Research on human motion prediction has made significant progress due to its importance in the development of various artificial intelligence applications.However,effectively capturing spatio-temporal features for smo...Research on human motion prediction has made significant progress due to its importance in the development of various artificial intelligence applications.However,effectively capturing spatio-temporal features for smoother and more precise human motion prediction remains a challenge.To address these issues,a robust human motion prediction method via integration of spatial and temporal cues(RISTC)has been proposed.This method captures sufficient spatio-temporal correlation of the observable sequence of human poses by utilizing the spatio-temporal mixed feature extractor(MFE).In multi-layer MFEs,the channel-graph united attention blocks extract the augmented spatial features of the human poses in the channel and spatial dimension.Additionally,multi-scale temporal blocks have been designed to effectively capture complicated and highly dynamic temporal information.Our experiments on the Human3.6M and Carnegie Mellon University motion capture(CMU Mocap)datasets show that the proposed network yields higher prediction accuracy than the state-of-the-art methods.展开更多
Predicting human motion based on historical motion sequences is a fundamental problem in computer vision,which is at the core of many applications.Existing approaches primarily focus on encoding spatial dependencies a...Predicting human motion based on historical motion sequences is a fundamental problem in computer vision,which is at the core of many applications.Existing approaches primarily focus on encoding spatial dependencies among human joints while ignoring the temporal cues and the complex relationships across non-consecutive frames.These limitations hinder the model’s ability to generate accurate predictions over longer time horizons and in scenarios with complex motion patterns.To address the above problems,we proposed a novel multi-level spatial and temporal learning model,which consists of a Cross Spatial Dependencies Encoding Module(CSM)and a Dynamic Temporal Connection Encoding Module(DTM).Specifically,the CSM is designed to capture complementary local and global spatial dependent information at both the joint level and the joint pair level.We further present DTM to encode diverse temporal evolution contexts and compress motion features to a deep level,enabling the model to capture both short-term and long-term dependencies efficiently.Extensive experiments conducted on the Human 3.6M and CMU Mocap datasets demonstrate that our model achieves state-of-the-art performance in both short-term and long-term predictions,outperforming existing methods by up to 20.3% in accuracy.Furthermore,ablation studies confirm the significant contributions of the CSM and DTM in enhancing prediction accuracy.展开更多
Intersections are quite important and complex traffic scenarios,where the future motion of surrounding vehicles is an indispensable reference factor for the decision-making or path planning of autonomous vehicles.Cons...Intersections are quite important and complex traffic scenarios,where the future motion of surrounding vehicles is an indispensable reference factor for the decision-making or path planning of autonomous vehicles.Considering that the motion trajectory of a vehicle at an intersection partly obeys the statistical law of historical data once its driving intention is determined,this paper proposes a long short-term memory based(LSTM-based)framework that combines intention prediction and trajectory prediction together.First,we build an intersection prior trajectories model(IPTM)by clustering and statistically analyzing a large number of prior traffic flow trajectories.The prior trajectories model with fitted probabilistic density is used to approximate the distribution of the predicted trajectory,and also serves as a reference for credibility evaluation.Second,we conduct the intention prediction through another LSTM model and regard it as a crucial cue for a trajectory forecast at the early stage.Furthermore,the predicted intention is also a key that is associated with the prior trajectories model.The proposed framework is validated on two publically released datasets,next generation simulation(NGSIM)and INTERACTION.Compared with other prediction methods,our framework is able to sample a trajectory from the estimated distribution,with its accuracy improved by about 20%.Finally,the credibility evaluation,which is based on the prior trajectories model,makes the framework more practical in the real-world applications.展开更多
