Kanata Suzuki, Hanon Nakamurama, Kana Miyamoto
et al.
Robotic laboratory automation has traditionally relied on carefully engineered motion pipelines and task-specific hardware interfaces, resulting in high design cost and limited flexibility. While recent imitation learning techniques can generate general robot behaviors, their large model sizes often require high-performance computational resources, limiting applicability in practical laboratory environments. In this study, we propose a compact imitation learning framework for laboratory automation using small foundation models. The proposed method, TVF-DiT, aligns a self-supervised vision foundation model with a vision-language model through a compact adapter, and integrates them with a Diffusion Transformer-based action expert. The entire model consists of fewer than 500M parameters, enabling inference on low-VRAM GPUs. Experiments on three real-world laboratory tasks - test tube cleaning, test tube arrangement, and powder transfer - demonstrate an average success rate of 86.6%, significantly outperforming alternative lightweight baselines. Furthermore, detailed task prompts improve vision-language alignment and task performance. These results indicate that small foundation models, when properly aligned and integrated with diffusion-based policy learning, can effectively support practical laboratory automation with limited computational resources.
Vítor Tinoco, Manuel F. Silva, Filipe Neves dos Santos
et al.
Agriculture needs to produce more with fewer resources to satisfy the world’s demands. Labor shortages, especially during harvest seasons, emphasize the need for agricultural automation. However, the high cost of commercially available robotic manipulators, ranging from EUR 3000 to EUR 500,000, is a significant barrier. This research addresses the challenges posed by low-cost manipulators, such as inaccuracy, limited sensor feedback, and dynamic uncertainties. Three control strategies for a low-cost agricultural SCARA manipulator were developed and benchmarked: a Sliding Mode Controller (SMC), a Reinforcement Learning (RL) Controller, and a novel Proportional-Integral (PI) controller with a self-tuning feedforward element (PIFF). The results show the best response time was obtained using the SMC, but with joint movement jitter. The RL controller showed sudden breaks and overshot upon reaching the setpoint. Finally, the PIFF controller showed the smoothest reference tracking but was more susceptible to changes in system dynamics.
Improving public transport connectivity is crucial for decarbonisation and economic growth. Current transport planning approaches to addressing connectivity problems rely on trial-and-error approaches to identify problems and generate options, limited by planners' incomplete knowledge and the overwhelming volume of available travel data.We introduce a machine-assisted approach to identify opportunities for connectivity enhancements from origin-destination data and generate prioritised intervention options. Using an origin-destination matrix for Greater London covering approximately 1200 activity centres, our method applies trajectory clustering to identify potential high-demand corridors with poor public transport quality.Our prototype automatically generates multiple public transport scheme options (local bus, express bus, metro) within these corridors along with approximate operating costs. These options are batch-tested using accelerated assignment modelling that optimises mode choice, frequency, and route generation, and the results are given ordered according to benefit-cost ratios.This approach is now being used to supplement human planners’ knowledge in the development of new express bus services in London.
Transportation and communications, Transportation engineering
Tareq Salameh, Mena Maurice Farag, Abdul-Kadir Hamid
et al.
This study addresses the fundamental challenge of accurately forecasting power generation from photovoltaic (PV) systems, which is crucial for effective grid integration and energy management. The intermittency and variability of solar power due to environmental factors pose significant difficulties in achieving reliable predictions. An adaptive neuro-fuzzy inference system (ANFIS) model is proposed for forecasting the performance of a 2.88 kW on-grid PV system in Sharjah, UAE. The model leverages extensive real-time data collected during the peak energy generation season to predict critical variables such as the maximum power point (MPP), voltage, and current. The ANFIS model achieves high prediction accuracy, with a Coefficient of Determination (R2) of 0.9967 for power generation, 0.9076 for voltage generation, and 0.9913 for current generation. These results highlight the model’s robustness in capturing the nonlinear dependencies between environmental factors and PV output. The study compares the ANFIS model with other established machine learning models, including Linear Regression, Decision Tree, Support Vector Machine (SVM), and Random Forest. The ANFIS model outperforms these models in terms of prediction accuracy, demonstrating its superior generalization capabilities. The findings underscore the potential of the ANFIS model for robust forecasting and effective PV performance management, providing a reliable tool for early fault detection and system assessment. Future work will focus on integrating fault detection capabilities and extending model validation across different seasons to ensure a comprehensive investigation of the system dynamics under fluctuating weather conditions.
