The University of Warwick

This paper leverages multi-agent reinforcement learning (MARL) to develop an efficient control system for a wind farm comprising a new type of wind turbines with hydrostatic transmission. The primary motivation for hydrostatic wind turbines (HWT) is increased reliability, and reduced manufacturing, operating, and maintaining costs by removing troublesome components and reducing nacelle weight. Nevertheless, the high system complexity of HWT and the wake effect pose significant challenges for the control of HWT-based wind farms. We therefore propose a MARL algorithm named multi-agent policy optimization (MAPO), which allows agents (turbines) to gradually improve their control policies by repeatedly interacting with the environment to learn an optimal operation curve for wind farms. Simulation results based on a wind farm simulator, FAST.Farm, show that MAPO outperforms the greedy policy and a popular learning-based method, multi-agent deep deterministic policy gradient (MADDPG), in terms of power generation.

Training Deep Neural Networks (DNN) concurrently is becoming increasingly important for deep learning practitioners, e.g., hyperparameter optimization (HPO) and neural architecture search (NAS) . The GPU memory capacity is the impediment that prohibits multiple DNNs from being trained on the same GPU due to the large memory usage during training. In this paper, we propose Waterwave a GPU memory flow engine for concurrent deep learning training. Firstly, to address the memory explosion brought by the long time lag between memory allocation and deallocation time, we develop an allocator tailored for multi-streams. By making the allocator aware of the stream information, a prioritized allocation is conducted based on the chunk's synchronization attributes, allowing us to provide useable memory after scheduling rather than waiting it to be really released after GPU computation. Secondly, Waterwave partitions the compute graph to a set of continuous node groups and then performs finer-grained scheduling: NodeGroup pipeline execution , to guarantee a proper memory requests order. Waterwave can accomplish up to 96.8% of the maximum batch size of solo training. Additionally, in scenarios with high memory demand, Waterwave can outperform existing spatial sharing and temporal sharing by up to 12x and 1.49x, respectively.

Predicting the behaviour (i.e., manoeuvre/trajectory) of other road users, including vehicles, is critical for the safe and efficient operation of autonomous vehicles (AVs), a.k.a., automated driving systems (ADSs). Due to the uncertain future behaviour of vehicles, multiple future behaviour modes are often plausible for a vehicle in a given driving scene. Therefore, multimodal prediction can provide richer information than single-mode prediction, enabling AVs to perform a better risk assessment. To this end, we propose a novel multimodal prediction framework that can predict multiple plausible behaviour modes and their likelihoods. The proposed framework includes a bespoke problem formulation for manoeuvre prediction, a novel transformer-based prediction model, and a tailored training method for multimodal manoeuvre and trajectory prediction. The performance of the framework is evaluated using three public highway driving datasets, namely NGSIM, highD, and exiD. The results show that our framework outperforms the state-of-the-art multimodal methods in terms of prediction error and is capable of predicting plausible manoeuvre and trajectory modes.

A web service is a software interface that describes a collection of operations that can be accessed over the Internet using standard protocols. Though web services have significant features, centralized UDDI architecture is one of the most challenging issues which attract researchers for an efficient solution. In this paper, a cooperative distributed UDDI (dUDDI) architecture for P2P service networks has been proposed. dUDDI system decentralizes the traditional UDDI using a collection of minimum traffic components which maintains the service provider discovery start list. Service providers act cooperatively on the service discovery operation by linking to other providers who offer similar services. A comprehensive description of the various elements in the dUDDI architecture and their internal component is presented. We also presented an effective algorithm for service publish and discovery operations using dUDDI architecture. The proposed model improves the efficiency of service resource retrieval and also applies different security measures. The proposed dUDDI model is evaluated with the best-in-class working decentralized UDDI models by considering different conditions like the registry size, QoS factors and discovery of the relevant services based on user request. A testbed has been generated consisting of 1000 web services of various domains and services are manually divided into 21 domains with different QoS requirement combinations. The experimentation results justify that the proposed model outperforms the existing decentralized UDDI models in terms of precision, recall and f-measure factors.

