Pub Date : 2025-11-27DOI: 10.1109/THMS.2025.3627872
Sizhuang Zhang;Ying Sun;Derui Ding;Hui Yu
High-fidelity 3-D face reconstruction is critical for enhancing personalized and immersive human–machine interaction experiences. However, existing methods struggle to capture the full spectrum of facial textures, particularly fine-scale details, such as wrinkles and pores, due to limitations in multiscale representation. To address this challenge, we propose a self-supervised multiscale hierarchical network to hierarchically model fine geometric details in multiple scales in this study. We design a global and local Markov random field loss and a detail perception loss to provide a global and local sensory field of view guidance for retaining fine-scale detail structure information of the face. In addition, we introduce a learnable Gabor-aware texture enhancement module to enhance the network’s sensitivity to fine textures. Extensive experiments show that the proposed method can reconstruct fine-scale details of the face and has superior performance to the state-of-the-art methods in terms of reconstruction accuracy and visual effect.
{"title":"SMHNet: Self-Supervised Multiscale Hierarchical Network for High Fidelity 3-D Face Reconstruction","authors":"Sizhuang Zhang;Ying Sun;Derui Ding;Hui Yu","doi":"10.1109/THMS.2025.3627872","DOIUrl":"https://doi.org/10.1109/THMS.2025.3627872","url":null,"abstract":"High-fidelity 3-D face reconstruction is critical for enhancing personalized and immersive human–machine interaction experiences. However, existing methods struggle to capture the full spectrum of facial textures, particularly fine-scale details, such as wrinkles and pores, due to limitations in multiscale representation. To address this challenge, we propose a self-supervised multiscale hierarchical network to hierarchically model fine geometric details in multiple scales in this study. We design a global and local Markov random field loss and a detail perception loss to provide a global and local sensory field of view guidance for retaining fine-scale detail structure information of the face. In addition, we introduce a learnable Gabor-aware texture enhancement module to enhance the network’s sensitivity to fine textures. Extensive experiments show that the proposed method can reconstruct fine-scale details of the face and has superior performance to the state-of-the-art methods in terms of reconstruction accuracy and visual effect.","PeriodicalId":48916,"journal":{"name":"IEEE Transactions on Human-Machine Systems","volume":"56 1","pages":"114-123"},"PeriodicalIF":4.4,"publicationDate":"2025-11-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146045339","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-11-25DOI: 10.1109/THMS.2025.3631490
Xinrui Tao;Yuping Tu;Jiadi Liu;Ying Wang;Fan Yang;Quyuan Wang
In the era of rapid artificial intelligence development, human–agent collaboration holds the potential to significantly enhance work efficiency. While existing studies have explored various collaboration strategies and resource methods, there remains a notable lack of in-depth research on how to economically allocate a limited budget to acquire both human and agent computing capacities. To address this gap, we first construct a developer model based on theories from psychology and economics, providing a quantitative description of human working time and efficiency. Building upon this, we further investigate the impact of dynamic budget allocation strategies on consumer decision-making. Specifically, a novel budget recycling mechanism is introduced to redistribute unused resources, thereby enhancing system responsiveness. Experimental results demonstrate a 56% improvement in resource utilization and a 32% increase in task completion. This confirms the effectiveness of our proposed method in optimizing collaboration and supporting sustainable project execution.
