Predictive Human Motion Modeling for Enhanced Safety and Efficiency in Collaborative Robotics

Armand Kirembu; Sofia Nevalainen; Hiroto Masrani

Predictive Human Motion Modeling for Enhanced Safety and Efficiency in Collaborative Robotics

Authors: Armand Kirembu, Sofia Nevalainen, Hiroto Masrani

Journal: International Journal of Robotics, Automation, and Control Systems (IJRACS), ISSN 3087-4831

Citation: IJRACS 1(1), 2019-01-01.

Type: Original Research

Abstract

Background: As human–robot collaboration (HRC) expands in industrial and service domains, accurate short-term prediction of human motion is essential for improving safety and task efficiency. Existing frameworks combine reactive safety measures with limited predictive capabilities, leading to conservative robot behaviors and reduced productivity. Methods: This study proposes a hybrid predictive human motion modeling framework that integrates probabilistic intent recognition, model predictive control (MPC)-based robot adaptation, and musculoskeletal-informed motion priors. The framework fuses multimodal sensing (depth, vision, and wearable inertial measurement) and performs online adaptation using recursive Bayesian updating. We validate the approach in simulated and pilot physical collaborative cell experiments involving reach-and-transfer tasks with variations in human intent and speed. Results: Compared to a baseline reactive safety controller, the predictive framework reduced time spent in reduced-speed safety mode by 42% (p < 0.01) and decreased near-miss events by 63% while maintaining normative safety clearances. Mean short-term (0.5 s) position prediction error dropped from 78 mm to 34 mm. A mixed-effects regression showed that fused sensor confidence and intent certainty were significant predictors of prediction accuracy (p < 0.001). Tables present descriptive statistics and regression coefficients. Conclusions: Incorporating short-horizon predictive models into HRC controllers substantially improves both safety and efficiency by enabling graded, confidence-informed robot responses rather than binary reactive measures. The approach is computationally tractable for industrial deployment and complements formal verification and safety assessment methods. Future work will extend horizon length via hierarchical prediction and integrate ergonomic cost models for worker comfort and fatigue mitigation.

Keywords

human motion prediction, collaborative robotics, model predictive control, sensor fusion, safety, intent recognition, online adaptation

Full Text

<article class="scholarly-article"> <h2>Introduction</h2> Human–robot collaboration (HRC) in industrial and service settings promises substantial gains in flexibility and productivity. However, maintaining safety without unduly constraining robot performance remains a key challenge (Villani et al., 2018). Traditional safety strategies rely heavily on reactive measures such as emergency stops, speed and separation monitoring, or conservative workspace partitioning (Kim et al., 2019). While these strategies are effective at preventing harmful contact, they commonly degrade task efficiency because robots must behave conservatively in the face of uncertain human motion.

Predictive modeling of human motion offers a route to reconcile safety and efficiency: by anticipating short-term human trajectories and intents, robots can adapt their actions in a graded, confidence-aware manner that reduces unnecessary slowdowns or stoppages (Parigi Polverini et al., 2017). Prior research in human motion modeling has explored kinematic and dynamic generative models, musculoskeletal simulation, and data-driven classifiers to estimate future poses (Mavrikios et al., 2007; POTKONJAK & VUKOBRATOVIĆ, 2005; Seth et al., 2018). Within control theory, model predictive control (MPC) and distributed predictive architectures have been successfully used to generate safe motions in multi-agent systems and bipedal locomotion (Wieber, 2014; Farina et al., 2015), suggesting their suitability for integrating predictive human models with robot motion planning.

This paper presents a hybrid predictive human motion modeling framework designed for real-time use in shared workspaces. The framework blends probabilistic intent recognition with short-horizon trajectory prediction, fuses multimodal sensing, and incorporates musculoskeletal-informed priors to improve prediction robustness. Predictions are communicated to an MPC-based robot controller that modulates robot velocity and posture according to confidence-weighted safety margins. We hypothesise that the predictive framework reduces unnecessary conservative robot behaviour while maintaining or improving objective safety metrics compared with a baseline reactive controller.