After the first Earth Orientation Parameters Prediction Comparison Campaign(1 st EOP PCC),the traditional method using least-squares extrapolation and autoregressive(LS+AR)models was considered as one of the polar mot...After the first Earth Orientation Parameters Prediction Comparison Campaign(1 st EOP PCC),the traditional method using least-squares extrapolation and autoregressive(LS+AR)models was considered as one of the polar motion prediction methods with higher accuracy.The traditional method predicts individual polar motion series separately,which has a single input data and limited improvement in prediction accuracy.To address this problem,this paper proposes a new method for predicting polar motion by combining the difference between polar motion series.The X,Y,and Y-X series were predicted separately using LS+AR models.Then,the new forecast value of X series is obtained by combining the forecast value of Y series with that of Y-X series;the new forecast value of Y series is obtained by combining the forecast value of X series with that of Y-X series.The hindcast experimental comparison results from January 1,2011 to April 4,2021 show that the new method achieves a maximum improvement of 12.95%and 14.96%over the traditional method in the X and Y directions,respectively.The new method has obvious advantages compared with the differential method.This study tests the stability and superiority of the new method and provides a new idea for the research of polar motion prediction.展开更多
A new method improves prediction of the motion of a hybrid monohull in regular waves. Stem section hydrodynamic coefficients of a hybrid monohull with harmonic oscillation were computed using the Reynolds Averaged Nav...A new method improves prediction of the motion of a hybrid monohull in regular waves. Stem section hydrodynamic coefficients of a hybrid monohull with harmonic oscillation were computed using the Reynolds Averaged Navier-Stokes Equations （RANSE）. The governing equations were solved using the finite volume method. The VOF method was used for free surface treatment, and RNGK-ε turbulence model was employed in viscous flow calculation. The whole computational domain was divided into many blocks each with structured grids, and the dynamic process was treated with moving grids. Using a 2-D strip method and 2.5D theory with the correction hydrodynamic coefficients allows consideration of the viscous effect when predicting longitudinal motion of a hybrid monohull in regular waves. The method is effective at predicting motion of a hybrid monohull, showing that the viscous effect on a semi-submerged body cannot be ignored.展开更多
High-precision polar motion(PM) prediction is of important significance in astronomy, geodesy, aviation,hydrographic mapping, interstellar navigation, and so on. SSA can effectively extract the trend and period terms ...High-precision polar motion(PM) prediction is of important significance in astronomy, geodesy, aviation,hydrographic mapping, interstellar navigation, and so on. SSA can effectively extract the trend and period terms of PM,in the process of achieving high-precision medium-and long-term polar motion prediction,it is necessary to solve the end effect problem and overfitting problem of SSA forecasting method;therefore, ARMA was applied to decreasethe end effect, and a suitable combination of reconstructed components was determined to avoid the high variance reaction of SSA overfitting. Based on the decomposition and reconstruction of the PM by SSA, the reconstructed components are determined to participate in the SSA iterative fitting model according to the variance contribution rate. The combination of the reconstructed components representing the polar motion period term and the trend term is determined according to the correlation analysis of the selected reconstructed components. After the above work, the principal component prediction sequence is obtained by fitting the period term and the trend term to convergence, respectively, and then, the SSA end effect is modified, and the residual term is predicted based on ARMA. The test results show that he prediction accuracy of SSA + ARMA at the front of the X and Y directions are improved by 96.90% and 97.53% compared with those of SSA, respectively,and the forecast accuracy of 365 days are improved by 37.93% and 19.53% in the X and Y directions compared with those of Bulletin A, respectively.展开更多
A lightweight multi-layer residual temporal convolutional network model(RTCN)is proposed to target the highly complex kinematics and temporal correlation of human motion.RTCN uses 1-D convolution to efficiently obtain...A lightweight multi-layer residual temporal convolutional network model(RTCN)is proposed to target the highly complex kinematics and temporal correlation of human motion.RTCN uses 1-D convolution to efficiently obtain the spatial structure information of human motion and extract the correlation in the time series of human motion.The residual structure is applied to the proposed network model to alleviate the problem of gradient disappearance in the deep network.Experiments on the Human 3.6M dataset demonstrate that the proposed method effectively reduces the errors of motion prediction compared with previous methods,especially of long-term prediction.展开更多