{"title":"A Recycling-Driven Dynamic Budget Allocation Strategy for Human–Agent Collaboration","authors":"Xinrui Tao;Yuping Tu;Jiadi Liu;Ying Wang;Fan Yang;Quyuan Wang","doi":"10.1109/THMS.2025.3631490","DOIUrl":"https://doi.org/10.1109/THMS.2025.3631490","url":null,"abstract":"In the era of rapid artificial intelligence development, human–agent collaboration holds the potential to significantly enhance work efficiency. While existing studies have explored various collaboration strategies and resource methods, there remains a notable lack of in-depth research on how to economically allocate a limited budget to acquire both human and agent computing capacities. To address this gap, we first construct a developer model based on theories from psychology and economics, providing a quantitative description of human working time and efficiency. Building upon this, we further investigate the impact of dynamic budget allocation strategies on consumer decision-making. Specifically, a novel budget recycling mechanism is introduced to redistribute unused resources, thereby enhancing system responsiveness. Experimental results demonstrate a 56% improvement in resource utilization and a 32% increase in task completion. This confirms the effectiveness of our proposed method in optimizing collaboration and supporting sustainable project execution.","PeriodicalId":48916,"journal":{"name":"IEEE Transactions on Human-Machine Systems","volume":"56 1","pages":"182-191"},"PeriodicalIF":4.4,"publicationDate":"2025-11-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146045327","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-11-25DOI: 10.1109/THMS.2025.3627578
Hyesun Chung;X. Jessie Yang
This study aims to explore the associations between individuals’ trust dynamics in automated/autonomous technologies and their personal characteristics, and to further examine whether personal characteristics can be used to predict a user’s trust dynamics type. The experimental data involved 130 participants who performed a simulated surveillance task that consisted of a compensatory tracking task and a threat detection task. An imperfect automated threat detector assisted participants in the detection task. Using a pre-experimental survey covering 12 constructs and 28 dimensions, we collected data on participants’ personal characteristics. Based on the experimental data, we performed k-means clustering and identified three trust dynamics types. Subsequently, we conducted one-way analyses of variance to evaluate differences among the three trust dynamics types in terms of personal characteristics, behaviors, performance, and postexperimental ratings. Participants were clustered into three groups, namely Bayesian decision makers, disbelievers, and oscillators. Results showed that the clusters differ significantly in seven personal characteristics: masculinity, positive affect, extraversion, neuroticism, intellect, performance expectancy, and high expectations. The disbelievers tend to have high neuroticism and low performance expectancy. The oscillators tend to have higher scores in masculinity, positive affect, extraversion, and intellect. We also found significant differences in behaviors, performance, and postexperimental ratings across the three groups. The disbelievers are the least likely to blindly follow the recommendations made by the automated threat detector. Based on the significant personal characteristics, we developed a decision tree model to predict the trust dynamics type with an accuracy of 70% . This model offers promising implications for identifying individuals whose trust dynamics may deviate from a Bayesian pattern.
{"title":"Predicting Trust Dynamics Type Using Seven Personal Characteristics","authors":"Hyesun Chung;X. Jessie Yang","doi":"10.1109/THMS.2025.3627578","DOIUrl":"https://doi.org/10.1109/THMS.2025.3627578","url":null,"abstract":"This study aims to explore the associations between individuals’ trust dynamics in automated/autonomous technologies and their personal characteristics, and to further examine whether personal characteristics can be used to predict a user’s trust dynamics type. The experimental data involved 130 participants who performed a simulated surveillance task that consisted of a compensatory tracking task and a threat detection task. An imperfect automated threat detector assisted participants in the detection task. Using a pre-experimental survey covering 12 constructs and 28 dimensions, we collected data on participants’ personal characteristics. Based on the experimental data, we performed k-means clustering and identified three trust dynamics types. Subsequently, we conducted one-way analyses of variance to evaluate differences among the three trust dynamics types in terms of personal characteristics, behaviors, performance, and postexperimental ratings. Participants were clustered into three groups, namely Bayesian decision makers, disbelievers, and oscillators. Results showed that the clusters differ significantly in seven personal characteristics: masculinity, positive affect, extraversion, neuroticism, intellect, performance expectancy, and high expectations. The disbelievers tend to have high <italic>neuroticism</i> and low <italic>performance expectancy</i>. The oscillators tend to have higher scores in <italic>masculinity</i>, <italic>positive affect</i>, <italic>extraversion</i>, and <italic>intellect</i>. We also found significant differences in behaviors, performance, and postexperimental ratings across the three groups. The disbelievers are the least likely to blindly follow the recommendations made by the automated threat detector. Based on the significant personal characteristics, we developed a decision tree model to predict the trust dynamics type with an accuracy of 70% . This model offers promising implications for identifying individuals whose trust dynamics may deviate from a Bayesian pattern.","PeriodicalId":48916,"journal":{"name":"IEEE Transactions on Human-Machine Systems","volume":"56 1","pages":"147-159"},"PeriodicalIF":4.4,"publicationDate":"2025-11-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146045347","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-11-19DOI: 10.1109/THMS.2025.3627559
Bohao Tian;Dinghao Xue;Yilei Zheng;Shijun Zhang;Yuru Zhang;Dangxiao Wang
The ability to rapidly acquire novel visuomotor skills is essential for daily functioning tasks such as motor rehabilitation, surgical operation, and mechanical assembly. Previous research suggested that experiencing flow can enhance learning outcomes. Although dynamic difficulty adjustment (DDA) has been commonly used to induce flow and maximize engagement, most existing methods rely on model-free, stepwise adaptations that lack quantitative, model-based support. In this study, we proposed a personalized, model-driven DDA approach to facilitate visuomotor skill acquisition by enhancing flow during the learning process. We implemented an adaptive fine fingertip force control task with DDA based on optimal control principles to train the visuomotor skills. This DDA updated task difficultly using real-time multiple performance metrics, with parameters derived from an individually fitted model that captures each user's motor behavior during the task. A user study, involving two groups, compared the effects of a model-driven adaptive task with a model-free control task. Results from the flow state scale and physiological recordings demonstrated that the model-driven task elicited significantly higher levels of flow than the model-free task. Moreover, participants in the model-driven group showed a notably higher learning rate in visuomotor skills (19%) compared to the model-free group (8%). These findings underscore the potential of integrating personalized modeling and optimal control theory to optimize user experience and accelerate learning outcomes in DDA frameworks when building adaptive human–machine interaction systems.