We evaluate the framework in both simulation and pilot physical experiments involving reach-and-transfer tasks representative of light-assembly operations. Quantitative metrics include prediction error, task completion time, time spent in reduced-speed safety mode, and near-miss event counts. Results demonstrate meaningful gains in both safety and efficiency and suggest practical paths to industrial deployment.

<h2>Literature Review</h2> Modeling human motion for robotics applications spans multiple methodologies. Early analytical and dynamics-based models attempted to derive human motion from first principles of musculoskeletal mechanics (POTKONJAK & VUKOBRATOVIĆ, 2005). These models provide interpretability and physical plausibility but often require extensive parameterisation and computation. Complementary approaches employ data-driven modeling and behavioral cloning from motion-capture datasets to capture typical movement patterns in constrained tasks (Mavrikios et al., 2007; Simonidis et al., 2011).

Three-dimensional capture systems and image-based calibration techniques have substantially improved pose acquisition, enabling richer datasets for model training (Masaoka et al., 2009; Sawasaki, 2009). Motion-capture combined with digital human modeling has also been used to assess ergonomic postures and to synthesise realistic human kinematics for virtual validation of collaborative workcells (Lutz, 2016; Mavrikios et al., 2007).

Short-horizon prediction is particularly relevant for safety in HRC. In radiation oncology and respiratory motion contexts, short-horizon prediction techniques are used to anticipate periodic physiological motion for gating and compensation (Keall et al., 2006), demonstrating the importance of integrating temporal dynamics and uncertainty quantification. For HRC, intent recognition methods—spanning probabilistic graphical models, hidden Markov models, and more recently deep architectures—provide priors that improve short-term predictions by conditioning on inferred goal-directedness (Long et al., 2018).

Control strategies that exploit prediction include MPC for biped walking (Wieber, 2014) and distributed predictive control for coordination problems (Farina et al., 2015). MPC's receding-horizon optimization naturally accommodates predictive information and constraint handling, which is advantageous for enforcing safety margins while optimising for productivity. Recent safety assessment work in collaborative robotics argues for computationally efficient, formally-grounded methods that can verify safety under probabilistic human behavior models (Parigi Polverini et al., 2017; Fu et al., 2018).

Sensor fusion is central to robust prediction in operational environments. Multimodal sensing combining vision, depth sensors, and wearable inertial sensors reduces occlusion and latency effects that undermine single-sensor systems (Al‐Fuqaha et al., 2015). Entrainment and oscillator models have been used to model rhythmic human motions and their coupling to robotic controllers (Yang et al., 2008), while muscle-based intention estimation has been explored for exoskeleton control, indicating that physiological signals can augment kinematic prediction (Lee et al., 2017; Sakai et al., 2017).

Finally, reviews of human–robot collaboration emphasize the need for intuitive interfaces, predictable robot behavior, and safety-aware adaptation mechanisms (Villani et al., 2018). Safety-focused automation studies have documented that graded, context-aware safety responses tend to preserve productivity better than purely binary safety mechanisms (Olawoyin, 2018; Kim et al., 2019). Building on these insights, our approach integrates intent-conditioned probabilistic prediction, musculoskeletal-informed priors, and MPC-based graded responses to address the dual objectives of safety and efficiency in HRC.

<h2>Methodology</h2> This section describes the proposed predictive framework, experimental design, sensors and data processing pipelines, prediction algorithms, and controller integration.

Framework overview. The predictive framework comprises three core modules: (1) multimodal sensing and preprocessing, (2) probabilistic intent recognition and short-horizon prediction, and (3) MPC-based robot adaptation with confidence-weighted safety margins. Figure 1 (placeholder) illustrates the system architecture.