With its growing use in safety/security-critical applications, Deep Learning (DL) has raised increasing concerns regarding its dependability. In particular, DL has a notorious problem of lacking robustness. Input added with adversarial perturbations, i.e. Adversarial Examples (AEs) are easily mis-predicted by the DL model. Despite recent efforts made in detecting AEs via state-of-the-art attack and testing methods, they are normally input distribution agnostic and/or disregard the perceptual quality of adversarial perturbations. Consequently, the detected AEs are irrelevant inputs in the application context or unrealistic that can be easily noticed by humans. This may lead to a limited effect on improving the DL model’s dependability, as the testing budget is likely to be wasted on detecting AEs that are encountered very rarely in its real-life operations. In this paper, we propose a new robustness testing approach for detecting AEs that considers both the feature level distribution and the pixel level distribution, capturing the perceptual quality of adversarial perturbations. The two considerations are encoded by a novel hierarchical mechanism. First, we select test seeds based on the density of feature level distribution and the vulnerability of adversarial robustness. The vulnerability of test seeds are indicated by the auxiliary information, that are highly correlated with local robustness. Given a test seed, we then develop a novel genetic algorithm based local test case generation method, in which two fitness functions work alternatively to control the perceptual quality of detected AEs. Finally, extensive experiments confirm that our holistic approach considering hierarchical distributions is superior to the state-of-the-arts that either disregard any input distribution or only consider a single (non-hierarchical) distribution, in terms of not only detecting imperceptible AEs but also improving the overall robustness of the DL model under testing.

In a large variety of systems (biological, physical, social etc.), synchronization occurs when different oscillating objects tune their rhythm when they interact with each other. The different underlying network defining the connectivity properties among these objects drives the global dynamics in a complex fashion and affects the global degree of synchrony of the system. Here we study the impact of such types of different network architectures, such as Fully-Connected, Random, Regular ring lattice graph, Small-World and Scale-Free in the global dynamical activity of a system of coupled Kuramoto phase oscillators. By fixing the external stimulation parameters, we choose different fractions of nodes from the system first randomly and then informed by their respective strong/weak connectivity properties (centrality, shortest path length and clustering coefficient) and we measure the global degree of synchrony. Our main finding is, that in Scale-Free and Random networks a sophisticated choice of nodes based on graph connectivity properties exhibits a systematic trend in achieving higher degree of synchrony. For the other types of graphs considered, the choice of the stimulated nodes (randomly vs selectively using the aforementioned criteria) seems to not have a noticeable effect.

The capacity retention of commercially-sourced pouch cells with single crystal Al surface-doped Ni-rich cathodes (LiNi0.834Mn0.095Co0.071O2) is examined. The degradation-induced capacity fade becomes more pronounced as the upper-cut-off voltage (UCV) increases from 4.2 V to 4.3 V (vs. graphite) at a fixed cycling temperature (either 25 or 40 °C). However, cycles with 4.3 V UCV (slightly below the oxygen loss onset) show better capacity retention upon increasing the cycling temperature from 25 °C to 40 °C. Namely, after 500 cycles at 4.3 V UCV, cycling temperature at 40 °C retains 85.5% of the initial capacity while cycling at 25 °C shows 75.0% capacity retention. By employing a suite of electrochemical, X-ray spectroscopy and secondary ion mass spectrometry techniques, we attribute the temperature-induced improvement of the capacity retention at high UCV to the combined effects of Al surface-dopants, electrochemically resilient single crystal Ni-rich particles, and thermally-improved Li kinetics translating into better electrochemical performance. If cycling remains below the lattice oxygen loss onset, improved capacity retention in industrial cells should be achieved in single crystal Ni-rich cathodes with the appropriate choice of cycling parameter, particle quality, and particle surface dopants.

This article considers how theories of social cooperation might be helpful in developing policy levers for changing travel behaviours towards environmentally beneficial outcomes, especially in reducing private car use. ‘Theories of cooperation’ can be described as a shift away from a ‘traditional’ economic focus on selfish individuals to one where individuals care what those around them are doing and even sometimes identify with, and think as, groups. We use a simplified ‘game’ to show how game theory offers a very constrained backdrop to thinking about cooperation in a transport setting: it neglects important social factors, both strategic ones and the general social interactions and ease that may be required as a backdrop to cooperation in real life. We then apply this to ‘use cases’ (lift sharing, on-site travel planning, safe cycle storage and peer-to-peer information sharing) that bridge the gap between the abstractions of theories of cooperation, on the one hand, and the practicalities of policymaking and lived reality, on the other. In doing this, we show how cooperation in travel behaviour can develop in two different ways: as emergent social phenomena (for example, the informal-economy approach to car or bicycle repair) and purposeful policy initiatives (for example, rail-fare discounts for two people travelling together, such as the UK’s ‘two together’ railcard). Somewhat reductively, these could be described as ‘bottom-up’ and ‘top-down’ elements within behaviour-change processes. The article shows that: (1) cooperation exists ‘naturally’ in the ‘travel-behaviour policy space’; (2) there is a wealth of opportunities for policy to help make cooperation happen more and/or work better; and (3) this includes opportunities to create the conditions required for cooperation to exist and flourish.

