Non-embedded Systems Research Articles

Design and Evaluation of CPU-, GPU-, and FPGA-Based Deployment of a CNN for Motor Imagery Classification in Brain-Computer Interfaces

Brain–computer interfaces (BCIs) have gained popularity in recent years. Among noninvasive BCIs, EEG-based systems stand out as the primary approach, utilizing the motor imagery (MI) paradigm to discern movement intentions. Initially, BCIs were predominantly focused on nonembedded systems. However, there is now a growing momentum towards shifting computation to the edge, offering advantages such as enhanced privacy, reduced transmission bandwidth, and real-time responsiveness. Despite this trend, achieving the desired target remains a work in progress. To illustrate the feasibility of this shift and quantify the potential benefits, this paper presents a comparison of deploying a CNN for MI classification across different computing platforms, namely, CPU-, embedded GPU-, and FPGA-based. For our case study, we utilized data from 29 participants included in a dataset acquired using an EEG cap for training the models. The FPGA solution emerged as the most efficient in terms of the power consumption–inference time product. Specifically, it delivers an impressive reduction of up to 89% in power consumption compared to the CPU and 71% compared to the GPU and up to a 98% reduction in memory footprint for model inference, albeit at the cost of a 39% increase in inference time compared to the GPU. Both the embedded GPU and FPGA outperform the CPU in terms of inference time.

Compact groups from semi-analytical models of galaxy formation – V. Their assembly channels as a function of the environment

ABSTRACT We delved into the assembly pathways and environments of compact groups (CGs) of galaxies using mock catalogues generated from semi-analytical models (SAMs) on the Millennium simulation. We investigate the ability of SAMs to replicate the observed CG environments and whether CGs with different assembly histories tend to inhabit specific cosmic environments. We also analyse whether the environment or the assembly history is more important in tailoring CG properties. We find that about half of the CGs in SAMs are non-embedded systems, 40 per cent are inhabiting loose groups or nodes of filaments, while the rest distribute evenly in filaments and voids, in agreement with observations. We observe that early-assembled CGs preferentially inhabit large galaxy systems ($\sim 60~{{\ \rm per\ cent}}$), while around 30 per cent remain non-embedded. Conversely, lately formed CGs exhibit the opposite trend. We also obtain that lately formed CGs have lower velocity dispersions and larger crossing times than early-formed CGs, but mainly because they are preferentially non-embedded. Those lately formed CGs that inhabit large systems do not show the same features. Therefore, the environment plays a strong role in these properties for lately formed CGs. Early-formed CGs are more evolved, displaying larger velocity dispersions, shorter crossing times, and more dominant first-ranked galaxies, regardless of the environment. Finally, the difference in brightness between the two brightest members of CGs is dependent only on the assembly history and not on the environment. CGs residing in diverse environments have undergone varied assembly processes, making them suitable for studying their evolution and the interplay of nature and nurture on their traits.

A Minimally Invasive Low-Power Platform for Real-Time Brain Computer Interaction Based on Canonical Correlation Analysis

A growing trend in human–computer interaction is to integrate computational capabilities into wearable devices, to enable sophisticated and natural interaction modalities. Acting directly by decoding neural activity is a very natural way of interaction and one of the fundamental paradigms of brain computer interfaces (BCIs) as well. In this paper, we present a wearable Internet of Things node designed for BCI spelling. The system is based on visual evoked potentials detection and runs the canonical correlation analysis on a low power microcontroller. Neural data is acquired by an array of electroencephalography active dry electrodes, suitable for a minimally intrusive interface. To evaluate our solution, we optimized the system on eight subjects and tested it on five different subjects for four and eight stimuli, reaching a peak transfer rate of 1.57 b/s, comparable with those achieved by state-of-the-art nonembedded systems. The power consumption of the device is less than 30 mW, resulting in 122 h of operation with a standard 1000-mAh battery.

Options for non-embedded optical systems of rotation angle sensors based on the BR-180º prism and photoelectric autocollimator

Options for non-embedded optical systems of rotation angle sensors based on the BR-180º prism and photoelectric autocollimator

Where are compact groups in the local Universe?

