Roxana-Mariana Nechita
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Roxana-Mariana Nechita has a total of 47 co-authors. Here are the top co-authors by number of shared publications:
- Dana Corina Deselnicu
Department of Entrepreneurship and Management, Faculty of Entrepreneurship, Business Engineering and Management, National University of Science and Technology POLITEHNICA Bucharest, 060042 Bucharest, Romania
10 shared publications - Iuliana Grecu
Department of Entrepreneurship and Management, Faculty of Entrepreneurship Business Engineering and Management, National University of Science and Technology POLITEHNICA Bucharest, 060042 Bucharest, Romania
9 shared publications - Cozmin Adrian Cristoiu
Faculty of Industrial Engineering and Robotics, National University of Science and Technology Politehnica Bucharest, 060042 Bucharest, Romania
8 shared publications - Corina-Ionela Dumitrescu
Department of Economics, Faculty of Entrepreneurship Business Engineering and Management, National University of Science and Technology POLITEHNICA Bucharest, 060042 Bucharest, Romania
6 shared publications - Marius-Valentin Drăgoi
Faculty of Industrial Engineering and Robotics, National University of Science and Technology Politehnica Bucharest, 060042 Bucharest, Romania
5 shared publications - Oliver Ulerich
National University of Science and Technology Politehnica Bucharest
5 shared publications
Latest Publications
- Open Access
- 15 October 2026
- Published by MDPI AG in Biomimetics
- Vol. 11 (10) , 740
Spatial Decoupling of Foot-Substrate Cross-Talk in Plantar Pressure Sensor Arrays
Measuring plantar pressure with smart insoles is widely used in gait analysis, medical rehabilitation, and clinical monitoring, but the continuous elastomeric layers of hybrid footwear transmit force between adjacent sensors and produce mechanical cross-talk that distorts local readings. The objective of this study was to develop a deterministic framework that identifies the spatial mechanical influence between sensors and compensates for it after signal acquisition. Controlled vertical loads were applied individually to each sensor, while the responses of all sensors were recorded each time as a vector and placed in a matrix; the resulting force data, complemented by a distance-dependent transfer function, formed the direct influence matrix of the Decision-Making Trial and Evaluation Laboratory (DEMATEL) method. Normalization and inversion generated a static decoupling operator applied to the vector of raw measurements. Secondary signals, which initially ranged from 40 N to 970 N, were reduced to a residual band from 3 N to 75 N (97.8% attenuation). The system, once calibrated in this way, provides a clearer spatial representation of local plantar pressure without applying changes to the physical structure of the insole and can be executed through a single matrix-vector multiplication on microcontrollers used in wearable platforms with limited resources.
- 15 October 2026
- Published by MDPI AG in Biomimetics
- Vol. 11 (10) , 740
Open AccessSpatial Decoupling of Foot-Substrate Cross-Talk in Plantar Pressure Sensor Arrays
Measuring plantar pressure with smart insoles is widely used in gait analysis, medical rehabilitation, and clinical monitoring, but the continuous elastomeric layers of hybrid footwear transmit force between adjacent sensors and produce mechanical cross-talk that distorts local readings. The objective of this study was to develop a deterministic framework that identifies the spatial mechanical influence between sensors and compensates for it after signal acquisition. Controlled vertical loads were applied individually to each sensor, while the responses of all sensors were recorded each time as a vector and placed in a matrix; the resulting force data, complemented by a distance-dependent transfer function, formed the direct influence matrix of the Decision-Making Trial and Evaluation Laboratory (DEMATEL) method. Normalization and inversion generated a static decoupling operator applied to the vector of raw measurements. Secondary signals, which initially ranged from 40 N to 970 N, were reduced to a residual band from 3 N to 75 N (97.8% attenuation). The system, once calibrated in this way, provides a clearer spatial representation of local plantar pressure without applying changes to the physical structure of the insole and can be executed through a single matrix-vector multiplication on microcontrollers used in wearable platforms with limited resources.