Stem removal from harvested fruits remains one of the most labor-intensive tasks in fruit harvesting, directly affecting the fruit quality and marketability. Accurate and rapid fruit and stem segmentation techniques are essential for automating this process. This study proposes an enhanced You Only Look Once (YOLO) model, AppleStem (AS)-YOLO, which uses a ghost bottleneck and global attention mechanism to segment apple stems. The proposed model reduces the number of parameters and enhances the computational efficiency using the ghost bottleneck while improving feature extraction capabilities using the global attention mechanism. The model was evaluated using both a custom-built and an open dataset, which will be later released to ensure reproducibility. Experimental results demonstrated that the AS-YOLO model achieved high accuracy, with a mean average precision (mAP)@50 of 0.956 and mAP@50–95 of 0.782 across all classes, along with a real-time inference speed of 129.8 frames per second (FPS). Compared with state-of-the-art segmentation models, AS-YOLO exhibited superior performance. The proposed AS-YOLO model demonstrates the potential for real-time application in automated fruit-harvesting systems, contributing to the advancement of agricultural automation.
Fanyu Wang, Chetan Arora, Chakkrit Tantithamthavorn
et al.
Automated software testing has significant potential to enhance efficiency and reliability within software development processes. However, its broader adoption faces considerable challenges, particularly concerning alignment between test generation methodologies and software requirements. REquirements-Driven Automated Software Testing (REDAST) addresses this gap by systematically leveraging requirements as the foundation for automated test artifact generation. This systematic literature review (SLR) critically examines the REDAST landscape, analyzing the current state of requirements input formats, transformation techniques, generated test artifacts, evaluation methods, and prevailing limitations. We conducted a thorough analysis of 156 relevant studies selected through a rigorous multi-stage filtering process from an initial collection of 27,333 papers sourced from six major research databases. Our findings highlight the predominance of functional requirements, model-based specifications, and natural language formats. Rule-based techniques are extensively utilized, while machine learning-based approaches remain relatively underexplored. Furthermore, most existing frameworks are sequential and dependent on singular intermediate representations, and while test cases, structured textual formats, and requirements coverage are common, full automation remains rare. We identify significant gaps related to automation completeness and dependency on input quality. This comprehensive synthesis provides a detailed overview of REDAST research and limitations, offering clear, evidence-based recommendations to guide future advancements in automated software testing.
Geothermal energy provides continuous low emission potential but is underused in Qatar because of high capital costs, drilling risks, and uncertainty in subsurface conditions. This study examines how automation can improve the techno economic and environmental feasibility of geothermal deployment through three pathways: Enhanced Geothermal Systems in the Dukhan Basin, repurposed oil and gas wells, and ground source heat pumps for district cooling. Using geological datasets and financial modeling, the analysis shows that full automation reduces capital expenditure by 12 to 14 percent and operating expenditure by 14 to 17 percent. The Levelized Cost of Energy decreases from 145 USD per MWh to 125 USD per MWh, and payback periods shorten by up to two years. Environmental results indicate that geothermal substitution can avoid between 4000 and 17600 tons of CO2 per year for each project. Automation also reduces uncertainty in investment outcomes based on Monte Carlo simulations. Overall, the results show that automation strengthens the economic viability of geothermal systems and supports their integration into Qatars long term energy diversification and decarbonization strategies.
As artificial intelligence (AI) systems become increasingly embedded in critical societal functions, the need for robust red teaming methodologies continues to grow. In this forum piece, we examine emerging approaches to automating AI red teaming, with a particular focus on how the application of automated methods affects human-driven efforts. We discuss the role of labor in automated red teaming processes, the benefits and limitations of automation, and its broader implications for AI safety and labor practices. Drawing on existing frameworks and case studies, we argue for a balanced approach that combines human expertise with automated tools to strengthen AI risk assessment. Finally, we highlight key challenges in scaling automated red teaming, including considerations around worker proficiency, agency, and context-awareness.
