József Váncza

In manufacturing scheduling problems, such as Job Shop Scheduling (JSP), Flow Shop Scheduling (FSP), and Flexible Job Shop Scheduling (FJSP), a single optimal solution may be insufficient to capture the full production context, as it can oversimplify the problem by ignoring its multi-objective nature and lacking contextual adaptation. Human expertise plays a critical role in addressing this issue by identifying the gap between the optimal schedule and the actual production context. However, existing methods lack an effective mechanism for actively integrating human expertise into the optimization process. To address this gap, we propose the human-in-the-loop genetic algorithm (HiLGA), which integrates a two-layer, three-agent framework for multi-objective scheduling optimization. The first layer connects the human operator with the optimization core through two agents: an Interpreter agent which translates natural-language input into structured hint vector encodings, and an Explainer agent which evaluates the Pareto schedules returned by the optimization core and provides conversational support for schedule selection. The second layer serves as the optimization core, where a Generator agent integrates these hints into a genetic algorithm and produces Pareto-optimal schedules.Experiments across JSP, FSP, and FJSP evaluate each agent and agent coupling in sequence. The Generator agent alone achieves rapid convergence and attains 3-19% optimality gaps against the MILP benchmark across the tested configurations, confirming its viability as an evolutionary search operator. When the Interpreter agent is activated, HiLGA achieves 132% mean hypervolume improvement over combined baselines, with 51.7% of Pareto solutions achieving at least 80% preference alignment. Ablation against direct LLM scheduling baselines confirms that neither component alone is sufficient: direct LLM approaches fail in 40.1-43.7% of runs due to invalid or missing schedules, while GA-only search never reaches high-alignment regions, establishing the Interpreter-Generator coupling as the key driver of preference-aware Pareto solutions. The Explainer agent closes the human-in-the-loop cycle by decomposing each Pareto schedule's preference satisfaction into machine-selection, machine-avoidance, and operation-sequence components, allowing operators to query trade-offs and select schedules conversationally over verified optimization results. Overall, HiLGA establishes that natural-language human expertise can be represented as an optimization objective, optimized through Pareto search, and explained at the component level, enabling human workers to actively guide scheduling decisions rather than passively execute algorithmic outputs.

Blank localization (also known as workpiece referencing) is an essential task in machining. It aims to precisely establish the geometric relation of the machine tool (mill, lathe, etc.) and the workpiece. We introduced the concept of multi-operation blank localization to address this task for drilling and milling scenarios in a semi-automated way, which allows positioning different machining features (e.g., different holes) separately in order to exploit the tolerances on the relative position of those features to compensate the small errors of the blank. The method takes as input the measured rough geometry and the machining CNC code, and computes the best possible position of each feature considering machining allowances and tolerances by solving a convex quadratically constrained quadratic program (QCQP). The versatility and extensibility of the Julia language helped the development of this algorithm, materializing in the BlankLocalizationCore.jl package. Its flexibility and ease of use make it an excellent research tool that can be deployed in production as well.

Digital transformation and artificial intelligence are creating an opportunity for innovation across all levels of industry and are transforming the world of work by enabling factories to embrace cutting edge Information Technologies (ITs) into their manufacturing processes. Manufacturing Execution Systems (MESs) are abandoning their traditional role of legacy executing middle-ware for embracing the much wider vision of functional interoperability enablers among autonomous, distributed, and collaborative Cyber-Physical Production System (CPPS). In this paper, we propose a basic methodology for universally modeling, digitalizing, and integrating services offered by a variety of isolated workcells into a single, standardized, and augmented production system. The result is a reliable, reconfigurable, and interoperable manufacturing architecture, which privileges Open Platform Communications Unified Architecture (OPC UA) and its rich possibilities for information modeling at a higher level of the common service interoperability, along with Message Queuing Telemetry Transport (MQTT) lightweight protocols at lower levels of data exchange. The proposed MES architecture has been demonstrated and validated in several use-cases at a research manufacturing laboratory of excellence for industrial testbeds.

Manufacturing became one of the main targets of the coronavirus disease 2019 pandemic. Important questions such as how to deal with the drastic changes (up and down) in demand, and how to keep production ongoing with a decimated workforce and crippled supply chains arose. This technical note gives a short assessment of the praiseworthy reactions of the manufacturing industry, which prevented society from sinking into an even deeper crisis. The pandemic, at the same time, pointed out some weaknesses of present-day manufacturing. Possible long-term consequences of the lessons given and perhaps learned are summarized herein, with a focus on substantial changes and new paradigms that could provide an appropriate armory for the manufacturing industry when responding to similar or even more drastic situations in the future. However, it is yet undecided as to whether the ongoing changes in manufacturing will be accelerated or dismissed, causing manufacturers to fall back into the old groove. Some related questions and dilemmas are also exposed in the end of the paper.