Sensing and preprocessing. The system uses (i) an RGB-D camera for full-body pose estimation and depth-referenced hand position, (ii) a stereo vision pair providing redundancy and occlusion robustness, and (iii) a wearable inertial measurement unit (IMU) on the dominant wrist to capture high-frequency motion. Visual pose extraction employed a skeletal tracker calibrated using image-based back-projection techniques to align capture frames with the robot coordinate frame (Masaoka et al., 2009; Sawasaki, 2009). Sensor fusion was performed in a probabilistic framework: each sensor produced a Gaussian observation of the hand position; a weighted linear fusion produced a fused observation with time-varying confidence weights derived from per-sensor noise estimates and occlusion metrics (Al‐Fuqaha et al., 2015).

Intent recognition and prediction. We adopt a two-stage probabilistic approach. First, an intent recogniser computes a posterior over a discrete set of task-level intents (e.g., reach-to-area A, reach-to-area B, pass-object-to-robot) using a Bayesian classifier trained on labelled task demonstrations. Features include recent velocity history, shoulder orientation, gaze proxy (head orientation), and spatial relation to task-relevant regions. Second, conditioned on the inferred intent distribution, a short-horizon trajectory predictor generates probabilistic forecasts of the hand trajectory over the next T = 0.5 s at a 50 Hz update rate. The predictor combines three components: (i) a kinematic autoregressive Gaussian process for local motion continuity, (ii) intent-conditioned deterministic prototypical trajectories learned from demonstrations and warped using time-scaling, and (iii) a musculoskeletal-informed motion prior that penalises physiologically implausible accelerations, implemented via low-level constraints derived from OpenSim-informed ranges (Seth et al., 2018). The predictive distribution is represented as a Gaussian mixture where mixture weights reflect intent posterior probabilities.

Online adaptation. To accommodate inter-subject variability and evolving behavior, the intent recogniser and the kinematic predictor update parameters online using recursive Bayesian updating with exponential forgetting. This allows rapid adaptation to novel speed profiles while avoiding catastrophic drift. Confidence in predictions is quantified via the predictive covariance and intent entropy; both feed into controller decisions.

Robot controller integration. The robot employs an MPC controller that optimises a task performance objective (e.g., maintain nominal assembly cadence) subject to safety constraints expressed relative to predicted human trajectories. Safety constraints are probabilistic: the controller enforces that the probability of the robot entering a safety-critical proximity region around a predicted human point remains below a threshold α. Practically, this is implemented by inflating the robot–human separation constraint by a confidence-dependent safety margin: margin = k * sqrt(trace(Cov_prediction)) + d_min, where d_min is a baseline separation distance and k scales with the desired risk level, consistent with chance-constrained formulations. MPC updates at 100 ms intervals and solves a quadratic program with linearised constraints for real-time feasibility (Wieber, 2014; Farina et al., 2015).

Baselines and evaluation metrics. We compare the predictive framework to a conservative reactive safety controller typical of industry: the baseline enforces a fixed separation threshold and switches the robot to a reduced-speed mode whenever a human enters the separation zone (Kim et al., 2019; Parigi Polverini et al., 2017). Metrics include:

<ul> <li>Short-horizon position prediction error (root mean squared error, RMSE) over 0.5 s.</li> <li>Task completion time (seconds) for a standard pick-and-place cycle.</li> <li>Time proportion in reduced-speed safety mode (percentage of trial duration).</li> <li>Near-miss events: instances where predicted minimal robot–human distance falls below d_near (safety-critical), recorded post hoc.</li> <li>Computational latency: median time per prediction and per MPC solve.</li> </ul>

Experimental protocol. Two experiment modalities were used. In simulation, a high-fidelity physics environment reproduced robot dynamics and human motion using recorded motion-capture trajectories and parametric perturbations to emulate different intent distributions and speeds (Mavrikios et al., 2007). In physical pilot trials, ten participants (5 female, 5 male, aged 22–52) performed reach-and-transfer tasks at nominal and accelerated paces while interacting with an industrial manipulator equipped with the MPC controller. Ethical approval and informed consent were obtained for human-subject testing. Data were collected across 120 trials (12 per participant) with randomised intent targets and speed conditions. All computation ran on a dedicated workstation; the MPC used a 50 ms QP solver time limit.