The purpose of this work is to perform a statistical analysis of the location of compact groups in the Universe from observational and semi-analytical points of view. We used the velocity-filtered compact group sample extracted from the Two Micron All Sky Survey for our analysis. We also used a new sample of galaxy groups identified in the 2M++ galaxy redshift catalogue as tracers of the large-scale structure. We defined a procedure to search in redshift space for compact groups that can be considered embedded in other overdense systems and applied this criterion to several possible combinations of different compact and galaxy group subsamples. We also performed similar analyses for simulated compact and galaxy groups identified in a 2M++ mock galaxy catalogue constructed from the Millennium Run Simulation I plus a semi-analytical model of galaxy formation. We observed that only $\sim27\%$ of the compact groups can be considered to be embedded in larger overdense systems, that is, most of the compact groups are more likely to be isolated systems. The embedded compact groups show statistically smaller sizes and brighter surface brightnesses than non-embedded systems. No evidence was found that embedded compact groups are more likely to inhabit galaxy groups with a given virial mass or with a particular dynamical state. We found very similar results when the analysis was performed using mock compact and galaxy groups. Based on the semi-analytical studies, we predict that $70\%$ of the embedded compact groups probably are 3D physically dense systems. Finally, real space information allowed us to reveal the bimodal behaviour of the distribution of 3D minimum distances between compact and galaxy groups. The location of compact groups should be carefully taken into account when comparing properties of galaxies in environments that are a priori different.

Comparing development approaches and reuse strategies: An empirical evaluation of developer views from the aerospace industry

ContextThere is a debate in the aerospace industry whether lessons from reuse successes and failures in nonembedded software can be applied to embedded software. Reuse supposedly reduces development time and errors. The aerospace industry was an early advocate of reuse, but in Aerospace, not all reuse experiences have been as successful as expected. Some major projects experienced large overruns in time, budget, as well as inferior performance, at least in part, due to the gap between reuse expectations and reuse outcomes. This seemed to be especially the case for embedded systems. ObjectiveOur goal is to discover software reuse practices in the aerospace industry. In particular, we wish to learn whether practitioners who develop embedded systems use the same development approaches and artifacts as software practitioners who develop nonembedded systems. We wish to learn whether reuse influences selection of development approaches and artifacts and whether outcomes are impacted. MethodWe developed a survey given to software practitioners in a major Aerospace Corporation developing either embedded or nonembedded systems. The survey probed to identify development methods used, artifacts reused and outcomes resulting from the reuse. We used qualitative and quantitative methods such as descriptive statistics, MANOVA, Principle Component Analysis and an analysis of freeform comments to compare reuse practices between embedded systems and nonembedded systems development. ResultsWe found that embedded systems were more likely to include component based development, product line development and model based development in their development approach, whereas nonembedded systems were more likely to include Ad Hoc and COTS/GOTS in their development approach. Embedded systems developers tended to reuse more and different reuse artifacts. ConclusionWe found that, while outcomes were nearly identical, the development approaches and artifacts used did, in fact, differ. In particular, the tight coupling between code and the platform in embedded systems often dictated the development approach and reuse artifacts and identified some of the reasons.

The extreme vulnerability of interdependent spatially embedded networks

Recent studies show that in interdependent networks a very small failure in one network may lead to catastrophic consequences. Above a critical fraction of interdependent nodes, even a single node failure can invoke cascading failures that may abruptly fragment the system, whereas below this critical dependency a failure of a few nodes leads only to a small amount of damage to the system. So far, research has focused on interdependent random networks without space limitations. However, many real systems, such as power grids and the Internet, are not random but are spatially embedded. Here we analytically and numerically study the stability of interdependent spatially embedded networks modelled as lattice networks. Surprisingly, we find that in lattice systems, in contrast to non-embedded systems, there is no critical dependency and any small fraction of interdependent nodes leads to an abrupt collapse. We show that this extreme vulnerability of very weakly coupled lattices is a consequence of the critical exponent describing the percolation transition of a single lattice. Networks of networks are vulnerable: a failure in one sub-network can bring the rest crashing down. Previous simulations have suggested that randomly positioned networks might offer some limited robustness under certain circumstances. Analysis now shows, however, that real-world interdependent networks, where nodes are positioned according to geographical constraints, might not be so resilient.

