SDSumit Kumar Das

Sumit Kumar Das

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article

Previous Organizations / Institutes

The organizations / institutes from the scholar's other articles in the past ten years


Sumit Kumar Das has a total of 29 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Comparative Analysis of Post Hoc Explainable Methods for Robotic Grasp Failure Prediction

    by Aneseh Alvanpour, Cagla Acun, Kyle Spurlock, Christopher K. Robinson, Sumit K. Das, Dan O. Popa, Olfa Nasraoui

    In human–robot collaborative environments, predicting and explaining robotic grasp failures is crucial for effective operation. While machine learning models can predict failures accurately, they often lack transparency, limiting their utility in critical applications. This paper presents a comparative analysis of three post hoc explanation methods—Tree-SHAP, LIME, and TreeInterpreter—for explaining grasp failure predictions from white-box and black-box models. Using a simulated robotic grasping dataset, we evaluate these methods based on their agreement in identifying important features, similarity in feature importance rankings, dependency on model type, and computational efficiency. Our findings reveal that Tree-SHAP and TreeInterpreter demonstrate stronger consistency with each other than with LIME, particularly for correctly predicted failures. The choice of ML model significantly affects explanation consistency, with simpler models yielding more agreement across methods. TreeInterpreter offers a substantial computational advantage, operating approximately 24 times faster than Tree-SHAP and over 2000 times faster than LIME for complex models. All methods consistently identify effort in joint 1 across fingers 1 and 3 as critical factors in grasp failures, aligning with mechanical design principles. These insights contribute to developing more transparent and reliable robotic grasping systems, enabling better human–robot collaboration through improved failure understanding and prevention.


  • Robot failure mode prediction with deep learning sequence models

    by Khalil Damak, Mariem Boujelbene, Cagla Acun, Aneseh Alvanpour, Sumit K. Das, Dan O. Popa, Olfa Nasraoui


  • Open Access

    Neuroadaptive Controller for Physical Interaction With an Omni-Directional Mobile Nurse Assistant Robot

    by Mohammad Nasser Saadatzi, Shamsudeen Abubakar, Sumit Kumar Das, Mohammad Hossein Saadatze, Dan Popa

    Robot-assisted healthcare could help alleviate the shortage of nursing staff in hospitals and is a potential solution to assist with safe patient handling and mobility. In an attempt to off-load some of the physically-demanding tasks and automate mundane duties of overburdened nurses, we have developed the Adaptive Robotic Nursing Assistant (ARNA), which is a custom-built

  • Open Access

    Acceptability of Using a Robotic Nursing Assistant in Health Care Environments: Experimental Pilot Study

    by Mohammad Nasser Saadatzi, M Cynthia Logsdon, Shamsudeen Abubakar, Sumit Das, Penelope Jankoski, Heather Mitchell, Diane Chlebowy, Dan O Popa

    Background: According to the US Bureau of Labor Statistics, nurses will be the largest labor pool in the United States by 2022, and more than 1.1 million nursing positions have to be filled by then in order to avoid a nursing shortage. In addition, the incidence rate of musculoskeletal disorders in nurses is above average in comparison with other occupations. Robot-assisted health care has the potential to alleviate the nursing shortage by automating mundane and routine nursing tasks. Furthermore, robots in health care environments may assist with safe patient mobility and handling and may thereby reduce the likelihood of musculoskeletal disorders. Objective: This pilot study investigates the perceived ease of use and perceived usefulness (acceptability) of a customized service robot as determined by nursing students (as proxies for nursing staff in health care environments). This service robot, referred to as the Adaptive Robotic Nurse Assistant (ARNA), was developed to enhance the productivity of nurses through cooperation during physical tasks (eg, patient walking, item fetching, object delivery) as well as nonphysical tasks (eg, patient observation and feedback). This pilot study evaluated the acceptability of ARNA to provide ambulatory assistance to patients. Methods: We conducted a trial with 24 participants to collect data and address the following research question: Is the use of ARNA as an ambulatory assistive device for patients acceptable to nurses? The experiments were conducted in a simulated hospital environment. Nursing students (as proxies for nursing staff) were grouped in dyads, with one participant serving as a nurse and the other acting as a patient. Two questionnaires were developed and administrated to the participants based on the Technology Acceptance Model with respect to the two subscales of perceived usefulness and perceived ease of use metrics. In order to evaluate the internal consistency/reliability of the questionnaires, we calculated Cronbach alpha coefficients. Furthermore, statistical analyses were conducted to evaluate the relation of each variable in the questionnaires with the overall perceived usefulness and perceived ease of use metrics. Results: Both Cronbach alpha values were acceptably high (.93 and .82 for perceived usefulness and perceived ease of use questionnaires, respectively), indicating high internal consistency of the questionnaires. The correlation between the variables and the overall perceived usefulness and perceived ease of use metrics was moderate. The average perceived usefulness and perceived ease of use metrics among the participants were 4.13 and 5.42, respectively, out of possible score of 7, indicating a higher-than-average acceptability of this service robot. Conclusions: The results served to identify factors that could affect nurses’ acceptance of ARNA and aspects needing improvement (eg, flexibility, ease of operation, and autonomy level).