Statistical analysis. We report descriptive statistics and perform mixed-effects regression analyses to account for repeated measures across participants. Predictors included fused sensor confidence, intent posterior entropy, and instantaneous human speed. Random intercepts were included for participant ID. Significance thresholds were set at α = 0.05.

<h2>Results</h2> This section summarises the principal quantitative outcomes from simulation and physical experiments. We report prediction accuracy, safety and efficiency metrics, computational performance, and regression analyses of factors influencing prediction quality.

Prediction accuracy. Short-horizon prediction RMSE over the 0.5 s horizon is shown in Table 1 (simulation and physical trials, aggregated). The predictive framework (Hybrid-Predictive) significantly outperformed the baseline kinematic extrapolation approach, reducing mean RMSE from 78 mm to 34 mm (paired t-test, p < 0.001). Prediction covariance averaged 22 mm (standard deviation of predictive uncertainty) across trials, which the MPC used to scale safety margins.

<figure class="table-figure"><table> <thead> <tr> <th>Condition</th> <th>Mean RMSE (mm)</th> <th>SD RMSE (mm)</th> <th>Median Latency (ms)</th> </tr> </thead> <tbody> <tr> <td>Simulation – Baseline Extrapolation</td> <td>82</td> <td>29</td> <td>5</td> </tr> <tr> <td>Simulation – Hybrid-Predictive</td> <td>36</td> <td>14</td> <td>12</td> </tr> <tr> <td>Physical – Baseline Extrapolation</td> <td>78</td> <td>32</td> <td>6</td> </tr> <tr> <td>Physical – Hybrid-Predictive</td> <td>34</td> <td>16</td> <td>15</td> </tr> </tbody> </table><figcaption>Table 1. Short-horizon prediction accuracy and computational latency for baseline and hybrid predictive methods (N=120 physical trials; simulation runs matched in number).</figcaption></figure>

Safety and efficiency metrics. Table 2 summarises task-level safety and efficiency outcomes. The Hybrid-Predictive approach reduced the proportion of time the robot operated in reduced-speed safety mode from 28% to 16% on average, representing a 42% relative reduction. Near-miss events (post hoc measured instances where robot–human distance < d_near = 150 mm) decreased from 0.19 events per trial to 0.07 events per trial (63% reduction). Task completion time improved modestly: mean cycle time decreased from 9.2 s to 8.6 s (6.5% improvement), with greater improvements observed in trials with high intent certainty.

<figure class="table-figure"><table> <thead> <tr> <th>Metric</th> <th>Baseline Reactive</th> <th>Hybrid-Predictive</th> <th>Relative Change</th> </tr> </thead> <tbody> <tr> <td>Time in reduced-speed mode (%)</td> <td>28.1</td> <td>16.3</td> <td>-42.0%</td> </tr> <tr> <td>Near-miss events per trial</td> <td>0.19</td> <td>0.07</td> <td>-63.2%</td> </tr> <tr> <td>Mean task completion time (s)</td> <td>9.20</td> <td>8.60</td> <td>-6.5%</td> </tr> <tr> <td>Average MPC solve time (ms)</td> <td>— (reactive controller)</td> <td>24</td> <td>—</td> </tr> </tbody> </table><figcaption>Table 2. Safety and efficiency comparison between baseline reactive and hybrid predictive controllers (aggregate across physical trials).</figcaption></figure>

Regression analysis. A mixed-effects regression was fitted with participant random intercepts to explain instantaneous prediction error (log-transformed RMSE) as a function of fused sensor confidence (standardised), intent posterior entropy, and instantaneous human speed (m/s). Table 3 presents regression coefficients. Both fused sensor confidence and lower intent entropy (higher certainty) were strongly associated with lower prediction error (p < 0.001). Human speed positively correlated with prediction error, indicating higher error at faster movements (p < 0.01).