{"title":"Personalized Model-Driven Adaptive Task Facilitates Visuomotor Skill Learning Mediated by Promoting Flow Experience","authors":"Bohao Tian;Dinghao Xue;Yilei Zheng;Shijun Zhang;Yuru Zhang;Dangxiao Wang","doi":"10.1109/THMS.2025.3627559","DOIUrl":"https://doi.org/10.1109/THMS.2025.3627559","url":null,"abstract":"The ability to rapidly acquire novel visuomotor skills is essential for daily functioning tasks such as motor rehabilitation, surgical operation, and mechanical assembly. Previous research suggested that experiencing flow can enhance learning outcomes. Although dynamic difficulty adjustment (DDA) has been commonly used to induce flow and maximize engagement, most existing methods rely on model-free, stepwise adaptations that lack quantitative, model-based support. In this study, we proposed a personalized, model-driven DDA approach to facilitate visuomotor skill acquisition by enhancing flow during the learning process. We implemented an adaptive fine fingertip force control task with DDA based on optimal control principles to train the visuomotor skills. This DDA updated task difficultly using real-time multiple performance metrics, with parameters derived from an individually fitted model that captures each user's motor behavior during the task. A user study, involving two groups, compared the effects of a model-driven adaptive task with a model-free control task. Results from the flow state scale and physiological recordings demonstrated that the model-driven task elicited significantly higher levels of flow than the model-free task. Moreover, participants in the model-driven group showed a notably higher learning rate in visuomotor skills (19%) compared to the model-free group (8%). These findings underscore the potential of integrating personalized modeling and optimal control theory to optimize user experience and accelerate learning outcomes in DDA frameworks when building adaptive human–machine interaction systems.","PeriodicalId":48916,"journal":{"name":"IEEE Transactions on Human-Machine Systems","volume":"56 1","pages":"160-170"},"PeriodicalIF":4.4,"publicationDate":"2025-11-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146045302","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-11-18DOI: 10.1109/THMS.2025.3620237
Matthew J. Ball;Amin Khademi;Alfredo M. Carbonell;Jackie S. Cha
Laparoscopic surgery training tasks are a prime example of tasks requiring psychomotor skills. A typical evaluation of psychomotor skills can be accomplished using learning curves. Learning curves have been generated from the performance metrics within specific psychomotor skill tasks previously; however, these learning curves do not directly measure the information that a repetition of the task provides. Thus, we propose an information-theoretic model to measure the information gained from task repetitions. A total of 20 participants repeated a laparoscopic matchboard training task until a proficiency metric was reached. The proposed probability model, used in the information gain framework, was then calibrated using 16 randomly chosen participants’ trials to proficiency and validated by simulating four new sample trials tested against the remaining participant’s data. It was found that the average number of trials to proficiency was 27 trials corresponding to an information gain of 0.0136 units for an extra repetition. Utilizing the information gained for stopping training complements available proficiency metrics for adjudicating proficiency in motor skills tasks and provides several advantages.