  • SkinSim: A Design and Simulation Tool for Robot Skin With Closed-Loop pHRI Controllers

    by Sven Cremer, Mohammad N. Saadatzi, Indika B. Wijayasinghe, Sumit K. Das, Mohammad H. Saadatzi, Dan O. Popa

    The importance of tactile sensing for physical human–robot interaction (pHRI) and dexterous manipulation is well known. SkinSim is a novel simulation environment for tactile robot skins in which one can study design tradeoffs involved in deploying whole-body, dense sensor arrays. In this article, scalable modeling approaches are presented for simulating pressure-sensitive robot skin patches, with simultaneous consideration of sensing element geometry and mechanical structure, signal quality, data processing, and closed-loop force controller performance. The open-source simulation architecture of SkinSim is compatible with Gazebo and robot operating system (ROS) programming environments and supports both robot skin dynamic models, as well as tactile sensing element models. An experimentally validated force dispersion model was introduced for the simulation of sensors embedded in a mechanical damping layer. Simulation examples of robot skin with different tactile resolutions are presented using parameter values extracted from an experimental testbed. Thus, simulation results were experimentally validated and the skin sensor density impact on a simple pHRI controller performance was evaluated. Performance measures include center of pressure (COP) estimation error and control signal settling time, overshoot, and steady-state errors. Results suggest that while COP errors decrease in denser sensor skins, controller performance also deteriorates. Therefore, optimal robot skin designs will have to consider application-dependent tradeoffs. Similar results were confirmed in simulation with a larger skin patch containing approximately 4000 tactels and deployed on the end-effector of a collaborative mobile manipulator. Note to Practitioners —This article was motivated by the expensive and time-consuming process of designing tactile skin for robots. When designing pressure-sensitive whole-body sensor arrays for physical human–robot interaction, there are tradeoffs related to sensor resolution, accuracy, and response time, among others. Instead of building prototypes and evaluating them experimentally, the SkinSim simulation environment allows automatic testing of skins with simultaneous consideration of sensing element geometry and mechanical structure, signal quality, data processing, and closed-loop force controller performance. The user can specify several configurations to test and thereby explore the best tradeoffs before actually prototyping any hardware.


  • Open Access

    Acceptability of Using a Robotic Nursing Assistant in Health Care Environments: Experimental Pilot Study (Preprint)

    by Mohammad Nasser Saadatzi, M Cynthia Logsdon, Shamsudeen Abubakar, Sumit Das, Penelope Jankoski, Heather Mitchell, Diane Chlebowy, Dan O Popa

    BACKGROUND According to the US Bureau of Labor Statistics, nurses will be the largest labor pool in the United States by 2022, and more than 1.1 million nursing positions have to be filled by then in order to avoid a nursing shortage. In addition, the incidence rate of musculoskeletal disorders in nurses is above average in comparison with other occupations. Robot-assisted health care has the potential to alleviate the nursing shortage by automating mundane and routine nursing tasks. Furthermore, robots in health care environments may assist with safe patient mobility and handling and may thereby reduce the likelihood of musculoskeletal disorders. OBJECTIVE This pilot study investigates the perceived ease of use and perceived usefulness (acceptability) of a customized service robot as determined by nursing students (as proxies for nursing staff in health care environments). This service robot, referred to as the Adaptive Robotic Nurse Assistant (ARNA), was developed to enhance the productivity of nurses through cooperation during physical tasks (eg, patient walking, item fetching, object delivery) as well as nonphysical tasks (eg, patient observation and feedback). This pilot study evaluated the acceptability of ARNA to provide ambulatory assistance to patients. METHODS We conducted a trial with 24 participants to collect data and address the following research question: Is the use of ARNA as an ambulatory assistive device for patients acceptable to nurses? The experiments were conducted in a simulated hospital environment. Nursing students (as proxies for nursing staff) were grouped in dyads, with one participant serving as a nurse and the other acting as a patient. Two questionnaires were developed and administrated to the participants based on the Technology Acceptance Model with respect to the two subscales of perceived usefulness and perceived ease of use metrics. In order to evaluate the internal consistency/reliability of the questionnaires, we calculated Cronbach alpha coefficients. Furthermore, statistical analyses were conducted to evaluate the relation of each variable in the questionnaires with the overall perceived usefulness and perceived ease of use metrics. RESULTS Both Cronbach alpha values were acceptably high (.93 and .82 for perceived usefulness and perceived ease of use questionnaires, respectively), indicating high internal consistency of the questionnaires. The correlation between the variables and the overall perceived usefulness and perceived ease of use metrics was moderate. The average perceived usefulness and perceived ease of use metrics among the participants were 4.13 and 5.42, respectively, out of possible score of 7, indicating a higher-than-average acceptability of this service robot. CONCLUSIONS The results served to identify factors that could affect nurses’ acceptance of ARNA and aspects needing improvement (eg, flexibility, ease of operation, and autonomy level).