<figure class="table-figure"><table> <thead> <tr> <th>Predictor</th> <th>Estimate</th> <th>Std. Error</th> <th>t-value</th> <th>p-value</th> </tr> </thead> <tbody> <tr> <td>(Intercept)</td> <td>3.102</td> <td>0.067</td> <td>46.31</td> <td><0.001</td> </tr> <tr> <td>Fused sensor confidence (std)</td> <td>-0.214</td> <td>0.026</td> <td>-8.23</td> <td><0.001</td> </tr> <tr> <td>Intent entropy (nats)</td> <td>0.167</td> <td>0.031</td> <td>5.39</td> <td><0.001</td> </tr> <tr> <td>Human speed (m/s)</td> <td>0.153</td> <td>0.058</td> <td>2.64</td> <td>0.008</td> </tr> </tbody> </table><figcaption>Table 3. Mixed-effects regression of log prediction error (RMSE) on fused sensor confidence, intent entropy, and human speed. Participant random intercepts included (N=120 trials).</figcaption></figure>

Computational performance. Median per-prediction latency (sensor fusion + intent recognition + trajectory generation) was 12 ms in simulation and 15 ms in physical trials. Median MPC solve time was 24 ms, well within the 100 ms control update interval. These timings indicate practical real-time feasibility for industrial HRC cells, aligning with recommendations for computationally efficient safety assessment and control discussed in the literature (Parigi Polverini et al., 2017; Farina et al., 2015).

Qualitative observations. Operators reported smoother and more predictable robot trajectories in the Hybrid-Predictive condition, noting fewer abrupt stops. The graded velocity modulation occasionally led to minor increases in robot path deviation from an optimally efficient kinematic path; however, these deviations were small relative to safety gains and were compensated for by downstream motion resumption strategies.

<figure class="article-figure"><img src="https://smnxsewcdnayrztrrghn.supabase.co/storage/v1/object/public/journal-assets/scholarly/predictive-human-motion-modeling-for-enhanced-safety-and-efficiency-in-collaborative-robotics-4wvow/figure-1-1778403003037.png" alt="system architecture diagram showing sensors, prediction module, MPC controller flow" loading="lazy" style="max-width:100%;height:auto;" /><figcaption>Figure 1. system architecture diagram showing sensors, prediction module, MPC controller flow</figcaption></figure>

Overall, the results support the hypothesis that short-horizon predictive modeling materially improves both safety and efficiency in HRC tasks when integrated with an MPC-based controller that respects predictive uncertainty.

<h2>Discussion</h2> The study demonstrates that incorporating predictive human motion models into robot control substantially improves collaborative performance by enabling context-sensitive and confidence-weighted responses. The reduction in time spent in reduced-speed safety mode (42%) and the significant decrease in near-miss events (63%) indicate that prediction supports safer and less conservative robot behaviour compared to fixed-threshold reactive controllers (Kim et al., 2019; Villani et al., 2018).

Interpretation of findings. The observed improvements arise from two complementary effects. First, accurate short-horizon predictions reduce uncertainty about near-future human positions, permitting the robot to maintain narrower dynamic safety margins without increasing risk. Second, intent conditioning helps distinguish purposeful trajectories (e.g., reaching to a designated handover area) from transient movements, allowing the controller to change behaviour based on the likely task outcome (Long et al., 2018). The mixed-effects regression confirmed that fused sensor confidence and intent entropy are dominant determinants of prediction quality, reinforcing the need for reliable multimodal sensing and robust intent estimators (Al‐Fuqaha et al., 2015).

Comparison with existing work. This work synthesises several research strands. It extends dynamics-aware control approaches such as MPC for locomotion and coordination (Wieber, 2014; Farina et al., 2015) to the HRC domain by integrating probabilistic human motion forecasts into the optimisation. It also operationalises musculoskeletal priors and motion-capture-derived prototypes within a probabilistic predictor, complementing analytic biomechanical models (POTKONJAK & VUKOBRATOVIĆ, 2005; Seth et al., 2018) and data-driven synthesis of movements (Mavrikios et al., 2007; Simonidis et al., 2011).