- Open Access
A Multi-Stage Framework for Examining the Internal Activity and Refinement of the Cybersecurity Risk Mitigation System in Financial-Banking Environments
The rapid digitalization of financial banking services has increased the need for effective cybersecurity risk mitigation as institutions rely on interconnected digital infrastructures. Understanding how technological, organizational, and human-related factors interact is important for supporting cybersecurity planning and risk management. This study examines the internal structure of cybersecurity risk mitigation in financial banking systems through a framework that combines bibliometric analysis, the Decision Making Trial and Evaluation Laboratory (DEMATEL) method, and the Matrix Impact Cross-reference Multiplication Applied to Classification (MICMAC) technique. The bibliometric analysis identified the factors most frequently associated with cybersecurity risk mitigation in scientific publications indexed in the Web of Science Core Collection. These factors were subsequently evaluated by a panel of IT and cybersecurity experts from the banking sector. DEMATEL was applied to examine the influence relationships among the identified factors, and MICMAC was used to refine the system by classifying the factors according to their driving and dependence power. The analysis provides a structured representation of the cybersecurity risk mitigation system and clarifies the role of each factor within the overall structure. The proposed framework may support managers, cybersecurity specialists and decision makers in cybersecurity planning, resource allocation, and strategic decision making within financial institutions.
Open AccessA Multi-Stage Framework for Examining the Internal Activity and Refinement of the Cybersecurity Risk Mitigation System in Financial-Banking Environments
The rapid digitalization of financial banking services has increased the need for effective cybersecurity risk mitigation as institutions rely on interconnected digital infrastructures. Understanding how technological, organizational, and human-related factors interact is important for supporting cybersecurity planning and risk management. This study examines the internal structure of cybersecurity risk mitigation in financial banking systems through a framework that combines bibliometric analysis, the Decision Making Trial and Evaluation Laboratory (DEMATEL) method, and the Matrix Impact Cross-reference Multiplication Applied to Classification (MICMAC) technique. The bibliometric analysis identified the factors most frequently associated with cybersecurity risk mitigation in scientific publications indexed in the Web of Science Core Collection. These factors were subsequently evaluated by a panel of IT and cybersecurity experts from the banking sector. DEMATEL was applied to examine the influence relationships among the identified factors, and MICMAC was used to refine the system by classifying the factors according to their driving and dependence power. The analysis provides a structured representation of the cybersecurity risk mitigation system and clarifies the role of each factor within the overall structure. The proposed framework may support managers, cybersecurity specialists and decision makers in cybersecurity planning, resource allocation, and strategic decision making within financial institutions.
- Open Access
- 31 August 2026
- Published by MDPI AG in Technologies
- Vol. 14 (9) , 538
The Synergy Between Industrial Robots and Artificial Intelligence at a Global Level in the Industry 5.0 Era
The transition from Industry 4.0 to Industry 5.0 has led to a rapid increase in scientific publications addressing human–machine collaboration, artificial intelligence and sustainable manufacturing, which makes it difficult to obtain a coherent overview of the main research directions and global collaboration patterns in this field. This study aims to clarify the development of Industry 5.0 by combining a conceptual overview of its enabling technologies with a bibliometric analysis of scientific articles indexed in the Web of Science Core Collection. The paper outlines the role of key technologies such as computer vision, machine learning, edge computing and collaborative robots in supporting adaptive, human-centred and resource-efficient production systems. The bibliometric component examines publication trends, country-level research activity and thematic relationships between keywords in order to identify dominant topics and emerging areas of interest. The results indicate a concentration of research output in countries with strong investments in advanced manufacturing and artificial intelligence, as well as a growing association between Industry 5.0 and themes such as human–robot collaboration, predictive maintenance, digital twins and ethical aspects of intelligent systems. The analysis also highlights the appearance of new research clusters related to workforce adaptation, resilient supply chains and sustainable production strategies. By providing both a synthesis of technological developments and a quantitative mapping of the literature, this study offers an updated overview of how Industry 5.0 is currently interpreted within the scientific community and supports the identification of future research directions and industrial implementation priorities.
- 31 August 2026
- Published by MDPI AG in Technologies
- Vol. 14 (9) , 538
Open AccessThe Synergy Between Industrial Robots and Artificial Intelligence at a Global Level in the Industry 5.0 Era
The transition from Industry 4.0 to Industry 5.0 has led to a rapid increase in scientific publications addressing human–machine collaboration, artificial intelligence and sustainable manufacturing, which makes it difficult to obtain a coherent overview of the main research directions and global collaboration patterns in this field. This study aims to clarify the development of Industry 5.0 by combining a conceptual overview of its enabling technologies with a bibliometric analysis of scientific articles indexed in the Web of Science Core Collection. The paper outlines the role of key technologies such as computer vision, machine learning, edge computing and collaborative robots in supporting adaptive, human-centred and resource-efficient production systems. The bibliometric component examines publication trends, country-level research activity and thematic relationships between keywords in order to identify dominant topics and emerging areas of interest. The results indicate a concentration of research output in countries with strong investments in advanced manufacturing and artificial intelligence, as well as a growing association between Industry 5.0 and themes such as human–robot collaboration, predictive maintenance, digital twins and ethical aspects of intelligent systems. The analysis also highlights the appearance of new research clusters related to workforce adaptation, resilient supply chains and sustainable production strategies. By providing both a synthesis of technological developments and a quantitative mapping of the literature, this study offers an updated overview of how Industry 5.0 is currently interpreted within the scientific community and supports the identification of future research directions and industrial implementation priorities.