{"title":"Application of an Information Gain Model in a Motor Learning Laparoscopic Surgery Task","authors":"Matthew J. Ball;Amin Khademi;Alfredo M. Carbonell;Jackie S. Cha","doi":"10.1109/THMS.2025.3620237","DOIUrl":"https://doi.org/10.1109/THMS.2025.3620237","url":null,"abstract":"Laparoscopic surgery training tasks are a prime example of tasks requiring psychomotor skills. A typical evaluation of psychomotor skills can be accomplished using learning curves. Learning curves have been generated from the performance metrics within specific psychomotor skill tasks previously; however, these learning curves do not directly measure the information that a repetition of the task provides. Thus, we propose an information-theoretic model to measure the information gained from task repetitions. A total of 20 participants repeated a laparoscopic matchboard training task until a proficiency metric was reached. The proposed probability model, used in the information gain framework, was then calibrated using 16 randomly chosen participants’ trials to proficiency and validated by simulating four new sample trials tested against the remaining participant’s data. It was found that the average number of trials to proficiency was 27 trials corresponding to an information gain of 0.0136 units for an extra repetition. Utilizing the information gained for stopping training complements available proficiency metrics for adjudicating proficiency in motor skills tasks and provides several advantages.","PeriodicalId":48916,"journal":{"name":"IEEE Transactions on Human-Machine Systems","volume":"56 1","pages":"105-113"},"PeriodicalIF":4.4,"publicationDate":"2025-11-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146045311","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-11-18DOI: 10.1109/THMS.2025.3620160
Kai Xu;Zhenyu Wang;Yuxuan Long;Rui Zhang
Dialogue policy is a core module in pipeline dialogue systems as it drives conversation generation. Personalized dialogue policies aim to equip chatbots with tailored personalities, making them behave like real users, providing more accurate action responses, and improving the anthropomorphic capabilities of personal assistants. Yet existing dialogue policy approaches often overlook individual personalities because obtaining explicit user profiles is costly and time-consuming. In this article, we propose a personalized dialogue policy learning framework, named PDL. It dynamically learns implicit user profiles from successful dialogue trajectories. Specifically, we collect a lot of success histories from human–computer interactions to extract sequences of user belief states and agent actions. The extracted sequences are processed via a loop clipping operation and modeled with an autoregressive transformer to mimic human analytical behavior. After that, a new user’s latent personalized preferences are predicted based on the autoregressive transformer model. The personalized preferences are employed to implement dialogue policies via three categories of reinforcement learning algorithms, including value-based approaches, policy-based approaches, and model-based approaches. The experiments are conducted on three different task-oriented dialogue datasets, and the results show that the proposed PDL framework achieves state-of-the-art results compared to other comparative approaches.
{"title":"Personalized Dialogue Policy Learning Framework Based on Implicit User Profiles","authors":"Kai Xu;Zhenyu Wang;Yuxuan Long;Rui Zhang","doi":"10.1109/THMS.2025.3620160","DOIUrl":"https://doi.org/10.1109/THMS.2025.3620160","url":null,"abstract":"Dialogue policy is a core module in pipeline dialogue systems as it drives conversation generation. Personalized dialogue policies aim to equip chatbots with tailored personalities, making them behave like real users, providing more accurate action responses, and improving the anthropomorphic capabilities of personal assistants. Yet existing dialogue policy approaches often overlook individual personalities because obtaining explicit user profiles is costly and time-consuming. In this article, we propose a personalized dialogue policy learning framework, named PDL. It dynamically learns implicit user profiles from successful dialogue trajectories. Specifically, we collect a lot of success histories from human–computer interactions to extract sequences of user belief states and agent actions. The extracted sequences are processed via a loop clipping operation and modeled with an autoregressive transformer to mimic human analytical behavior. After that, a new user’s latent personalized preferences are predicted based on the autoregressive transformer model. The personalized preferences are employed to implement dialogue policies via three categories of reinforcement learning algorithms, including value-based approaches, policy-based approaches, and model-based approaches. The experiments are conducted on three different task-oriented dialogue datasets, and the results show that the proposed PDL framework achieves state-of-the-art results compared to other comparative approaches.","PeriodicalId":48916,"journal":{"name":"IEEE Transactions on Human-Machine Systems","volume":"56 1","pages":"95-104"},"PeriodicalIF":4.4,"publicationDate":"2025-11-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146045337","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-11-04DOI: 10.1109/THMS.2025.3619064
Long Chen;Jiatong He;Lei Zhang;Minpeng Xu;Zhongpeng Wang;Dong Ming
