reinforcement learning

LGBTIQ+ lives under colonial shadows: The legacy of German colonial law in today’s Tanzania and Namibia

Colonial legacies have profoundly shaped the post-colonial landscapes of African countries, influencing all aspects of life, including the reinforcement of a singular, heteronormative conception of sexuality. This legacy continues to shape the perceptions of sexual orientation, gender identity, gender expression and sex characteristics (SOGIESC) in Africa today, as ongoing Western influence dictates both legal frameworks and societal norms. This creates serious lasting challenges for lesbian, gay, bisexual, trans, intersex and queer (LGBTIQ+) communities, especially in relation to discriminatory laws that were established in colonial times and continue to persist across many African jurisdictions. Realising the relatively short period of German colonial rule, in this chapter I take a closer look at the role of German colonial legislation in relation to contemporary legal and societal attitudes towards the rights of LGBTIQ+ communities in what today are Namibia and Tanzania. Given that both countries have experienced different legal and societal developments on issues of sexuality and gender until today, this chapter aims to highlight historical injustices and their ongoing repercussions in a comparative approach. In doing this, the chapter offers three key learnings: First, the influence of German colonial rule on contemporary legal frameworks regarding same-sex sexual practices in Namibia and Tanzania is complex and difficult to trace. Nonetheless, the colonial imposition of rigid heteronormative structures and binary gender norms, used as instruments of power, shaped social perceptions of sexuality in ways that continue to disadvantage LGBTIQ+ individuals. Second, both countries illustrate how colonial legacies interact with post-colonial trajectories: despite divergent post-independence paths, inherited societal norms continue to influence approaches to gender and sexuality, reflecting broader African struggles to abolish colonial frameworks. Third, German colonial legal histories not only illuminate internal historical realities, but also inform global debates on law, sexuality and human rights, highlighting specific colonial structures whose effects persist today.

A cascaded PID-reinforcement learning-based virtual inertia control in microgrid load frequency control system using electric vehicle energy storage

The islanded microgrids increasingly depend on renewable energy sources for power generation and introduce significant frequency control challenges due to the renewable sources’ intermittent nature and low system inertia. Traditional energy storage systems, though employed and effective for frequency stabilization, are often limited by high costs and power density requirements. Accounting issues of microgrid frequency performance under stochastic renewable energy source integration and limitations of conventional energy storage systems, this paper proposes a virtual inertia control strategy that leverages electric vehicle battery storage, supported by a cascade proportional integral derivative-reinforcement learning-based auxiliary controller, to enhance frequency regulation of the load frequency control system in an islanded microgrid. The proposed cascade proportional integral derivative-reinforcement learning-based virtual inertia controller is implemented and tested in a MATLAB/Simulink environment and evaluated under diverse operating conditions involving dynamic load and renewable energy source disturbances, while comparing with other control strategies. The comparative results demonstrate that the proposed virtual inertia control strategy outperforms other compared controllers in terms of frequency stability and overall dynamic response.

Artificial Intelligence for Economic Resilience and Global Stability

The current world economies are in a highly volatile framework characterized by thick interdependencies, and quick changing risk factors. The classical econometric models with their assumption of the stasis and restrictive data granularity cannot predict systemic shocks and timely interventions. This chapter provides a technical review of how Artificial Intelligence can increase economic resilience via predictive governance. It is a conceptualization of AI as a multi-layered analytical architecture that achieves high-frequency data streams, machine-learning prediction models, and policy optimization structures. The capabilities of deep learning, reinforcement learning, and network-based models to detect the emergent signals, forecast macroeconomic anomalies and simulate the counterfactual policy outcomes are put under stress. Transparency in algorithms, interpretable models and ethics in macro-level applications are also discussed in the chapter. It can be argued that AI-enabled predictive governance is necessary to develop resilient, adaptive, sustainable economic systems.