- Open Access
Delimiting the Human Resource System for Project Success in R&D and Exploring Factor Grouping Patterns: Implications for Strategic Decision-Making and Organisational Systems
This study represents the third stage of a structured research program focused on human-related factors in the success of research and development (R&D) projects. Building on prior work that identified 49 success factors and examined relationships among 28 human-centric variables using the Decision-Making Trial and Evaluation Laboratory (DEMATEL) method, the present paper addresses the lack of formal validation of factor inclusion within the system. In this study, validation refers to assessing whether the identified factors are sufficiently integrated into the network of relationships to justify their inclusion within the human resource system associated with R&D project success. The research introduces a multi-layer analytical approach to refine the initial framework into a validated system. First, Matrix Impact Cross-Reference Multiplication Applied to a Classification (MICMAC) is applied to the influence matrix derived from DEMATEL to classify factors based on influence and dependence. Autonomous variables with low influence and low dependence are removed, resulting in a reduced set of structurally integrated factors. Second, Multidimensional Scaling (MDS) is used to examine similarity patterns among the remaining factors and to provide an additional perspective on the consistency of the system structure. Third, the coefficient is applied to analyse co-occurrence patterns and identify groups of factors that tend to appear together. The results support the delimitation of a core human resource system composed of factors that occupy central positions within the interaction structure and exhibit stable association patterns. This integrated approach reduces ambiguity in factor selection and provides a clearer representation of how human-related factors are organised within R&D projects. The findings contribute to the literature on project success by offering a structured procedure for system delimitation and by identifying priority areas that can support managerial and organisational decision-making in R&D environments.Open AccessDelimiting the Human Resource System for Project Success in R&D and Exploring Factor Grouping Patterns: Implications for Strategic Decision-Making and Organisational Systems
This study represents the third stage of a structured research program focused on human-related factors in the success of research and development (R&D) projects. Building on prior work that identified 49 success factors and examined relationships among 28 human-centric variables using the Decision-Making Trial and Evaluation Laboratory (DEMATEL) method, the present paper addresses the lack of formal validation of factor inclusion within the system. In this study, validation refers to assessing whether the identified factors are sufficiently integrated into the network of relationships to justify their inclusion within the human resource system associated with R&D project success. The research introduces a multi-layer analytical approach to refine the initial framework into a validated system. First, Matrix Impact Cross-Reference Multiplication Applied to a Classification (MICMAC) is applied to the influence matrix derived from DEMATEL to classify factors based on influence and dependence. Autonomous variables with low influence and low dependence are removed, resulting in a reduced set of structurally integrated factors. Second, Multidimensional Scaling (MDS) is used to examine similarity patterns among the remaining factors and to provide an additional perspective on the consistency of the system structure. Third, the coefficient is applied to analyse co-occurrence patterns and identify groups of factors that tend to appear together. The results support the delimitation of a core human resource system composed of factors that occupy central positions within the interaction structure and exhibit stable association patterns. This integrated approach reduces ambiguity in factor selection and provides a clearer representation of how human-related factors are organised within R&D projects. The findings contribute to the literature on project success by offering a structured procedure for system delimitation and by identifying priority areas that can support managerial and organisational decision-making in R&D environments. - Open Access
- 13 July 2026
- Published by MDPI AG in Sustainability
- Vol. 18 (14) , 7135
Causal Analysis of Reverse Logistics Determinants for Sustainable E-Commerce Supply Chains
The continued growth of e-commerce has changed the role of reverse logistics, turning it from a supporting activity into an important element of sustainable and resilient supply chain management. Handling product returns efficiently requires a clear understanding of how different logistics factors influence one another within the broader system. This study examines and models the causal links among six major factors through the Decision-Making Trial and Evaluation Laboratory (DEMATEL) method. Using evaluations provided by a panel of domain experts, the analysis measures the strength of interactions between the selected variables and identifies those that act as driving forces as well as those that mainly respond to influences from other factors. The resulting model offers a practical basis for decision-making by showing which factors generate the strongest effects across the reverse logistics network. This perspective helps managers direct available resources toward the elements capable of producing the greatest influence on overall system performance. As a result, organizations can improve the use of resources while supporting both operational efficiency and environmental objectives throughout the supply chain. By applying a causal analysis framework, this research delivers a useful instrument for understanding return management processes in e-commerce. The findings support more informed strategic decisions and provide guidance for companies seeking to improve sustainability performance within a market environment shaped by circular economy principles and frequent change.
