Search

Influenced by the advances in data and computing, the scientific practice increasingly involves machine learning and artificial intelligence driven methods which requires specialized capabilities at the system-, science- and service-level in addition to the conventional large-capacity supercomputing approaches. The latest distributed architectures built around the composability of data-centric applications led to the emergence of a new ecosystem for container coordination and integration. However, there is still a divide between the application development pipelines of existing supercomputing environments, and these new dynamic environments that disaggregate fluid resource pools through accessible, portable and re-programmable interfaces. New approaches for dynamic composability of heterogeneous systems are needed to further advance the data-driven scientific practice for the purpose of more efficient computing and usable tools for specific scientific domains. In this paper, we present a novel approach for using composable systems in the intersection between scientific computing, artificial intelligence (AI), and remote sensing domain. We describe the architecture of a first working example of a composable infrastructure that federates Expanse, an NSF-funded supercomputer, with Nautilus, a Kubernetes-based GPU geo-distributed cluster. We also summarize a case study in wildfire modeling, that demonstrates the application of this new infrastructure in scientific workflows: a composed system that bridges the insights from edge sensing, AI and computing capabilities with a physics-driven simulation., Comment: 18th IEEE International Conference on eScience (2022)

We investigate the coherence and non-Markovianity of a quantum tunneling system whose barrier is fluctuated by a telegraph noise, and its energy gap is modulated by Gaussian noise. With the help of averaging method, the system dynamics are analytically derived, and the analytical expression for coherence measure and non-Markovianity for the very limited parameter regimes for both initially coherent and non-coherent states are obtained. We observe non-Markovian dynamics in a situation where the Kubo number is high. It is also found that there is no strong relation between the coherence of the system and non-Markovianity dynamics except in a region in which these two tend to change their behavior at the intermediate noise color for two initial states., Comment: 15 pages, 3 figures

5G networks provide higher flexibility and improved performance compared to previous cellular technologies. This has raised expectations on the possibility to support advanced V2X services using the cellular network via Vehicle-to-Network (V2N) and V2N2V connections. Replacing direct V2V connections by V2N/V2N2V communications to support critical V2X services requires low V2N/V2N2V latencies. It is then necessary to quantify the latency values that V2N and V2N2V connections can achieve over 5G end-to-end (E2E) connections, but these depend on the particular 5G network deployments and configurations and to date the number of related studies is limited. This paper progresses the state-of-the-art by introducing a novel E2E latency model to quantify the latency of 5G V2N/V2N2V communications. The model includes the latency introduced at the radio, transport, core, Internet, peering points and application server (AS) for single MNO and multi-MNO scenarios. This paper estimates the E2E latency for a large variety of possible 5G network deployments that are being discussed or envisioned to support V2X services. This includes the possibility to deploy the V2X AS from the edge of the network to the cloud. The model is utilized to analyze the impact of different 5G network deployments and configurations on the E2E latency. The analysis helps identify which 5G network deployments and configurations are more suitable to meet V2X latency requirements. The conducted analysis highlights the challenge for centralized network deployments that locate the V2X AS at the cloud to meet the latency requirements of advanced V2X services. Locating the V2X AS closer to the cell edge reduces the latency. However, it requires a higher number of ASs and also a careful dimensioning of the network and its configuration to ensure sufficient network and AS resources are dedicated to serve the V2X traffic., Comment: 15 pages, 8 figures, 15 tables

5G has been designed to support applications such as connected and automated driving. To this aim, 5G includes a highly flexible New Radio (NR) interface that can be configured to utilize different subcarrier spacings (SCS), slot durations, scheduling, and retransmissions mechanisms. This flexibility can be exploited to support advanced V2X services with strict latency and reliability requirements using V2N2V (Vehicle-to-Network-to-Vehicles) communications instead of direct or sidelink V2V (Vehicle-to-Vehicle). To analyze this possibility, this paper presents a novel analytical model that estimates the latency of 5G at the radio network level. The model accounts for the use of different numerologies (SCS, slot durations and Cyclic Prefixes), modulation and coding schemes, full-slots or mini-slots, semi-static and dynamic scheduling, different retransmission mechanisms, and broadcast/multicast or unicast transmissions. The model has been used to first analyze the impact of different 5G NR radio configurations on the latency. We then identify which radio configurations and scenarios can 5G NR satisfy the latency and reliability requirements of V2X services using V2N2V communications. This paper considers cooperative lane changes as a case study. The results show that 5G can support advanced V2X services at the radio network level using V2N2V communications under certain conditions that depend on the radio configuration, bandwidth, service requirements and cell traffic load., Comment: 14 pages, 15 figures, 2 tables

The solution of the shortest path problem on a surface is not only a theoretical problem to be solved in the field of mathematics, but also problems that need to be solved in very different fields such as medicine, defense and construction technologies. When it comes to the land specific, solution algorithms for these problems are also of great importance in terms of determination of the shortest path in an open area where the road will pass in the field of civil engineering, or route determination of manned or unmanned vehicles for various logistic needs, especially in raw terrains. In addition, path finding problems in the raw terrains are also important for manned and unmanned ground vehicles (UGV) used in the defense industry. Within the scope of this study, a method that can be used for instant route determinations within sight range or for route determinations covering wider areas is proposed. Although the examples presented within the scope of the study are land-based, the method can be applied to almost all problem types of similar nature. The approach used in the study can be briefly described as the mechanical analysis of a surface transformed into a structural load bearing system based on mechanical analogies. In this approach, the determination of the shortest path connecting two points can be realized by following the stress-strain values that will occur by moving the points away from each other or by following a linear line that will be formed between two points during the mechanical analysis. If the proposed approach is to be carried out with multiple rigid body dynamics approaches instead of flexible bodies mechanics, it can be carried out easily and very quickly by determining the shortest path between two points or by tracking the forces. However, the proposed approach in this study is presented by simulating examples of flexible bodies using FEM., Comment: 27 pages, 9 figures

The size and frequency of wildland fires in the western United States have dramatically increased in recent years. On high-fire-risk days, a small fire ignition can rapidly grow and become out of control. Early detection of fire ignitions from initial smoke can assist the response to such fires before they become difficult to manage. Past deep learning approaches for wildfire smoke detection have suffered from small or unreliable datasets that make it difficult to extrapolate performance to real-world scenarios. In this work, we present the Fire Ignition Library (FIgLib), a publicly available dataset of nearly 25,000 labeled wildfire smoke images as seen from fixed-view cameras deployed in Southern California. We also introduce SmokeyNet, a novel deep learning architecture using spatiotemporal information from camera imagery for real-time wildfire smoke detection. When trained on the FIgLib dataset, SmokeyNet outperforms comparable baselines and rivals human performance. We hope that the availability of the FIgLib dataset and the SmokeyNet architecture will inspire further research into deep learning methods for wildfire smoke detection, leading to automated notification systems that reduce the time to wildfire response.