Objective: Emergency anticipation (EA) refers to the brain’s rapid perceptual, cognitive, and motor preparation in response to imminent emergencies. Timely decoding of EA can facilitate proactive responses before full behavioral execution, which is critical in real-world scenarios such as avoiding hazards or mitigating accidents. However, the cortical activation underlying the EA process has not been fully explored. This study aims to analyze the neural activity of the EA process and explore the feasibility of detecting emergency motor intention in conjunction with brain-computer interface (BCI) technology. Methods: We designed a new emergency state induction paradigm in the virtual environment, including a target task (emergency anticipation, EA) and two baseline tasks (emergency anticipation execution, EAE, visual observation, VO). A total of 31 healthy subjects were recruited for the offline experiment. The cortical responses during the EA process were quantified by analyzing event-related potential, movement-related cortical potential, and event-related spectral perturbation. Discriminative canonical pattern matching, common spatial patterns, and shrinkage linear discriminant analysis were employed to perform binary classification. Six subjects participated in the pseudo-online asynchronous experiment to valid the feasibility of identifying emergency motor intention. Results: The results showed that the cascading process associated with EA existed in both the temporal and spectral domains. Particularly, temporal domain feature demonstrated superior classification performance, with averages of 90.13% (>80% chance level). The pseudo-online evaluation showed that the system response time with an average of 257.12 ms, which was 35 ms faster than the behavioral response. Significance: Our work demonstrated the cascading process of perceptual recognition, cognitive evaluation, and motor preparation during the EA processes and provided preliminary evidence supporting the feasibility of detecting emergency motor intentions. These findings lay a theoretical foundation for extending the application of BCI technology to rapid control scenarios.
{"title":"Emergency Motor Intention Detection Based on Unpredictable Anticipatory Activity: An EEG Study","authors":"Long Chen;Jiatong He;Lei Zhang;Minpeng Xu;Zhongpeng Wang;Dong Ming","doi":"10.1109/THMS.2025.3619064","DOIUrl":"https://doi.org/10.1109/THMS.2025.3619064","url":null,"abstract":"Objective: Emergency anticipation (EA) refers to the brain’s rapid perceptual, cognitive, and motor preparation in response to imminent emergencies. Timely decoding of EA can facilitate proactive responses before full behavioral execution, which is critical in real-world scenarios such as avoiding hazards or mitigating accidents. However, the cortical activation underlying the EA process has not been fully explored. This study aims to analyze the neural activity of the EA process and explore the feasibility of detecting emergency motor intention in conjunction with brain-computer interface (BCI) technology. Methods: We designed a new emergency state induction paradigm in the virtual environment, including a target task (emergency anticipation, EA) and two baseline tasks (emergency anticipation execution, EAE, visual observation, VO). A total of 31 healthy subjects were recruited for the offline experiment. The cortical responses during the EA process were quantified by analyzing event-related potential, movement-related cortical potential, and event-related spectral perturbation. Discriminative canonical pattern matching, common spatial patterns, and shrinkage linear discriminant analysis were employed to perform binary classification. Six subjects participated in the pseudo-online asynchronous experiment to valid the feasibility of identifying emergency motor intention. Results: The results showed that the cascading process associated with EA existed in both the temporal and spectral domains. Particularly, temporal domain feature demonstrated superior classification performance, with averages of 90.13% (>80% chance level). The pseudo-online evaluation showed that the system response time with an average of 257.12 ms, which was 35 ms faster than the behavioral response. Significance: Our work demonstrated the cascading process of perceptual recognition, cognitive evaluation, and motor preparation during the EA processes and provided preliminary evidence supporting the feasibility of detecting emergency motor intentions. These findings lay a theoretical foundation for extending the application of BCI technology to rapid control scenarios.","PeriodicalId":48916,"journal":{"name":"IEEE Transactions on Human-Machine Systems","volume":"55 6","pages":"993-1005"},"PeriodicalIF":4.4,"publicationDate":"2025-11-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145652132","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-11-04DOI: 10.1109/THMS.2025.3621275
Naveed Ahmad Khan;Prashant K. Jamwal;Fahad Hussain;Wayne Spratford;Shahid Hussain
Optimizing energy transfer during physical human–robot interactions is important for enhancing neurotherapeutic outcomes and ensuring patient safety. Energy transfer dynamics are particularly complex, involving a delicate balance between kinetic and potential energies as the robot assists or resists movement, adapting to the patient’s needs in real time. Traditional methods, which often rely on predefined robot control strategies, often struggle in dynamic environments where the interplay of forces and motions becomes unpredictable. Therefore, this work integrates the computational intelligence of quantum computing with transformer models to estimate the dynamics of energy transfer between human and gait rehabilitation robot, specifically designed based on the Stephenson III six-bar linkage mechanism. The principles of quantum computing, such as superposition and entanglement, combined with the attention mechanisms of transformer models, explore a much larger solution space. It provides accurate predictions of the complex, nonlinear interactions of energy flows between the robot and the human lower limb. The quantum transformer network was trained on the experimental data obtained from the interaction of seven male and one female healthy human subjects with the gait rehabilitation robot operated at low and high impedance control modes.