- 13 July 2026
- Published by MDPI AG in Sustainability
- Vol. 18 (14) , 7135
Open AccessCausal Analysis of Reverse Logistics Determinants for Sustainable E-Commerce Supply Chains
The continued growth of e-commerce has changed the role of reverse logistics, turning it from a supporting activity into an important element of sustainable and resilient supply chain management. Handling product returns efficiently requires a clear understanding of how different logistics factors influence one another within the broader system. This study examines and models the causal links among six major factors through the Decision-Making Trial and Evaluation Laboratory (DEMATEL) method. Using evaluations provided by a panel of domain experts, the analysis measures the strength of interactions between the selected variables and identifies those that act as driving forces as well as those that mainly respond to influences from other factors. The resulting model offers a practical basis for decision-making by showing which factors generate the strongest effects across the reverse logistics network. This perspective helps managers direct available resources toward the elements capable of producing the greatest influence on overall system performance. As a result, organizations can improve the use of resources while supporting both operational efficiency and environmental objectives throughout the supply chain. By applying a causal analysis framework, this research delivers a useful instrument for understanding return management processes in e-commerce. The findings support more informed strategic decisions and provide guidance for companies seeking to improve sustainability performance within a market environment shaped by circular economy principles and frequent change.
- Open Access
- 11 June 2026
- Published by MDPI AG in Sustainability
- Vol. 18 (12) , 6008
A Methodological Framework for Environmental Compliance Assessment Under the Do No Significant Harm (DNSH) Principle in EU-Funded Projects
The assessment of the “Do No Significant Harm” (DNSH) principle in European Union (EU)-funded projects currently relies on narrative justification, which generates subjective evaluations, inconsistent results, and high administrative effort. This study aims to develop an operational framework and project-level tool to standardize how environmental impact is measured across multiple sectors and project types. The methodology applies a stepwise, non-compensatory approach, combining typology-based filtering, financial thresholds derived from carbon intensity and sustainability coefficients, checklists, spatial analysis, and quantitative indicators such as the circular economy transition metric. Each environmental objective is evaluated independently, ensuring that compliance cannot be offset by positive performance in other areas. The framework was preliminarily validated using a dataset of 1406 projects implemented in Romania, indicating its potential to distinguish low-risk from high-risk projects, reduce evaluator subjectivity, and improve the proportionality of analytical effort. While the tool is tested on Romanian case studies, its design allows for application across various European funding programmemes. The tool supports early-stage screening, encourages green procurement, and aligns project planning with EU environmental objectives, including climate mitigation, adaptation, water resource protection, pollution prevention, circular economy, and biodiversity conservation. The proposed methodology provides a clear, reproducible, and practical approach, offering evaluators a consistent mechanism for DNSH compliance verification and integrating environmental protection into project design and implementation.
- 11 June 2026
- Published by MDPI AG in Sustainability
- Vol. 18 (12) , 6008
Open AccessA Methodological Framework for Environmental Compliance Assessment Under the Do No Significant Harm (DNSH) Principle in EU-Funded Projects
The assessment of the “Do No Significant Harm” (DNSH) principle in European Union (EU)-funded projects currently relies on narrative justification, which generates subjective evaluations, inconsistent results, and high administrative effort. This study aims to develop an operational framework and project-level tool to standardize how environmental impact is measured across multiple sectors and project types. The methodology applies a stepwise, non-compensatory approach, combining typology-based filtering, financial thresholds derived from carbon intensity and sustainability coefficients, checklists, spatial analysis, and quantitative indicators such as the circular economy transition metric. Each environmental objective is evaluated independently, ensuring that compliance cannot be offset by positive performance in other areas. The framework was preliminarily validated using a dataset of 1406 projects implemented in Romania, indicating its potential to distinguish low-risk from high-risk projects, reduce evaluator subjectivity, and improve the proportionality of analytical effort. While the tool is tested on Romanian case studies, its design allows for application across various European funding programmemes. The tool supports early-stage screening, encourages green procurement, and aligns project planning with EU environmental objectives, including climate mitigation, adaptation, water resource protection, pollution prevention, circular economy, and biodiversity conservation. The proposed methodology provides a clear, reproducible, and practical approach, offering evaluators a consistent mechanism for DNSH compliance verification and integrating environmental protection into project design and implementation.