This dissertation addresses a topic that I have worked on over the past decade: the automation of next-to-leading order electroweak corrections in the Standard Model of particle physics. After introducing the basic concepts and techniques of next-to-leading order QCD calculations that underpin the MadGraph5_aMC@NLO framework, I present a few key features relevant to the automated next-to-leading order electroweak contributions to short-distance cross sections, with an emphasis on the mixed QCD and electroweak coupling expansions. These include the FKS subtraction, the renormalization and electroweak input parameter schemes, and the complex mass scheme for dealing with unstable particles. Issues related to the initial or final photons and leptons are also discussed. Two remaining challenges are highlighted if one wishes to go beyond next-to-leading order computations. Some phenomenological applications at the LHC are given to demonstrate the relevance of electroweak corrections at colliders. Finally, an outlook on future studies concludes the dissertation.
Fangyuan Zhou, Zhaohui Tang, Xiaolong Zhang
et al.
With the development of electrified railways towards high speed and heavy load, the peak power of traction loads is increasing, and the maximum demand and negative sequence current of traction substations are also increasing. Therefore, this article proposes an energy storage system (ESS) based on Li-ion batteries for regulating the maximum demand of traction substations. An ESS is connected to the DC bus of a railway power conditioner (RPC), which is connected to the two power supply arms of the traction substation. In response to the large fluctuation of traction load, this paper proposes a maximum demand active regulation method based on short-term prediction of traction load. The short-term prediction of traction load adopts a time series short-term load prediction method based on BP neural network error correction. Then, based on the load prediction value of the traction substation and the state of charge of the ESS, a collaborative control strategy for ESS and RPC is formulated to enable RPC to achieve a negative sequence suppression function simultaneously. Finally, simulation experiments were conducted using MATLAB, and the results showed that compared with the traditional maximum demand regulation method based on peak power reference values, the method proposed in this paper significantly reduces the number of ESS charging and discharging cycles, improves the regulation effect of maximum demand, and has a higher net income during the lifecycle. At the same time, it also takes into account the negative sequence current suppression function, thereby improving the comprehensive economic benefits of railways and the quality of power grids.
Iacob Postavaru, Emilia Bunea, Crina Pungulescu
et al.
This paper explores the potential of large language models to enhance economics education through computational humor. We employ OpenAI’s GPT-4 model to infuse humor into summaries of three Nobel laureates’ contributions to economics and conduct a small empirical exercise with undergraduate students to test the pedagogical efficacy of computational humor. The results suggest that computer-generated humor may be an effective learning aid: the results of the students who rate the humorous versions of the instructional texts as genuinely funny are significantly better than the results of their peers who are not amused. Encouragingly for teachers who try to be funny but fail, we do not find evidence that ineffectual humor is detrimental to learning.
Thomas Schilling, Rebecca Müller, Thomas Ellwart
et al.
Many organizations use decision support systems (DSS) to support DSS users in their daily work demands (e.g., high workload, insufficient information, ambiguous situations). A key question regarding their interaction is how the decision-control is divided between the DSS and the user, represented by the system's level of automation (LoA). To investigate the need for an adaptable DSS where users can manually adjust the LoA across situations, we used a vignette design to examine whether users prefer different LoA in different situations (i.e., six situational criteria, each manipulated by two specifications; e.g., low vs. high workload). In the twelve vignettes, the 116 participants should imagine working in an emergency control-center—a setting they were familiar with from previous experiments. Our results showed significant differences between the two corresponding vignettes, indicating that users prefer different LoA across situations. However, after controlling for the participants' overall preference for a situation-independent baseline LoA, the significant differences between all paired vignettes disappear. Our results have implications for whether situational or individual criteria are more important regarding LoA preferences, adaptable DSS, and for human-centered design based on user profiles. We discuss our findings in relation to the broader literature on trust and acceptance of DSS.