{"title":"Quantum Enhanced Transformer Network for Learning Transactive Energy During Physical Human-Robot Interaction","authors":"Naveed Ahmad Khan;Prashant K. Jamwal;Fahad Hussain;Wayne Spratford;Shahid Hussain","doi":"10.1109/THMS.2025.3621275","DOIUrl":"https://doi.org/10.1109/THMS.2025.3621275","url":null,"abstract":"Optimizing energy transfer during physical human–robot interactions is important for enhancing neurotherapeutic outcomes and ensuring patient safety. Energy transfer dynamics are particularly complex, involving a delicate balance between kinetic and potential energies as the robot assists or resists movement, adapting to the patient’s needs in real time. Traditional methods, which often rely on predefined robot control strategies, often struggle in dynamic environments where the interplay of forces and motions becomes unpredictable. Therefore, this work integrates the computational intelligence of quantum computing with transformer models to estimate the dynamics of energy transfer between human and gait rehabilitation robot, specifically designed based on the Stephenson III six-bar linkage mechanism. The principles of quantum computing, such as superposition and entanglement, combined with the attention mechanisms of transformer models, explore a much larger solution space. It provides accurate predictions of the complex, nonlinear interactions of energy flows between the robot and the human lower limb. The quantum transformer network was trained on the experimental data obtained from the interaction of seven male and one female healthy human subjects with the gait rehabilitation robot operated at low and high impedance control modes.","PeriodicalId":48916,"journal":{"name":"IEEE Transactions on Human-Machine Systems","volume":"55 6","pages":"930-939"},"PeriodicalIF":4.4,"publicationDate":"2025-11-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145652184","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-11-04DOI: 10.1109/THMS.2025.3617576
Le Anh Dao;Marco Maccarini;Matteo Lavit Nicora;Matteo Meregalli Falerni;Marta Mondellini;Palaniappan Veerappan;Lorenzo Mantovani;Dario Piga;Simone Formentin;Matteo Malosio;Loris Roveda
Black-box optimization involves solving optimization problems where the objective function and/or constraints are unknown, inaccessible, or do not explicitly exist. In many applications, particularly those involving human interaction, the optimization problem can only be accessed through physical experiments, with the available outcomes based on the preference of one candidate over one or more others. Accordingly, algorithms for active preference learning have been developed to exploit this specific information in constructing a surrogate of the objective function. This surrogate is then used to define an acquisition function that suggests new decision vectors to search for the optimal solution iteratively. Based on this idea, our approach aims to extend active preference learning algorithms to leverage further information effectively, which can be obtained in reality, such as: a five-point Likert-type scale for the outcomes of the preference query (i.e., the preference can be described not only as “this is better than that” but also as “this is much better than that”), or multiple outcomes for a single preference query with possible additive information on how certain the outcomes are. The validation of the proposed algorithm is done through some standard benchmark functions, and, in practice, through tuning parameters for robot sealing and human–robot collaboration experiments, showing a promising improvement with respect to the state-of-the-art algorithm in the same context.