- Open Access
- 4 June 2026
- Published by MDPI AG in Biomimetics
- Vol. 11 (6) , 395
Hybrid Decision-Making Management for Material Selection in the Design of Wearable Pressure-Sensing Orthoses in Neurorehabilitation
Wearable pressure-sensing orthoses are increasingly used in neurorehabilitation to support gait recovery, monitor plantar pressure distribution, and improve patient mobility during repetitive therapy sessions. The performance of these devices depends strongly on the materials used in the skin-contact layer, since material properties influence comfort, flexibility, durability, and force transmission during daily use. This study proposes a hybrid multi-criteria decision-making framework based on the Analytic Hierarchy Process (AHP) and the VIKOR method for material selection in sensor-integrated plantar orthoses. Five candidate materials, ethylene vinyl acetate (EVA), polyethylene (PE), polyurethane (PU), cobalt–chromium–molybdenum alloy (CoCrMo), and polypropylene (PP), were evaluated using five criteria: comfort and skin compatibility, elasticity, fatigue resistance, density, and energy dissipation. AHP was applied to determine the relative importance of the evaluation criteria using expert judgment, while VIKOR was used to rank the material alternatives and identify the compromise solution. The results showed that polyurethane achieved the best overall performance due to its balanced behavior in comfort, elasticity, and fatigue resistance, which are essential properties for long-term wearable neurorehabilitation devices. A sensitivity analysis confirmed that moderate variations in expert weighting did not modify the final ranking. Compared with conventional selection approaches based mainly on isolated material properties, the proposed framework offers a clear and reproducible method for integrating mechanical and user-related requirements into the material selection process for wearable orthoses.
- 4 June 2026
- Published by MDPI AG in Biomimetics
- Vol. 11 (6) , 395
Open AccessHybrid Decision-Making Management for Material Selection in the Design of Wearable Pressure-Sensing Orthoses in Neurorehabilitation
Wearable pressure-sensing orthoses are increasingly used in neurorehabilitation to support gait recovery, monitor plantar pressure distribution, and improve patient mobility during repetitive therapy sessions. The performance of these devices depends strongly on the materials used in the skin-contact layer, since material properties influence comfort, flexibility, durability, and force transmission during daily use. This study proposes a hybrid multi-criteria decision-making framework based on the Analytic Hierarchy Process (AHP) and the VIKOR method for material selection in sensor-integrated plantar orthoses. Five candidate materials, ethylene vinyl acetate (EVA), polyethylene (PE), polyurethane (PU), cobalt–chromium–molybdenum alloy (CoCrMo), and polypropylene (PP), were evaluated using five criteria: comfort and skin compatibility, elasticity, fatigue resistance, density, and energy dissipation. AHP was applied to determine the relative importance of the evaluation criteria using expert judgment, while VIKOR was used to rank the material alternatives and identify the compromise solution. The results showed that polyurethane achieved the best overall performance due to its balanced behavior in comfort, elasticity, and fatigue resistance, which are essential properties for long-term wearable neurorehabilitation devices. A sensitivity analysis confirmed that moderate variations in expert weighting did not modify the final ranking. Compared with conventional selection approaches based mainly on isolated material properties, the proposed framework offers a clear and reproducible method for integrating mechanical and user-related requirements into the material selection process for wearable orthoses.
