1,721,104 research outputs found
A Peak-Shaving-Oriented Incentive Mechanism for Smart Grids
No full textProsumers play a crucial role in smart grids, especially within local energy communities (LECs), since they can both consume and produce energy. When peer-to-peer (P2P) energy trading is available, prosumers can exchange their produced energy with each other: if done properly, this may lead to better energy self-consumption throughout the grid, resulting in reduced transmission losses, lower energy costs, and decreased wear and tear to the grid. Previous work on this topic led to a mechanism capable of obtaining several such goals, like preventing intentional energy production curtailment, disincentivizing simultaneous energy consumption that may lead to congestions, encouraging users to consume their own produced energy as much as possible, and ensuring that even if users initially create schedules with a selfish approach, they will ultimately converge upon a configuration that garners mutual agreement. However, this mechanism has not yet been analyzed from the perspective of peak shaving. Therefore, this paper aims to cover this shortcoming. Our objective in this work is to create a new mechanism that, under certain conditions, guarantees the achievement of optimal peak shaving. We will use it as a baseline to compare the existing mechanisms and understand under which conditions it leads to peak shaving. We performed simulations on a dataset from a grid in Cardiff, UK, and the results show that the existing mechanisms achieve optimal peak shaving both if the users act selfishly, and if they are allowed to form coalitions among themselves
Survey on Videos Data Augmentation for Deep Learning Models
No full textIn most Computer Vision applications, Deep Learning models achieve state-of-the-art performances. One drawback of Deep Learning is the large amount of data needed to train the models. Unfortunately, in many applications, data are difficult or expensive to collect. Data augmentation can alleviate the problem, generating new data from a smaller initial dataset. Geometric and color space image augmentation methods can increase accuracy of Deep Learning models but are often not enough. More advanced solutions are Domain Randomization methods or the use of simulation to artificially generate the missing data. Data augmentation algorithms are usually specifically designed for single images. Most recently, Deep Learning models have been applied to the analysis of video sequences. The aim of this paper is to perform an exhaustive study of the novel techniques of video data augmentation for Deep Learning models and to point out the future directions of the research on this topic
Towards Seamless Human-Robot Dialogue through a Robot Action Ontology
No full textThis research paper introduces a novel methodology enabling the Zora humanoid robot to effectively engage in dynamic interactions by responding to user queries and complementing its responses with appropriate gestures. Notably, these inquiries may extend beyond mere questions to encompass action commands articulated by the user, which the robot proficiently recognizes and executes. The integration of a Large Language Model enhances the system's capabilities, particularly in the domain of questionanswering. To bolster the recognition and execution of action commands, we have employed a robot action ontology established in previous research endeavors. This ontology defines relevant classes and individuals, forming the basis for a nuanced understanding of user-inputted action commands. Further refinement involves the generation of succinct three-word strings for each action, ensuring semantic alignment with the user's verbal instructions. Importantly, our system operates in two distinctive modes: STATELESS and STATEFUL. In STATEFUL mode, the robot possesses awareness of its present posture, allowing it to execute action commands only when they align with its current state. This adaptive feature enhances the overall effectiveness of the system, catering to the dynamic nature of human-robot interactions and promoting a seamless and contextually aware dialogue between the NAO humanoid robot and its users
A cooperative game-theory approach for incentive systems in local energy communities
No full textProsumers have a central role in the context of smart grids, and in particular within local energy communities (LECs), as they are capable of being both energy producers and consumers. In a scenario where peer-to-peer (P2P) energy trading is allowed, prosumers can exchange the energy they produce with other prosumers: the primary outcome of this is the improvement of energy self-consumption across the grid, which leads to decreased transmission losses, as well as lower energy costs and diminished long-term damage to the grid itself. Previous work proposed a mechanism to achieve multiple objectives for a cooperative game theory perspective for small coalitions, but its behavior for coalitions of arbitrary size remains unexplored, and it does not consider the objective of peak shaving. This paper aims to (i) design an algorithm for calculating schedules for coalitions of arbitrary size, (ii) analyze the behavior of this mechanism for large coalitions, (iii) create a new incentive mechanism by proposing new selling functions that ensure that the resulting mechanism would optimize for the objective of peak shaving when all the prosumers work together in one large coalition, and (iv) demonstrate the performance of the existing mechanism in terms of peak shaving, by comparing against the mechanism specifically optimized for this objective. Simulations conducted on data from a grid in Cardiff, UK, reveal that the existing mechanism works particularly well for the non-cooperative game, achieving results for cost reduction and self-consumption almost identical to the cooperative game, no matter the size of the coalitions. More precisely, although all mechanisms achieve optimal peak shaving for the grand coalition, the existing mechanism achieves this objective even within the framework of the selfish game, resulting in a reduction of the peak by approximately 29% compared to alternative methods. Furthermore, the mechanism is proven to optimally achieve peak shaving in both cooperative and non-cooperative cases
Deep learning and time series-To-image encoding for financial forecasting