We present a multiscale method for simulating non-equilibrium lubrication flows. The effect of low pressure or tiny lubricating geometries that gives rise to rarefied gas effects means that standard Navier–Stokes solutions are invalid, while the large lateral size of the systems that need to be investigated is computationally prohibitive for Boltzmann solutions, such as the direct simulation Monte Carlo method (DSMC). The multiscale method we propose is applicable to time-varying, low-speed, rarefied gas flows in quasi-3D geometries that are now becoming important in various applications, such as next-generation microprocessor chip manufacturing, aerospace, sealing technologies and MEMS devices. Our multiscale simulation method provides accurate solutions, with errors of less than 1% compared to the DSMC benchmark results when all modeling conditions are met. It also shows computational gains over DSMC that increase when the lateral size of the systems increases, reaching 2–3 orders of magnitude even for relatively small systems, making it an effective tool for simulation-based design.

In this essay, we present Michael Polanyi’s theory of knowledge and outline its implications for theory development in organizational research. While Polanyi is best known in the field for his concept of tacit knowledge, we discuss here several other cognate concepts Polanyi has also introduced, notably: conviviality, indwelling, tradition and lore, coherence, and post-critical reason, and show how they help us better understand organizational theorizing. Specifically, we argue the following. First, when engaged in theory creation, organizational scholars integrate largely unspecifiable particulars, in search of deepening coherence, by dwelling in a fiduciary framework of previous theory, others’ narrativized experiences, and their own personal experiences. Secondly, driven by intellectual passions and commitments, organizational scholars bring about conceptual novelty by seeking to redirect intellectual attention to hitherto tacitly accepted subsidiary particulars, which they seek to re-integrate in novel ways. And thirdly, since all knowledge, no matter how abstract, necessarily involves skilful action, organizational scholars dwell in scientific practice and, therefore, interiorize – that is, they become subsidiarily aware of - the practice’s collective purpose, which they freely and responsibly enact through the exercise of public liberty.

Accurate inference of who infected whom in an infectious disease outbreak is critical for the delivery of effective infection prevention and control. The increased resolution of pathogen whole-genome sequencing has significantly improved our ability to infer transmission events. Despite this, transmission inference often remains limited by the lack of genomic variation between the source case and infected contacts. Although within-host genetic diversity is common among a wide variety of pathogens, conventional whole-genome sequencing phylogenetic approaches exclusively use consensus sequences, which consider only the most prevalent nucleotide at each position and therefore fail to capture low frequency variation within samples. We hypothesized that including within-sample variation in a phylogenetic model would help to identify who infected whom in instances in which this was previously impossible. Using whole-genome sequences from SARS-CoV-2 multi-institutional outbreaks as an example, we show how within-sample diversity is partially maintained among repeated serial samples from the same host, it can transmitted between those cases with known epidemiological links, and how this improves phylogenetic inference and our understanding of who infected whom. Our technique is applicable to other infectious diseases and has immediate clinical utility in infection prevention and control.

