This study presents a promising deep learning-based framework for automatic classification of broccoli maturity stages through various image processing analyses. By leveraging advanced image processing techniques and various deep learning models, the proposed system accurately segments broccoli heads from complex backgrounds, achieving a segmentation accuracy of 98.52%, even under challenging visual conditions such as shadows and partial occlusion by leaves. The subsequent maturity classification phase reaches a training accuracy of 99% and a testing accuracy of 91%, demonstrating strong generalization capabilities. A key strength of the proposed method lies in its resilience to suboptimal image quality, making it suitable for deployment in uncontrolled, real-world environments. Extensive experimental evaluations confirm the effectiveness of the framework, highlighting its potential as a feasible solution for object segmentation and stage classification tasks in pattern recognition scenarios.

Nowadays precision farming techniques uses modern machinery and numerous information technologies such as geolocation and autonomation, IoT, remote sensing. Some of them are designed to collect more data from the crop and soil, which will be the basis to support decisions and perform exact dosing at an exact place and time. The paper discusses a multilevel cybernetic plant model implemented in a smart digital twin of plants as a source of data from the crop in the field. It focuses on development of a model for stage calculation and yield forecast of a digital twin, as well as simulation of plant crops. Described realization of digital twin is aimed to implement precision farming technologies by accurate and adequate decision-making in crop cultivation process. The paper proposes to represent plant model as a complex adaptive system based on ontologies and multi-agent technologies. Scope of digital twins of plants could be widen for various crops due to use of knowledge bases of crop cultivation, which describe the rules of growth and development process for each specie in particular. Designed prototype of digital twin of winter wheat is illustrated in the paper. It also discusses the prospects of developing a digital ecosystem – the digital twin of real biogeocenosis and ecosystem of plant crops.

The paper is devoted to development of a digital twin (DT) of plant. It is built as a smart system based on the knowledge base on macrostages of plant development and multiagent technology that allows for detailed monitoring and control of plant development, recalculation of forecast, namely, assessment of vegetation quality, future yield and timing for onset of next stages. It uses transition rules between stages upon receipt of data from agronomists on the state of plant development, as well as the actual and forecast weather data. The paper proposes a conceptual plant model based on ontologies and multi-agent technology, which is a network of linked states and transition rules that correspond to macrostages of plant development with the possibility of recalculating their parameters. The paper also covers the main principles of agronomist work with the digital twin of plant.

Multi-agent systems (MAS) have been an area of high expectations of the industrial IT community. However, in reality, these expectations are still not met and, in practice, the industry very rarely uses the MAS design methodologies, technologies, and software tools despite the appearance of many new classes of applications for which the MAS paradigm could be the perfect match. This paper analyzes the barriers and trends of the mismatch between the recent industrial anticipations and the real state of the practical use of MAS. It identifies engineering problems with very little re-use of code that currently stops economics of scale and impedes the extensive industrial MAS deployment and the ways to overcome them.

The paper considers the task of constructing a schedule for target application of space remote sensing systems for solving problems of precision farming. It also proposes a method of their solution using multi-agent technology. Presented models and algorithms allow for solving the urgent problem of network interaction between heterogeneous spacecraft group devices for implementation of enlarged and detailed territory surveys, as well as adaptive planning of user requests for shooting. Results of experiments demonstrate higher efficiency of the developed algorithms on large-scale problems in comparison with methods of traditional centralized planning.

The multi-agent approach and technology for distributed scheduling, coordination and control of several UAV in group in order to perform joint tasks in real time was considered. The network-centric multi-agent system for real-time task scheduling for UAV's group was proposed. Two part of sheduling initial scheduling and dynamic rescheduling (task redistributed) were considered. For real-time task redistributed the randomized local voting protocol was used. The architecture of system consists of three basic subsystem. The approach to the organization of subsystems and agents within each subsystem interaction was described).

Holonic enterprises are considered as one of the new types of networking organizations used to increase business efficiency. In practice this requires to introduce a new type of management stimulating the “interpreneurial” approach, active interaction between actors to make coordinated and effective decisions on resource allocation, both within the individual project team and between the teams. In this paper we propose the concept of holonic enterprise based on the principles of leadership management and describe the multi-agent system for supporting decision making to maximize productivity of knowledge based resources. The system is called “Smart Enterprise” and combines top-down project planning with real time on-line bottom-up communication with employees. This provides more coordinated, flexible and efficient resource scheduling in case of unexpected events, which are an inevitable reality when the environment changes. The functionality and architecture of Smart Enterprise system is described, and its usage of ontology and multi-agent technology is presented.

This paper presents the principles of the knowledge-driven adaptive management in manufacturing. The problems of real-time resource allocation, reaction to the unexpected events, on-the-fly update of the knowledge stored in ontology are considered. The possibilities of simultaneous change of the existing factory or workshop processes and schedules according to the information provided by the users are described. Finally, the possible ways of development of the presented approach and its application in production resource management are considered.

This paper discusses the methodology for the development of a strategic planning tool, and particularly the architectural solution to be adopted, within an integrated ICT solution for the improvement of planning and scheduling systems for the manufacturing of small production lots of complex products. After the analysis of different possible architectural solutions, the paper proposes a hybrid solution that combines existing solvers with the multi-agent system principles. The proposed approach introduces several benefits, namely in terms of flexibility, robustness, and achievement of exploratory alternative solutions, playing interactive what-if simulation supporting decision-makers to take strategic decisions.

This paper describes the results of design and implementation of a multi-agent system for real time scheduling and controlling in the Izhevsk Axion-Holding Factory. Lessons learnt, key measured benefits of solution and perspectives are presented.

The paper describes main features of a real-time forecasting and scheduling multi-agent solution designed for the LEGO Company. The design is based on Knowledge Genesis Group own multi-agent platform and technology, which provide real-time adaptive event-driven scheduling to replenish products to LEGO Branded Retail stores. The prototype system has been used to schedule 20 US-based LEGO retail outlets for a yearlong trial period and has achieved the following results:Reduction of lost sale from 40% to 16%; Increase in service level from 66% to 86%;Increase in profitability 56% to 81%. The results show a considerable potential value for full scale LEGO supply chain multi-agent solution which would be able to dynamically and adaptively re-schedule deliveries in real time.