Full text linkIn the last decade, market financial forecasting has attracted high interests amongst the researchers in pattern recognition. Usually, the data used for analysing the market, and then gamble on its future trend, are provided as time series; this aspect, along with the high fluctuation of this kind of data, cuts out the use of very efficient classification tools, very popular in the state of the art, like the well known convolutional neural networks ( CNNs ) models such as Inception, ResNet, AlexNet, and so on. This forces the researchers to train new tools from scratch. Such operations could be very time consuming. This paper exploits an ensemble of CNNs, trained over Gramian angular fields ( GAF ) images, generated from time series related to the Standard - Poor's 500 index future; the aim is the prediction of the future trend of the U.S. market. A multi-resolution imaging approach is used to feed each CNN, enabling the analysis of different time intervals for a single observation. A simple trading system based on the ensemble forecaster is used to evaluate the quality of the proposed approach. Our method outperforms the buy-And-hold ( B - H ) strategy in a time frame where the latter provides excellent returns. Both quantitative and qualitative results are provided
Improving digital twin experience through big data, IoT and social analysis: An architecture and a case study
No full textIndustries such as construction and business companies are becoming increasingly digitized. The amount of data to be monitored and processed has increased significantly since the advent of the Internet of Things and the massive use of sensors. In addition to the data from these sensors, large amounts of data that require specific handling and processing are received. Much of this data is eventually represented in digital twins as a monitoring or decision-support tool. In this paper, we present an architecture to improve digital twin-based experiences that need to represent information from multiple sources. This architecture is demonstrated using the specific use case of a digital twin for an office of an Italian company. The implementation leverages the Matterport 3D media platform and integrates different technologies and sensors. An evaluation of the solution has also been carried out. The results show high user acceptance and the opening of multiple possibilities to enrich the virtual model with further data from different sources
An abstraction layer exploiting voice assistant technologies for effective human—robot interaction
No full textA lot of people have neuromuscular problems that affect their lives leading them to lose an important degree of autonomy in their daily activities. When their disabilities do not involve speech disorders, robotic wheelchairs with voice assistant technologies may provide appropriate human–robot interaction for them. Given the wide improvement and diffusion of Google Assistant, Apple’s Siri, Microsoft’s Cortana, Amazon’s Alexa, etc., such voice assistant technologies can be fully integrated and exploited in robotic wheelchairs to improve the quality of life of affected people. As such, in this paper, we propose an abstraction layer capable of providing appropriate human– robot interaction. It allows use of voice assistant tools that may trigger different kinds of applications for the interaction between the robot and the user. Furthermore, we propose a use case as a possible instance of the considered abstraction layer. Within the use case, we chose existing tools for each component of the proposed abstraction layer. For example, Google Assistant was employed as a voice assistant tool; its functions and APIs were leveraged for some of the applications we deployed. On top of the use case thus defined, we created several applications that we detail and discuss. The benefit of the resulting Human–Computer Interaction is therefore two-fold: on the one hand, the user may interact with any of the developed applications; on the other hand, the user can also rely on voice assistant tools to receive answers in the open domain when the statement of the user does not enable any of the applications of the robot. An evaluation of the presented instance was carried out using the Software Architecture Analysis Method, whereas the user experience was evaluated through ad-hoc questionnaires. Our proposed abstraction layer is general and can be instantiated on any robotic platform including robotic wheelchairs
Leveraging Return Prediction Approaches for Improved Value-at-Risk Estimation
Full text linkValue at risk is a statistic used to anticipate the largest possible losses over a specific time frame and within some level of confidence, usually 95% or 99%. For risk management and regulators, it offers a solution for trustworthy quantitative risk management tools. VaR has become the most widely used and accepted indicator of downside risk. Today, commercial banks and financial institutions utilize it as a tool to estimate the size and probability of upcoming losses in portfolios and, as a result, to estimate and manage the degree of risk exposure. The goal is to obtain the average number of VaR “failures” or “breaches” (losses that are more than the VaR) as near to the target rate as possible. It is also desired that the losses be evenly distributed as possible. VaR can be modeled in a variety of ways. The simplest method is to estimate volatility based on prior returns according to the assumption that volatility is constant. Otherwise, the volatility process can be modeled using the GARCH model. Machine learning techniques have been used in recent years to carry out stock market forecasts based on historical time series. A machine learning system is often trained on an in-sample dataset, where it can adjust and improve specific hyperparameters in accordance with the underlying metric. The trained model is tested on an out-of-sample dataset. We compared the baselines for the VaR estimation of a day (d) according to different metrics (i) to their respective variants that included stock return forecast information of d and stock return data of the days before d and (ii) to a GARCH model that included return prediction information of d and stock return data of the days before d. Various strategies such as ARIMA and a proposed ensemble of regressors have been employed to predict stock returns. We observed that the versions of the univariate techniques and GARCH integrated with return predictions outperformed the baselines in four different marketplaces
Semantic role labeling for knowledge graph extraction from text
Full text linkThis paper introduces TakeFive, a new semantic role labeling method that transforms a text into a frame-oriented knowledge graph. It performs dependency parsing, identifies the words that evoke lexical frames, locates the roles and fillers for each frame, runs coercion techniques, and formalizes the results as a knowledge graph. This formal representation complies with the frame semantics used in Framester, a factual-linguistic linked data resource. We tested our method on the WSJ section of the Peen Treebank annotated with VerbNet and PropBank labels and on the Brown corpus. The evaluation has been performed according to the CoNLL Shared Task on Joint Parsing of Syntactic and Semantic Dependencies. The obtained precision, recall, and F1 values indicate that TakeFive is competitive with other existing methods such as SEMAFOR, Pikes, PathLSTM, and FRED. We finally discuss how to combine TakeFive and FRED, obtaining higher values of precision, recall, and F1 measure
- …