Gemmatimonadota is a diverse bacterial phylum commonly found in environments such as soils, rhizospheres, fresh waters, and sediments. So far, the phylum contains just six cultured species (five of them sequenced), which limits our understanding of their diversity and metabolism. Therefore, we analyzed over 400 metagenome-assembled genomes (MAGs) and 5 culture-derived genomes representing Gemmatimonadota from various aquatic environments, hydrothermal vents, sediments, soils, and host-associated (with marine sponges and coral) species. The principal coordinate analysis based on the presence/absence of genes in Gemmatimonadota genomes and phylogenomic analysis documented that marine and host-associated Gemmatimonadota were the most distant from freshwater and wastewater species. A smaller genome size and coding sequences (CDS) number reduction were observed in marine MAGs, pointing to an oligotrophic environmental adaptation. Several metabolic pathways are restricted to specific environments. For example, genes for anoxygenic phototrophy were found only in freshwater, wastewater, and soda lake sediment genomes. There were several genomes from soda lake sediments and wastewater containing type IC/ID ribulose-1,5-bisphosphate carboxylase/oxygenase (RuBisCO). Various genomes from wastewater harbored bacterial type II RuBisCO, whereas RuBisCO-like protein was found in genomes from fresh waters, soil, host-associated, and marine sediments. Gemmatimonadota does not contain nitrogen fixation genes; however, the nosZ gene, involved in the reduction of N 2 O, was present in genomes from most environments, missing only in marine water and host-associated Gemmatimonadota. The presented data suggest that Gemmatimonadota evolved as an organotrophic species relying on aerobic respiration and then remodeled its genome inventory when adapting to particular environments. IMPORTANCE Gemmatimonadota is a rarely studied bacterial phylum consisting of a handful of cultured species. Recent culture-independent studies documented that these organisms are distributed in many environments, including soil, marine, fresh, and waste waters. However, due to the lack of cultured species, information about their metabolic potential and environmental role is scarce. Therefore, we collected Gemmatimonadota metagenome-assembled genomes (MAGs) from different habitats and performed a systematic analysis of their genomic characteristics and metabolic potential. Our results show how Gemmatimonadota have adapted their genomes to different environments.

There is a need for all industries, including healthcare, to reduce their greenhouse gas emissions. In anaesthetic practice, this not only requires a reduction in resource use and waste, but also a shift away from inhaled anaesthetic gases and towards alternatives with a lower carbon footprint. As inhalational anaesthesia produces greenhouse gas emissions at the point of use, achieving sustainable anaesthetic practice involves individual practitioner behaviour change. However, changing the practice of healthcare professionals raises potential ethical issues. The purpose of this paper is twofold. First, we discuss what moral duties anaesthetic practitioners have when it comes to practices that impact the environment. We argue that behaviour change among practitioners to align with certain moral responsibilities must be supplemented with an account of institutional duties to support this. In other words, we argue that institutions and those in power have second-order responsibilities to ensure that practitioners can fulfil their first-order responsibilities to practice more sustainably. The second goal of the paper is to consider not just the nature of second-order responsibilities but the content. We assess four different ways that second-order responsibilities might be fulfilled within healthcare systems: removing certain anaesthetic agents, seeking consensus, education and methods from behavioural economics. We argue that, while each of these are a necessary part of the picture, some interventions like nudges have considerable advantages.

The Soviet Union was one of the most secretive states that ever existed. Defended by a complex apparatus of rules and checks administered by the secret police, the Soviet state had seemingly unprecedented capabilities based on its near monopoly of productive capital, monolithic authority, and secretive decision making. But behind the scenes, Soviet secrecy was double-edged: it raised transaction costs, incentivized indecision, compromised the effectiveness of government officials, eroded citizens' trust in institutions and in each other, and led to a secretive society and an uninformed elite. The result is what this book calls the secrecy/capacity tradeoff: a bargain in which the Soviet state accepted the reduction of state capacity as the cost of ensuring its own survival. This book is the first comprehensive, analytical, multi-faceted history of Soviet secrecy in the English language. Harrison combines quantitative and qualitative evidence to evaluate the impact of secrecy on Soviet state capacity from the 1917 Bolshevik Revolution to the collapse of the Soviet Union in 1991. Based on multiple years of research in once-secret Soviet-era archives, this book addresses two gaps in history and social science: one the core role of secrecy in building and stabilizing the communist states of the twentieth century; the other the corrosive effects of secrecy on the capabilities of authoritarian states.

Combining the search and pursuit aspects of predator–prey interactions into a single game, where the payoff to the Searcher (predator) is the probability of finding and capturing the Hider (prey) within a fixed number of searches was proposed by Gal and Casas (J. R. Soc. Interface 11, 20140062 (doi:10.1098/rsif.2014.0062)). Subsequent models allowed the predator to continue its search (in another ‘round’) if the prey was found but escaped the chase. However, it is unrealistic to allow this pattern of prey relocation to go on forever, so here we introduce a limit of the total number of searches, in all ‘rounds’, that the predator can carry out. We show how habitat structural complexity affects the mean time until capture: the quality of the location with the lowest capture probability matters more than the number of hiding locations. Moreover, we observed that the parameter space defined by the capture probabilities in each location and the budget of the predator can be divided into distinct domains, defining whether the prey ought to play with pure or mixed hiding strategies.