- Open Access
- 1 June 2026
- Published by MDPI AG in Biomimetics
- Vol. 11 (6) , 386
Strategic Management of Design and Conceptualization Factors for Wearable Postural Rehabilitation Devices: A Causal Interdependency Analysis
The implementation of wearable systems in postural rehabilitation offers new perspectives for continuous monitoring, yet their success depends on the interaction between technical parameters and clinical requirements. This study analyzes clinical performance factors to identify strategic levers determining recovery effectiveness. It examines indicators such as postural deviation detection, sensitivity to minor motion changes, suitability for home monitoring, continuous monitoring capability, clinical relevance of extracted parameters, and the capability to assess patient progress over time. Using the DEMATEL methodology, the study highlights influences among these factors in a clinical context. This structural analysis separates primary drivers from rehabilitation outcomes. To refine the analysis, the MICMAC method classifies factors by driving and dependence power, distinguishing determinant, relay, dependent, and autonomous variables. The approach provides an objective basis for managers and designers to prioritize resources toward functionalities with the greatest systemic impact on patient progress. The combined DEMATEL–MICMAC framework enhances decision-making by linking causal relationships with clear hierarchical categorization. The findings may support the integration of wearable technologies into rehabilitation practice by identifying the clinical performance factors with the strongest influence within the system.
- 1 June 2026
- Published by MDPI AG in Biomimetics
- Vol. 11 (6) , 386
Open AccessStrategic Management of Design and Conceptualization Factors for Wearable Postural Rehabilitation Devices: A Causal Interdependency Analysis
The implementation of wearable systems in postural rehabilitation offers new perspectives for continuous monitoring, yet their success depends on the interaction between technical parameters and clinical requirements. This study analyzes clinical performance factors to identify strategic levers determining recovery effectiveness. It examines indicators such as postural deviation detection, sensitivity to minor motion changes, suitability for home monitoring, continuous monitoring capability, clinical relevance of extracted parameters, and the capability to assess patient progress over time. Using the DEMATEL methodology, the study highlights influences among these factors in a clinical context. This structural analysis separates primary drivers from rehabilitation outcomes. To refine the analysis, the MICMAC method classifies factors by driving and dependence power, distinguishing determinant, relay, dependent, and autonomous variables. The approach provides an objective basis for managers and designers to prioritize resources toward functionalities with the greatest systemic impact on patient progress. The combined DEMATEL–MICMAC framework enhances decision-making by linking causal relationships with clear hierarchical categorization. The findings may support the integration of wearable technologies into rehabilitation practice by identifying the clinical performance factors with the strongest influence within the system.
- Open Access
- 20 May 2026
- Published by MDPI AG in Applied Sciences
- Vol. 16 (10) , 5107
A Python-Based Framework for Learning-from-Demonstration in Robotic Object Sorting: Comparative Evaluation of Lightweight Classifiers
This paper presents a Python-based v3.12 framework for robotic object sorting in a virtual workcell, combining learning-from-demonstration with a comparative evaluation of classical machine learning classifiers. A user provides a minimal demonstration (e.g., one cube and one cylinder placed into two bins) from which a dynamic type-to-bin rule is inferred. In this study, learning-from-demonstration is implemented at the level of rule acquisition from minimal task examples rather than at the level of trajectory imitation or low-level motion teaching. This rule is used to relabel a larger dataset of pre-generated object positions, enabling training with a selectable number of file-based samples (2–1600) optionally augmented with manual samples. Five classifiers—decision tree, k-nearest neighbors, logistic regression, naive Bayes, and linear SVM—were trained and then used to drive autonomous pick-and-place execution while logging replication time and correctness (correct/incorrect moves and accuracy). Because the task reaches accuracy saturation under a deterministic rule, an additional offline inference benchmark was included to compare prediction throughput using 10,000 probes with repeated timing (median over 50 runs or mean ± standard deviation over 30 runs). To complement this nominal evaluation, the framework also included a perturbation-aware robustness protocol based on controlled positional perturbation, systematic bias, controlled shape corruption, repeated perturbation voting, and stability-aware scoring. This additional layer makes it possible to examine classifier behavior under controlled uncertainty, especially in reduced-data settings, without changing the compact simulator-based nature of the workflow. Results indicate identical sorting accuracy across models, while inference-time differences remain measurable, highlighting deployment-oriented trade-offs and confirming that end-to-end cycle time is dominated by robot motion rather than model computation.