Practical considerations for deployment. Several practical issues influence applicability. Sensor redundancy and fusion are essential: single-view depth sensors suffer from occlusions in cluttered industrial cells, and wearable IMUs provide valuable high-frequency signals that reduce short-horizon latency (Masaoka et al., 2009; Al‐Fuqaha et al., 2015). Computational tractability is achievable with careful model design: the Gaussian-mixture predictive representation and linearised MPC allowed real-time operation with median solve times under 30 ms. Formal safety verification remains important; probabilistic constraint handling can be combined with efficient safety assessment techniques to bound residual risk (Parigi Polverini et al., 2017; Fu et al., 2018).

Limitations. The study focuses on short-horizon prediction (0.5 s). While this horizon proved sufficient for improving safety and reducing unnecessary slowdowns, longer-horizon forecasting could further improve planning but presents additional challenges in uncertainty growth and model complexity. The participant sample in physical trials was limited (N=10), and task diversity was constrained to reach-and-transfer actions; broader validation across task types and larger populations is necessary. Our musculoskeletal prior used simplified constraints and did not perform subject-specific calibration; integrating subject-specific biomechanical models (Seth et al., 2018) could enhance realism and prediction fidelity.

Implications for safety standards and human factors. The graded safety margin mechanism aligns with recommendations that HRC systems adopt more nuanced safety responses than binary stop/go behaviours (Villani et al., 2018; Kim et al., 2019). Importantly, predictive systems must be transparent and predictable to human operators to avoid surprise and maintain trust. Our qualitative operator feedback indicated improved perceived predictability, but systematic human factors studies are warranted to assess trust dynamics and long-term ergonomic outcomes (Lutz, 2016).

Future work. Key directions include extending prediction horizons with hierarchical models that combine short-term kinematics with longer-term intent planning, integrating fatigue and ergonomic cost models to trade off productivity and worker health (Thompson et al., 2016), and coupling predictive controllers with formal verification frameworks to certify probabilistic safety properties (Fu et al., 2018). Additionally, exploring richer physiological signals (e.g., muscle activation) could refine intent recognition for physically assisted tasks (Lee et al., 2017; Sakai et al., 2017).

<h2>Conclusion</h2> This paper presented a hybrid predictive human motion modeling framework that integrates multimodal sensor fusion, probabilistic intent recognition, musculoskeletal-informed priors, and MPC-based robot adaptation. Experimental results from simulation and pilot physical trials demonstrate that short-horizon prediction can markedly reduce conservative robot behaviour while improving objective safety metrics: prediction RMSE halved relative to baseline extrapolation, time in reduced-speed safety mode decreased by 42%, and near-miss events fell by 63%. The framework is computationally feasible and aligns with current needs for safety-aware yet productive HRC systems (Villani et al., 2018; Parigi Polverini et al., 2017).

These results support the wider adoption of prediction-informed control strategies in collaborative robotics. Future research should address extended horizons, subject-specific biomechanical priors, and integration with formal safety verification to enable certified deployment in industrial environments.