Energy-cost-aware cooperative scheduling for cascade slurry pumping stations: A multi-agent deep reinforcement learning approach

Robust indoor automated guided vehicle navigation for logistics applications via imitation-augmented hierarchical reinforcement learning with a memory-attention generalized network

Comportamientos exitosos del trabajo en suelo durante el campeonato venezolano de Judo 2025, primera categoría: Estudio Observacional

The aim of this investigation was to determine sucessfull behaviors asociated to grounwork from the participants. Under a non-experimental design, 51 actions from 1477 performed attacks were evaluated by: Submission form (inmobilization, strangled, articular luxation); precedent maneuvers, opponent initial position, performed throws, promotor action (continous attack, direct attack, failed attack, scored transition). Results showed that inmovilization followed by strangulation were the main submissions, asociated by sex (p=0,030). They were preceded by an inmediate inmovilization after landing or a turnover; not sex associated (p=0,167) starting from a opponent in sideways or 4 points posittion; again not sex associated (p=0,271). From the 18 performed throws previos to ground work, there was a mild sex association (p=0,043) and a significant one to initial opponent possition (p=0,028); where ippon and seoi nage throws were associated to 4 points positions meanwhile Tai otoshi was with face down and lastly, O soto gari and O uchi gaeshi were with sideways and face up. In conlusion, athletes hold or choke by a turnover or throw trayectorie's control, similar to international reports. Groundwork learning reinforcement using the aforemention data could benefit combat skill and long term adherence

Cooperative control of multi-microgrids under uncertainty: a multi-agent actor-critic approach with LSTM and deep dense architecture

The high penetration of renewable energy poses significant source-load uncertainty challenges to the economic and stable operation of microgrids. To address the key challenges of multi-microgrid cooperative optimization, including temporal source-load uncertainty, heterogeneous inter-microgrid coordination, and real-time decentralized decision-making, this paper proposes a cooperative control strategy named Long Short-Term Memory-based Multi-Agent Actor-Critic (L-MAAC). Under the Centralized Training with Decentralized Execution (CTDE) framework, the proposed method learns cooperative policies using global system information during training, while enabling each microgrid to make decisions based only on local observations during execution. Specifically, the Actor network incorporates an LSTM module to extract temporal features from historical source-load data, thereby improving the ability to proactively respond to renewable generation and load fluctuations, while the Critic network integrates an individual attention mechanism with a Deep Dense architecture in Reinforcement Learning (D2RL), which enhances the representation of heterogeneous interactions among microgrids and alleviates information loss in deep value-function approximation. Ablation studies, reward sensitivity, and hyperparameter sensitivity analyses confirm the synergistic contributions of these components. Simulation results based on a modified IEEE 13-bus multi-microgrid system show that, compared with advanced algorithms such as MAAC, L-MAAC reduces the total system operating cost by 4.81% on the test set. Furthermore, under multiple uncertainty scenarios, including cloud dynamics, wind power randomness, sudden load changes, and combined electricity price/diesel generator failures, the proposed method demonstrates superior voltage regulation capability and robustness. These results indicate that L-MAAC can effectively improve both the economic efficiency and operational stability of multi-microgrid systems under uncertain environments.

Optimization of dynamic admission control for preemptible cloud services based on TPD-SAC reinforcement learning

Abstract Preemptible cloud services utilize idle resources from contract-based services. The resources allocated to these low-priority services can be preempted by contract-based services during periods of overload, so cloud platforms offer low prices to attract cost-sensitive users. While preemptible cloud services enhance resource utilization and boost revenue, excessive admission of service requests may induce overbooking effects, resulting in an excessively high preemption rate that undermines user satisfaction and ultimately lowers revenue. Conversely, insufficient admissions may result in underutilized resources. The stochastic nature of idle resource capacities necessitates real-time dynamic optimization of admission control. The primary difficulty in optimizing dynamic admission control lies in the fact that fluctuations in idle resource capacity are a special category of stochastic fluctuations characterized by non-stationarity, where the stochastic distribution patterns evolve over time. To address this difficulty, we first formulate the admission control for preemptible cloud services with overbooking effects as a Non-Stationary Markov Decision Process (NSMDP). Then, an improved soft actor-critic (SAC) framework is proposed for the NSMDP model. In this framework, the regularization and feature-enhanced LSTM algorithm is employed for turning point detection (TPD) in non-stationary stochastic fluctuation patterns. On this basis, a soft actor-critic algorithm based on turning point detection (TPD-SAC) is developed to achieve real-time resolution of NSMDP. Experiments with Google Cloud datasets validate the effectiveness of the proposed algorithm in non-stationary environments. Furthermore, simulation experiments were conducted to analyze the revenue contributions of dynamic admission strategies under two pricing mechanisms. Experimental results indicate that compared to strategies under the uniform discount mechanism, dynamic admission strategies achieve higher revenue contributions under the interruption-based discount mechanism.