  • Model-Free Online Neuroadaptive Controller With Intent Estimation for Physical Human–Robot Interaction

    by Sven Cremer, Sumit Kumar Das, Indika B. Wijayasinghe, Dan O. Popa, Frank L. Lewis

    With the rise of collaborative robots, the need for safe, reliable, and efficient physical human–robot interaction (pHRI) has grown. High-performance pHRI requires robust and stable controllers suitable for multiple degrees of freedom (DoF) and highly nonlinear robots. In this article, we describe a cascade-loop pHRI controller, which relies on human force and pose measurements and can adapt to varying robot dynamics online. It can also adapt to different users and simplifies the interaction by making the robot behave according to a prescribed dynamic model. In our controller formulation, two neural networks (NNs) in the “outer-loop” predict human motion intent and estimate a reference trajectory for the robot that the “inner-loop” controller follows. The inner-loop imposes a prescribed error dynamics (PED) with the help of a model-free neuroadaptive controller (NAC), which uses a NN to feedback linearize the robot dynamics. Lyapunov stability analysis gives weight tuning laws that guarantee that the error signals are bounded and the desired reference trajectory is achieved. Our control scheme was implemented on a Personal Robot 2 robot and validated through an exploratory experimental study in point-to-point collaborative motion. Results indicate fast convergence of our controller, and the resulting tracking error, motion jerk, and human control effort are comparable with other methods that require prior training, knowledge, and calibration.


  • Head-Eye Coordination of Humanoid Robot with Potential Controller

    by Indika B. Wijayasinghe, Sumit K. Das, Haylie L. Miller, Nicoleta L. Bugnariu, Dan O. Popa


  • Human-like object tracking and gaze estimation with PKD android

    by Indika B. Wijayasinghe, Haylie L. Miller, Sumit K. Das, Nicoleta L. Bugnariu, Dan O. Popa

    As the use of robots increases for tasks that require human-robot interactions, it is vital that robots exhibit and understand human-like cues for effective communication. In this paper, we describe the implementation of object tracking capability on Philip K. Dick (PKD) android and a gaze tracking algorithm, both of which further robot capabilities with regard to human communication. PKD’s ability to track objects with human-like head postures is achieved with visual feedback from a Kinect system and an eye camera. The goal of object tracking with human-like gestures is twofold: to facilitate better human-robot interactions and to enable PKD as a human gaze emulator for future studies. The gaze tracking system employs a mobile eye tracking system (ETG; SensoMotoric Instruments) and a motion capture system (Cortex; Motion Analysis Corp.) for tracking the head orientations. Objects to be tracked are displayed by a virtual reality system, the Computer Assisted Rehabilitation Environment (CAREN; MotekForce Link). The gaze tracking algorithm converts eye tracking data and head orientations to gaze information facilitating two objectives: to evaluate the performance of the object tracking system for PKD and to use the gaze information to predict the intentions of the user, enabling the robot to understand physical cues by humans.


  • AGC of a multi-area hydro-thermal system with BES and firefly optimized PID controller

    by Puja Dash, Lalit Chandra Saikia, Sumit Kumar Das

    This paper presents automatic generation control of three unequal area hydro-thermal systems. Appropriate generation rate constraints are considered in each area. Proportional - Integral - Derivative (PID) controller is used as the secondary controller in the areas. A Battery Energy Storage (BES) system is incorporated in Area1 and the system dynamic responses are compared with that of without BES system. Selection of position of BES in interconnected multi area system has been done first time as an individual and observed that BES in Area-1 with 1% SLP is the best. The PID controllers gains are optimized simultaneously using a more recent evolutionary computational technique called firefly algorithm. Investigations reveal that dynamic responses of the system with BES are better than without BES. Sensitivity analysis reveals that the optimum parameters obtained at nominal conditions are quite robust and need not be reset with wide change in system loading and change in step load perturbation (SLP). The system dynamics is evaluated by considering 1% SLP in area1.