Kopsavilkums
A key challenge in further improving infrared (IR) sensor capabilities is the development of efficient data pre-processing algorithms. This paper addresses this challenge by providing a mathematical model and synthetic data generation framework for an uncooled IR sensor. The developed model is capable of generating synthetic data for the design of data pre-processing algorithms of uncooled IR sensors. The mathematical model accounts for the physical characteristics of the focal plane array, bolometer readout, optics and the environment. The framework permits the sensor simulation with a range of sensor configurations, pixel defectiveness, non-uniformity and noise parameters.
Scopus procentile: 91 · SNIP: 1.42
- 2022
Digital Twins and AI in Smart Motion Control Applications(atveras jaunā logā)
Martin Cech, Ared-Jan Beltman, Kaspars Ozols
IEEE International Conference on Emerging Technologies and Factory Automation, ETFA, IEEE
Kopsavilkums
Recently, smart system integration was identified as a key competence for optimizing machines and robots. However, when one wants to 'tune' the entire production process a step further is necessary. We should evaluate performance indicators (e.g. energy and material consumption) over the whole machine life cycle in order to align the production with circular economy principles. To reach that target MBSE (model-based system engineering) should be covered by advanced digital twin approaches which allow continuous monitoring of machine performance, predict the failures and maintenance. Moreover, artificial intelligence and machine learning must be used to process big data sets gathered from the production lines. This paper identifies a common set of technologies and building blocks suitable to solve above mentioned problems for a large variety of industrial domains (semiconductor production, health-care robotics, CNC1 machining, high-speed packaging and others). It presents the first results of the large-scale IMOCO4.E2 project and shows the pathways for application of the technology on specific machines (so-called pilots). The authors believe the ideas presented could be inspiring also in other domains.
Indeksēts Scopus / WoS
- 2022
BlueTiSCH: A Multi-PHY Simulation of Low-Power 6TiSCH IoT Networks(atveras jaunā logā)
Chloe Bae, Shiwen Yang, Michael Baddeley, Atis Elsts, Israat Haque
IEEE Globecom 2022, IEEE
Kopsavilkums
Low-power wireless IoT networks have traditionally operated over a single physical layer (PHY) – many based on the IEEE 802.15.4 standard. However, recent low-power wireless chipsets offer both the IEEE 802.15.4 and all four PHYs of the Bluetooth 5 (BT 5) standard. This introduces the intriguing possibility that IoT solutions might not necessarily be bound by the limits of a single PHY, and could actively or proactively adapt their PHY depending on RF or networking conditions (e.g., to offer a higher throughput or a longer radio range). Several recent studies have explored such use-cases. However, these studies lack comprehensive evaluation over various metrics (such as reliability, latency, and energy) with regards to scalability and the Radio Frequency (RF) environment. In this work we evaluate the performance of IEEE 802.15.4 and the four BT 5 2.4GHz PHY options for the recently completed IETF 6TiSCH low-power wireless standard. To the best of our knowledge, this is the first work to directly compare these PHYs in identical settings. Specifically, we use a recently released 6TiSCH simulator, TSCH-Sim, to compare these PHY options in networks of up to 250 nodes over different RF environments (home, industrial, and outdoor), and highlight from these results how different PHY options might be better suited to particular application use-cases.
- 2022
The Contiki-NG open source operating system for next generation IoT devices(atveras jaunā logā)
George Oikonomou, Simon Duquennoy, Atis Elsts, Joakim Eriksson, Yasuyuki Tanaka, Nicolas Tsiftes
SoftwareX, Sēj. 18, Elsevier
Kopsavilkums
Contiki-NG (Next Generation) is an open source, cross-platform operating system for severely constrained wireless embedded devices. It focuses on dependable (reliable and secure) low-power communications and standardised protocols, such as 6LoWPAN, IPv6, 6TiSCH, RPL, and CoAP. Its primary aims are to (i) facilitate rapid prototyping and evaluation of Internet of Things research ideas, (ii) reduce time-to-market for Internet of Things applications, and (iii) provide an easy-to-use platform for teaching embedded systems-related courses in higher education. Contiki-NG started as a fork of the Contiki OS and retains many of its original features. In this paper, we discuss the motivation behind the creation of Contiki-NG, present the most recent version (v4.7), and highlight the impact of Contiki-NG through specific examples.
Scopus procentile: 63 · SNIP: 1.458