Social information quality and environmental volatility shape collective foraging behaviour

Abstract Collective foraging is widespread across the animal kingdom, allowing animals to more effectively discover resources. However, collective foragers need to balance a key trade-off between private exploration and using social information. Social information can come in very distinct forms, ranging from simple positional cues to complex pay-off information. However, how the types of available social cues and environmental volatility shape collective foraging behaviour is not well understood. We address this using a spatially explicit model in which agents track a mobile resource via multi-agent reinforcement learning (MARL). Agents choose between random exploration, private tracking and social attraction. We systematically varied resource volatility and the type of available social cues to analyse their effect on individual and collective behaviour. Our results show that the quality of social information dictates the emerging collective behaviour. Low-quality social cues (e.g. positions and actions) result in a fragile strategy that is effective in stable environments but fails as volatility increases. Conversely, high-quality social information (e.g. pay-offs) enables behavioural diversity: agents selectively copy others and flexibly change between individual tracking or exploration depending on the environmental volatility. Our findings identify the interplay between information quality and ecological context as an important mechanism governing the emergence of distinct forms of collective behaviour from individual decision rules.

AI-Driven Autonomous Test Automation Frameworks for Cloud Native Applications: Improving Software Reliability and Continuous Delivery

Cloud native applications are built and released through continuous integration and continuous delivery pipelines that call for fast feedback and frequent deployment. Traditional test automation, which depends on scripts that are written once and maintained by hand, was never really designed for this pace of change. When a microservice is redeployed, refactored, or scaled several times a day, the scripts that were written to test it tend to fall behind, and quality assurance teams end up spending more time repairing tests than writing new ones. This paper looks at how artificial intelligence can be used to build a test automation framework that adjusts itself as the application changes, rather than breaking every time something moves. We describe a framework, referred to here as AutoQA CN, that brings together three capabilities: automatic generation of new test cases from API specifications and usage data, self-healing of broken test scripts through similarity-based and learned matching, and risk-based prioritisation of test execution using a reinforcement learning agent. The framework is built to sit alongside existing continuous integration tooling in a containerised environment rather than replace it outright. We evaluate the approach through an illustrative case study across three representative microservice scenarios and compare it with a conventional scripted automation baseline on three measures: maintenance effort, time to detect faults, and pipeline throughput. The results point to meaningful reductions in maintenance work and faster fault detection, along with a modest gain in throughput. We close by discussing where these gains are likely to hold up in practice, where the approach still needs human oversight, and what would need to happen for a framework like this to be trusted in production settings.

AI-Based Optimization for Biofuel Production: Strategies for Utilizing Degraded Land for Climate Change Mitigation, Green Finance Mobilization, and Achieving United Nations Sustainable Development Goals

Global land degradation affects approximately 2 billion hectares, threatening food security, biodiversity, and climate stability while undermining the United Nations Sustainable Development Goals (SDGs). The concurrent urgency to decarbonize the energy system and mobilize green finance for sustainable transitions has created a rare policy window in which AI-optimized biofuel production on degraded lands can simultaneously serve multiple imperatives. This study presents a comprehensive secondary data analysis of AI-based optimization frameworks for deploying biofuel production systems on degraded lands, integrating an explicit green finance dimension that has been largely absent from prior synthesis literature. Drawing on 152 peer-reviewed studies and authoritative datasets from FAO, IEA, IRENA, UNCCD, the Green Climate Fund (GCF), and the World Bank, we analyze machine learning, deep learning, reinforcement learning, and hybrid AI architectures applied to feedstock selection, soil remediation, yield prediction, supply-chain logistics, and green finance risk-return optimization. Our findings reveal that AI-optimized biofuel systems on degraded lands recover 75-94% of prime-land bioenergy yields, sequester 8.3-10.5 t CO2e ha-1 over 30 years, reduce lifecycle GHG emissions by 55-88%, and generate internal rates of return of 9-22% when green finance instruments are systematically integrated. Green bonds, Article 6 carbon credits, GCF concessional finance, and blended finance structures are identified as the most impactful instruments, collectively capable of reducing project risk scores by 30-45% and expanding the investable universe of degraded-land biofuel projects by an estimated 340%. We develop the AI-Biofuel-Land Restoration (ABLR) conceptual framework with explicit green finance routing pathways and identify critical policy enablers for global deployment. This study advances the evidence base for policy-makers, investors, researchers, and development practitioners working at the intersection of artificial intelligence, bioenergy, green finance, and sustainable land management.

Numerical study on hydrogen-air explosion dynamic characteristics with deep reinforcement learning