In recent years,human motion prediction has become an active research topic in computer vision.However,owing to the complexity and stochastic nature of human motion,it remains a challenging problem.In previous works,h...In recent years,human motion prediction has become an active research topic in computer vision.However,owing to the complexity and stochastic nature of human motion,it remains a challenging problem.In previous works,human motion prediction has always been treated as a typical inter-sequence problem,and most works have aimed to capture the temporal dependence between successive frames.However,although these approaches focused on the effects of the temporal dimension,they rarely considered the correlation between different joints in space.Thus,the spatio-temporal coupling of human joints is considered,to propose a novel spatio-temporal network based on a transformer and a gragh convolutional network(GCN)(STTG-Net).The temporal transformer is used to capture the global temporal dependencies,and the spatial GCN module is used to establish local spatial correlations between the joints for each frame.To overcome the problems of error accumulation and discontinuity in the motion prediction,a revision method based on fusion strategy is also proposed,in which the current prediction frame is fused with the previous frame.The experimental results show that the proposed prediction method has less prediction error and the prediction motion is smoother than previous prediction methods.The effectiveness of the proposed method is also demonstrated comparing it with the state-of-the-art method on the Human3.6 M dataset.展开更多
Predicting the motion of other road agents enables autonomous vehicles to perform safe and efficient path planning.This task is very complex,as the behaviour of road agents depends on many factors and the number of po...Predicting the motion of other road agents enables autonomous vehicles to perform safe and efficient path planning.This task is very complex,as the behaviour of road agents depends on many factors and the number of possible future trajectories can be consid-erable(multi-modal).Most prior approaches proposed to address multi-modal motion prediction are based on complex machine learning systems that have limited interpret-ability.Moreover,the metrics used in current benchmarks do not evaluate all aspects of the problem,such as the diversity and admissibility of the output.The authors aim to advance towards the design of trustworthy motion prediction systems,based on some of the re-quirements for the design of Trustworthy Artificial Intelligence.The focus is on evaluation criteria,robustness,and interpretability of outputs.First,the evaluation metrics are comprehensively analysed,the main gaps of current benchmarks are identified,and a new holistic evaluation framework is proposed.Then,a method for the assessment of spatial and temporal robustness is introduced by simulating noise in the perception system.To enhance the interpretability of the outputs and generate more balanced results in the proposed evaluation framework,an intent prediction layer that can be attached to multi-modal motion prediction models is proposed.The effectiveness of this approach is assessed through a survey that explores different elements in the visualisation of the multi-modal trajectories and intentions.The proposed approach and findings make a significant contribution to the development of trustworthy motion prediction systems for autono-mous vehicles,advancing the field towards greater safety and reliability.展开更多
During ship operations,frequent heave movements can pose significant challenges to the overall safety of the ship and completion of cargo loading.The existing heave compensation systems suffer from issues such as dead...During ship operations,frequent heave movements can pose significant challenges to the overall safety of the ship and completion of cargo loading.The existing heave compensation systems suffer from issues such as dead zones and control system time lags,which necessitate the development of reasonable prediction models for ship heave movements.In this paper,a novel model based on a time graph convolutional neural network algorithm and particle swarm optimization algorithm(PSO-TGCN)is proposed for the first time to predict the multipoint heave movements of ships under different sea conditions.To enhance the dataset's suitability for training and reduce interference,various filter algorithms are employed to optimize the dataset.The training process utilizes simulated heave data under different sea conditions and measured heave data from multiple points.The results show that the PSO-TGCN model predicts the ship swaying motion in different sea states after 2 s with 84.7%accuracy,while predicting the swaying motion in three different positions.By performing a comparative study,it was also found that the present method achieves better performance that other popular methods.This model can provide technical support for intelligent ship control,improve the control accuracy of intelligent ships,and promote the development of intelligent ships.展开更多