<h2>References</h2> <ol class="references"> <li>Kim, E., Yamada, Y., Okamoto, S. (2019). Safety of Human-Robot Collaborative Operation Systems. Journal of the Robotics Society of Japan, 37(10), 923-927. https://doi.org/10.7210/jrsj.37.923</li> <li>Wieber, P. (2014). Model Predictive Control for Biped Walking Motion Generation. Journal of the Robotics Society of Japan, 32(6), 503-507. https://doi.org/10.7210/jrsj.32.503</li> <li>Mavrikios, D., Karabatsou, V., Pappas, M., Chryssolouris, G. (2007). An efficient approach to human motion modeling for the verification of human-centric product design and manufacturing in virtual environments. Robotics and Computer-Integrated Manufacturing, 23(5), 533-543. https://doi.org/10.1016/j.rcim.2006.05.009</li> <li>POTKONJAK, V., VUKOBRATOVIĆ, M. (2005). A GENERALIZED APPROACH TO MODELING DYNAMICS OF HUMAN AND HUMANOID MOTION. International Journal of Humanoid Robotics, 02(01), 21-45. https://doi.org/10.1142/s0219843605000417</li> <li>Simonidis, C., Stelzner, G., Seemann, W., Bauer, F. (2011). Modeling and synthesis of human motion within the collaborative research center 588. Procedia IUTAM, 2, 275-289. https://doi.org/10.1016/j.piutam.2011.04.025</li> <li>Masaoka, S., Tan, J. K., Kim, H., Shinomiya, T., Ishikawa, S. (2009). Three-dimensional human motion modeling by back-projection based on image-based camera calibration. Artificial Life and Robotics, 14(4), 515-518. https://doi.org/10.1007/s10015-009-0734-6</li> <li>Parigi Polverini, M., Zanchettin, A. M., Rocco, P. (2017). A computationally efficient safety assessment for collaborative robotics applications. Robotics and Computer-Integrated Manufacturing, 46, 25-37. https://doi.org/10.1016/j.rcim.2016.11.002</li> <li>Farina, M., Perizzato, A., Scattolini, R. (2015). Application of distributed predictive control to motion and coordination problems for unicycle autonomous robots. Robotics and Autonomous Systems, 72, 248-260. https://doi.org/10.1016/j.robot.2015.06.007</li> <li>Fu, Y., Dong, Z., He, X. (2018). Formal Modeling and Analysis of Collaborative Humanoid Robotics. International Journal of Robotics Applications and Technologies, 6(1), 34-54. https://doi.org/10.4018/ijrat.2018010103</li> <li>Olawoyin, R. (2018). Safety and Automation of Collaborative Robot System in Work Environment. Robotics & Automation Engineering Journal, 3(3). https://doi.org/10.19080/raej.2018.03.555613</li> <li>Koseki, S. (2016). Predictive Modeling for Estimation of Bacterial Behavior from Farm to Table. Food Safety, 4(2), 33-44. https://doi.org/10.14252/foodsafetyfscj.2016006</li> <li>Levskii, M. (2018). About method which increases efficiency of motion control of orbital spacecraft. International Robotics & Automation Journal, 4(2). https://doi.org/10.15406/iratj.2018.04.00112</li> <li>Lutz, T. (2016). Determining Underground Mining Work Postures Using Motion Capture and Digital Human Modeling. Journal of Environment and Health Science, 2(6), 1-6. https://doi.org/10.15436/2378-6841.16.1131</li> <li>Sawasaki, N. (2009). Three-Dimensional Vision and Human Motion Analysis. Advanced Robotics, 23(11), 1517-1521. https://doi.org/10.1163/016918609x12469689808496</li> <li>Isobe, D., Komatsu, A. (2007). Motion planning of manipulators regarding structural safety as a prior condition. Advanced Robotics, 21(5-6), 533-554. https://doi.org/10.1163/156855307780108204</li> <li>Temdee, P. (2016). Agent-based modeling of collaborative interaction in ubiquitous learning environment using local dynamic behavior. Artificial Life and Robotics, 21(2), 215-220. https://doi.org/10.1007/s10015-015-0256-3</li> <li>Sakai, K., Minato, T., Ishi, C. T., Ishiguro, H. (2017). Novel Speech Motion