Intrusion detection for resource-constrained embedded control systems in the power grid

The power grid depends on embedded control systems or SCADA systems to function properly. Securing these systems presents unique challenges—in addition to the resource restrictions inherent to embedded devices, SCADA systems must accommodate strict timing requirements that are non-negotiable, and their massive scale greatly amplifies costs such as power consumption. Together, these constraints make the conventional approach to host intrusion detection–using a hypervisor to create a safe environment from which a monitoring entity can operate–too costly or impractical for embedded control systems in the critical infrastructure.This paper discusses the design and implementation of Autoscopy, an experimental host-based intrusion detection mechanism that operates from within the kernel and leverages its built-in tracing framework to identify control-flow anomalies, which are most often caused by rootkits that hijack kernel hooks. The paper presents the concepts underlying the original Autoscopy prototype, highlights some of the issues that arose from it, and introduces the new system, dubbed Autoscopy Jr., which addresses the issues. Tests on non-embedded systems demonstrated that the monitoring scope could be managed to limit Autoscopy Jr.’s performance impact on its host to under 5%. The paper also describes the use of an optimized probe framework to reduce overhead and the test results obtained for a hardened kernel. The results demonstrate that Autoscopy Jr.’s design and effectiveness render it uniquely suited to intrusion detection for SCADA systems.

Architecture‐based semantic evolution of embedded remotely controlled systems

AbstractEvolution of a software system is a natural process. In most systems, evolution takes place during the maintenance phase of their life cycles. Those systems that have reached their limit in evolution have usually reached their end of useful life and may have to be replaced. However, there are systems in which evolution occurs during the operational phase of their life cycles. Such systems are designed to evolve while in use or, in other words, be adaptable. Semantically adaptable systems are of particular interest to industry as such systems often times adapt themselves to environment change with little or no intervention from their developing or maintaining organization. Since embedded systems usually have a restricted hardware configuration, it is difficult to apply the techniques developed for non‐embedded systems directly to embedded systems. This paper focuses on evolution through adaptation and develops the concepts and techniques for semantic evolution in embedded systems. As the first step in the development of a software solution, architectures of software systems themselves have to be made semantically evolvable. In this paper we explore various architectural alternatives for the semantic evolution of embedded systems—these architectures are based on four different techniques that we have identified for semantic evolution in embedded systems. The development of these architectures follows the systematic process provided by the non‐functional requirement (NFR) framework, which also permits the architectures to be rated in terms of their evolvability. As the field of embedded systems is vast, this paper concentrates on those embedded systems that can be remotely controlled. In this application domain the embedded system is connected to an external controller by a communication link such as ethernet, serial, radio frequency, etc., and receives commands from and sends responses to the external controller via the communication link. The architectures developed in this paper have been partly validated by applying them in a real embedded system—a test instrument used for testing cell phones. These architectures and techniques for semantic evolution in this application domain give a glimpse of what can be done in achieving semantic evolution in software‐implemented systems. Copyright © 2003 John Wiley & Sons, Ltd.

Multistream transmission for hybrid IBOC-AM with embedded/multidescriptive audio coding

Hybrid in band on channel (IBOC) digital audio broadcasting simultaneously with analog amplitude modulation (AM) has been proposed as a solution to DAB in the AM band. Since the AM band is crowded and the available bandwidth per program is limited, adding digital transmission presents a challenge. To achieve FM-like audio quality, an audio coder rate of 32-64 kb/sec may be required. One currently proposed hybrid IBOC-AM system is 30 kHz wide. Severe second adjacent interference may occur in certain areas, leading to a possible 40% loss of the effective audio bit rate. To cope with such harsh transmission conditions, we present a solution based on embedded/multidescriptive audio coding with matched multistream transmission in separate frequency bands. With the loss of one frequency band, the embedded system blends to a lower audio coder rate with a much better quality than analog AM. The nonembedded system, without multistream transmission, fails catastrophically when a little more than one sideband is severely interfered with, causing a severe discontinuity in quality while blending directly to analog AM. Some detailed robust embedded systems are outlined. We also show how multistream transmission schemes can be used with nonembedded audio coders. Both daytime and nighttime scenarios are included. This paper contains a catalog of possible systems for different audio quality levels and interference scenarios, including systems with 20 kHz bandwidth rather than 30 kHz.