{"title":"Experience in Engineering Complex Systems: Active Preference Learning With Multiple Outcomes and Certainty Levels","authors":"Le Anh Dao;Marco Maccarini;Matteo Lavit Nicora;Matteo Meregalli Falerni;Marta Mondellini;Palaniappan Veerappan;Lorenzo Mantovani;Dario Piga;Simone Formentin;Matteo Malosio;Loris Roveda","doi":"10.1109/THMS.2025.3617576","DOIUrl":"https://doi.org/10.1109/THMS.2025.3617576","url":null,"abstract":"Black-box optimization involves solving optimization problems where the objective function and/or constraints are unknown, inaccessible, or do not explicitly exist. In many applications, particularly those involving human interaction, the optimization problem can only be accessed through physical experiments, with the available outcomes based on the preference of one candidate over one or more others. Accordingly, algorithms for active preference learning have been developed to exploit this specific information in constructing a surrogate of the objective function. This surrogate is then used to define an acquisition function that suggests new decision vectors to search for the optimal solution iteratively. Based on this idea, our approach aims to extend active preference learning algorithms to leverage further information effectively, which can be obtained in reality, such as: a five-point Likert-type scale for the outcomes of the preference query (i.e., the preference can be described not only as “this is better than that” but also as “this is much better than that”), or multiple outcomes for a single preference query with possible additive information on how certain the outcomes are. The validation of the proposed algorithm is done through some standard benchmark functions, and, in practice, through tuning parameters for robot sealing and human–robot collaboration experiments, showing a promising improvement with respect to the state-of-the-art algorithm in the same context.","PeriodicalId":48916,"journal":{"name":"IEEE Transactions on Human-Machine Systems","volume":"55 6","pages":"898-908"},"PeriodicalIF":4.4,"publicationDate":"2025-11-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145652171","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-10-28DOI: 10.1109/THMS.2025.3620362
Han Zhang;Yuhan Liu;Liaoyang Zhan;Wanzhong Zhao
As heavy vehicles advance toward increased intelligence and modernization, the control of advanced driver assistance systems for ensuring driving safety faces significant challenges. To enhance the driving safety of heavy vehicles operated by drivers with varying driving styles, this article proposes a human–machine cooperative control (HMCC) strategy that combines steering and braking using deep deterministic policy gradient (DDPG) algorithm. First, a multiagent system is adopted as the framework for the driving safety assistance control system, wherein the active front steering (AFS) system and the differential braking control system (DBC) function as subsystems. These subsystems interact through control sequence information while managing yaw and roll stability. The optimal control performance of both the AFS and DBC is ensured using a distributed model predictive controller and Pareto optimality theory. Second, to analyze different drivers’ driving styles, safety characteristic parameters were collected from multiple drivers. By analyzing the effects of drivers on yaw and roll stability, drivers were classified into three types. Furthermore, an HMCC strategy based on DDPG is designed. Phase plane constraints that consider yaw and roll stability are incorporated into the design of the DDPG reward function, training the agents to allocate cooperative control weights between the driver and the AFS and DBC controllers. Finally, the proposed control strategy’s effectiveness is validated through the electro-hydraulic compound steering and braking hardware-in-the-loop test system, demonstrating its ability to improve driving safety for different driver characteristics.
{"title":"A Human–Machine Cooperative Control Strategy Based on Deep Reinforcement Learning to Enhance Heavy Vehicle Driving Safety","authors":"Han Zhang;Yuhan Liu;Liaoyang Zhan;Wanzhong Zhao","doi":"10.1109/THMS.2025.3620362","DOIUrl":"https://doi.org/10.1109/THMS.2025.3620362","url":null,"abstract":"As heavy vehicles advance toward increased intelligence and modernization, the control of advanced driver assistance systems for ensuring driving safety faces significant challenges. To enhance the driving safety of heavy vehicles operated by drivers with varying driving styles, this article proposes a human–machine cooperative control (HMCC) strategy that combines steering and braking using deep deterministic policy gradient (DDPG) algorithm. First, a multiagent system is adopted as the framework for the driving safety assistance control system, wherein the active front steering (AFS) system and the differential braking control system (DBC) function as subsystems. These subsystems interact through control sequence information while managing yaw and roll stability. The optimal control performance of both the AFS and DBC is ensured using a distributed model predictive controller and Pareto optimality theory. Second, to analyze different drivers’ driving styles, safety characteristic parameters were collected from multiple drivers. By analyzing the effects of drivers on yaw and roll stability, drivers were classified into three types. Furthermore, an HMCC strategy based on DDPG is designed. Phase plane constraints that consider yaw and roll stability are incorporated into the design of the DDPG reward function, training the agents to allocate cooperative control weights between the driver and the AFS and DBC controllers. Finally, the proposed control strategy’s effectiveness is validated through the electro-hydraulic compound steering and braking hardware-in-the-loop test system, demonstrating its ability to improve driving safety for different driver characteristics.","PeriodicalId":48916,"journal":{"name":"IEEE Transactions on Human-Machine Systems","volume":"55 6","pages":"1006-1015"},"PeriodicalIF":4.4,"publicationDate":"2025-10-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145652161","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：