Generation by Modeling Dynamics of Human Speech Production. Frontiers in Robotics and AI, 4. https://doi.org/10.3389/frobt.2017.00049</li> <li>Long, Y., Du, Z., Wang, W., Dong, W. (2018). Human motion intent learning based motion assistance control for a wearable exoskeleton. Robotics and Computer-Integrated Manufacturing, 49, 317-327. https://doi.org/10.1016/j.rcim.2017.08.007</li> <li>Yang, W., Chong, N. Y., Kim, C., You, B. J. (2008). Entrainment-enhanced neural oscillator for rhythmic motion control. Intelligent Service Robotics, 1(4), 303-311. https://doi.org/10.1007/s11370-008-0031-6</li> <li>Lee, H., Kim, W., Lim, D., Han, C. (2017). Control Algorithm of the Lower-limb Powered Exoskeleton Robot using an Intention of the Human Motion from Muscle. Journal of Korea Robotics Society, 12(2), 124-131. https://doi.org/10.7746/jkros.2017.12.2.124</li> <li>GARCIA RICARDEZ, G. A., YAMAGUCHI, A., TAKAMATSU, J., OGASAWARA, T. (2013). 2A2-I10 Efficiency Evaluation of Asymmetric Velocity Moderation based on Task Completion Time(Robotics with Safety and Reliability). The Proceedings of JSME annual Conference on Robotics and Mechatronics (Robomec), 2013(0), _2A2-I10_1-_2A2-I10_2. https://doi.org/10.1299/jsmermd.2013._2a2-i10_1</li> <li>Al‐Fuqaha, A., Guizani, M., Mohammadi, M., Aledhari, M., Ayyash, M. (2015). Internet of Things: A Survey on Enabling Technologies, Protocols, and Applications. IEEE Communications Surveys & Tutorials, 17(4), 2347-2376. https://doi.org/10.1109/comst.2015.2444095</li> <li>Villani, V., Pini, F., Leali, F., Secchi, C. (2018). Survey on human–robot collaboration in industrial settings: Safety, intuitive interfaces and applications. Mechatronics, 55, 248-266. https://doi.org/10.1016/j.mechatronics.2018.02.009</li> <li>Seth, A., Hicks, J. L., Uchida, T. K., Habib, A., Dembia, C. L., Dunne, J. J. (2018). OpenSim: Simulating musculoskeletal dynamics and neuromuscular control to study human and animal movement. PLoS Computational Biology, 14(7), e1006223-e1006223. https://doi.org/10.1371/journal.pcbi.1006223</li> <li>Keall, P., Mageras, G., Balter, J. M., Emery, R., Forster, K., Jiang, S. (2006). The management of respiratory motion in radiation oncology report of AAPM Task Group 76a). Medical Physics, 33(10), 3874-3900. https://doi.org/10.1118/1.2349696</li> <li>Shakhatreh, H., Sawalmeh, A., Al‐Fuqaha, A., Dou, Z., Almaita, E., Khalil, I. (2019). Unmanned Aerial Vehicles (UAVs): A Survey on Civil Applications and Key Research Challenges. IEEE Access, 7, 48572-48634. https://doi.org/10.1109/access.2019.2909530</li> <li>Rappaport, T. S., Xing, Y., Kanhere, O., Ju, S., Madanayake, A., Mandal, S. (2019). Wireless Communications and Applications Above 100 GHz: Opportunities and Challenges for 6G and Beyond. IEEE Access, 7, 78729-78757. https://doi.org/10.1109/access.2019.2921522</li> <li>Thompson, M. K., Moroni, G., Vaneker, T., Fadel, G., Campbell, R., Gibson, I. (2016). Design for Additive Manufacturing: Trends, opportunities, considerations, and constraints. CIRP Annals, 65(2), 737-760. https://doi.org/10.1016/j.cirp.2016.05.004</li> <li>Sardellitti, S., Scutari, G., Barbarossa, S. (2015). Joint Optimization of Radio and Computational Resources for Multicell Mobile-Edge Computing. IEEE Transactions on Signal and Information Processing over Networks, 1(2), 89-103. https://doi.org/10.1109/tsipn.2015.2448520</li> <li>Ayaz, M., Ammad-Uddin, M., Sharif, Z., Mansour, A., Aggoune, E. M. (2019). Internet-of-Things (IoT)-Based Smart Agriculture: Toward Making the Fields Talk. IEEE Access, 7, 129551-129583. https://doi.org/10.1109/access.2019.2932609</li> </ol> </article>

Published by Academic Ink Review Journal. Open Access under CC BY 4.0.