- 20 May 2026
- Published by MDPI AG in Applied Sciences
- Vol. 16 (10) , 5107
Open AccessA Python-Based Framework for Learning-from-Demonstration in Robotic Object Sorting: Comparative Evaluation of Lightweight Classifiers
This paper presents a Python-based v3.12 framework for robotic object sorting in a virtual workcell, combining learning-from-demonstration with a comparative evaluation of classical machine learning classifiers. A user provides a minimal demonstration (e.g., one cube and one cylinder placed into two bins) from which a dynamic type-to-bin rule is inferred. In this study, learning-from-demonstration is implemented at the level of rule acquisition from minimal task examples rather than at the level of trajectory imitation or low-level motion teaching. This rule is used to relabel a larger dataset of pre-generated object positions, enabling training with a selectable number of file-based samples (2–1600) optionally augmented with manual samples. Five classifiers—decision tree, k-nearest neighbors, logistic regression, naive Bayes, and linear SVM—were trained and then used to drive autonomous pick-and-place execution while logging replication time and correctness (correct/incorrect moves and accuracy). Because the task reaches accuracy saturation under a deterministic rule, an additional offline inference benchmark was included to compare prediction throughput using 10,000 probes with repeated timing (median over 50 runs or mean ± standard deviation over 30 runs). To complement this nominal evaluation, the framework also included a perturbation-aware robustness protocol based on controlled positional perturbation, systematic bias, controlled shape corruption, repeated perturbation voting, and stability-aware scoring. This additional layer makes it possible to examine classifier behavior under controlled uncertainty, especially in reduced-data settings, without changing the compact simulator-based nature of the workflow. Results indicate identical sorting accuracy across models, while inference-time differences remain measurable, highlighting deployment-oriented trade-offs and confirming that end-to-end cycle time is dominated by robot motion rather than model computation.
- Open Access
- 11 May 2026
- Published by MDPI AG in Applied Sciences
- Vol. 16 (10) , 4786
Python Software Application for Obstacle-Avoiding Path Planning in RoboDK Using Free Space Graph and Robot Level Validation
This paper presents a software application developed in Python (v3.9) for obstacle avoidance trajectory planning in the RoboDK (v6.0) virtual environment. The proposed method automatically scans the virtual station, identifies obstacles, discretizes the workspace into a three-dimensional free-space-graph (FSG) and searches for candidate routes between start and finish points. Each route is then verified at the robot level by inverse kinematics and collision control, and the validated solutions can be transformed into preview curves, intermediate points and motion programs executable by the robot. The study includes an initial test scenario performed with the ABB IRB 6650-125/3.2 robot and randomly generated obstacles, followed by a series of benchmark tests performed in different virtual scenarios and with four different robot models. In the comparative tests, the proposed method was evaluated together with a rapidly exploring random tree (RRT) reference planner and the native probabilistic roadmap (PRM) planner, embedded RoboDK. The final scenario included a robotic cell with realistic objects. The results show that the application can identify valid executable routes and, in some cases, several alternative variants for the same pair of target points. Overall, the benchmark suggests that the analyzed methods have different strengths and should be viewed as complementary solutions. In the tested scenarios, RRT was the method with the lowest computational times, while the proposed method offered the possibility of generating several alternative routes. At the current stage, the application can thus be used as an offline programming tool but also as a research and analysis tool for planning robotic trajectories in the presence of static obstacles.
- 11 May 2026
- Published by MDPI AG in Applied Sciences
- Vol. 16 (10) , 4786
Open AccessPython Software Application for Obstacle-Avoiding Path Planning in RoboDK Using Free Space Graph and Robot Level Validation
This paper presents a software application developed in Python (v3.9) for obstacle avoidance trajectory planning in the RoboDK (v6.0) virtual environment. The proposed method automatically scans the virtual station, identifies obstacles, discretizes the workspace into a three-dimensional free-space-graph (FSG) and searches for candidate routes between start and finish points. Each route is then verified at the robot level by inverse kinematics and collision control, and the validated solutions can be transformed into preview curves, intermediate points and motion programs executable by the robot. The study includes an initial test scenario performed with the ABB IRB 6650-125/3.2 robot and randomly generated obstacles, followed by a series of benchmark tests performed in different virtual scenarios and with four different robot models. In the comparative tests, the proposed method was evaluated together with a rapidly exploring random tree (RRT) reference planner and the native probabilistic roadmap (PRM) planner, embedded RoboDK. The final scenario included a robotic cell with realistic objects. The results show that the application can identify valid executable routes and, in some cases, several alternative variants for the same pair of target points. Overall, the benchmark suggests that the analyzed methods have different strengths and should be viewed as complementary solutions. In the tested scenarios, RRT was the method with the lowest computational times, while the proposed method offered the possibility of generating several alternative routes. At the current stage, the application can thus be used as an offline programming tool but also as a research and analysis tool for planning robotic trajectories in the presence of static obstacles.