Recently,human motion prediction has gained significant attention and achieved notable success.However,current methods primarily rely on training and testing with ideal datasets,overlooking the impact of variations in...Recently,human motion prediction has gained significant attention and achieved notable success.However,current methods primarily rely on training and testing with ideal datasets,overlooking the impact of variations in the viewing distance and viewing angle,which are commonly encountered in practical scenarios.In this study,we address the issue of model invariance by ensuring robust performance despite variations in view distances and angles.To achieve this,we employed Riemannian geometry methods to constrain the learning process of neural networks,enabling the prediction of invariances using a simple network.Furthermore,this enhances the application of motion prediction in various scenarios.Our framework uses Riemannian geometry to encode motion into a novel motion space to achieve prediction with an invariant viewing distance and angle using a simple network.Specifically,the specified path transport square-root velocity function is proposed to aid in removing the view-angle equivalence class and encode motion sequences into a flattened space.Motion coding by the geometry method linearizes the optimization problem in a non-flattened space and effectively extracts motion information,allowing the proposed method to achieve competitive performance using a simple network.Experimental results on Human 3.6M and CMU MoCap demonstrate that the proposed framework has competitive performance and invariance to the viewing distance and viewing angle.展开更多
Accurate motion prediction of free-tumbling satellites is crucial for the success of capture operations.This paper proposes a two-step method to estimate the motion states and parameters of such satellites,thereby ena...Accurate motion prediction of free-tumbling satellites is crucial for the success of capture operations.This paper proposes a two-step method to estimate the motion states and parameters of such satellites,thereby enabling precise long-term motion prediction.This paper begins with a measurement of the system's degree of observability,quantified through the Empirical Observability Gramian(EOG).Based on this measurement,a batch processing algorithm is first employed to estimate the satellite's constant parameters offline.Subsequently,an online filtering algorithm,utilizing a minimal state set,fine-tunes these parameters and estimates the motion states in real time.This integrated approach significantly enhances both convergence properties and estimation accuracy,particularly for systems with poor observability.Utilizing the predicted long-term motion of the satellite,a composite evaluation metric is formulated to identify the optimal capture point and moment.The base pose of the space robot is then adjusted to ensure that the optimal capture point lies within the manipulator's dexterous workspace,which is determined through a pre-constructed capability map.The effectiveness of the proposed method is demonstrated through both simulation and experimental results.展开更多
The traditional approach to probabilistic seismic hazard analysis(PSHA)relies on ground motion records,which restricts its application in regions with sparse seismic records or low seismicity.Recently,the 3D physics-b...The traditional approach to probabilistic seismic hazard analysis(PSHA)relies on ground motion records,which restricts its application in regions with sparse seismic records or low seismicity.Recently,the 3D physics-based simulation(PBS)has been recognized as a more effective tool,which offers the flexibility to generate time histories of simulated ground motions.The PBS methods are essential for obtaining ground motion parameters and compensating for lack of records.In this study,building on the theoretical framework of the China Probabilistic Seismic Hazard Analysis(CPSHA)method,we integrated the hierarchical potential focal region model from the fifth-generation seismic ground motion parameters zonation map of China and an appropriate seismicity model reflecting spatial distribution characteristics.Ground motion parameters at the target scale were calculated using PBS for near-field seismic simulations and ground motion prediction equations(GMPEs)for far-field seismic predictions,accounting for the uncertainties in ground motion attenuation from both methods to compute the seismic hazard of each site.In this manner,we established a comprehensive regional probabilistic seismic hazard analysis method combining PBS and GMPE.Using Tianjin as a case study,a probabilistic seismic hazard analysis was conducted with this method,providing seismic hazard curves for specific sites within each administrative region and zoning maps,which were then compared with the results of the fifth-generation zonation maps.The results indicate that the calculated seismic hazard values are generally consistent with the fifth-generation map at the lower limit,while the upper limit is slightly higher due to the near-fault effect.展开更多
In the face of the large number of people with motor function disabilities,rehabilitation robots have attracted more and more attention.In order to promote the active participation of the user's motion intention i...In the face of the large number of people with motor function disabilities,rehabilitation robots have attracted more and more attention.In order to promote the active participation of the user's motion intention in the assisted rehabilitation process of the robots,it is crucial to establish the human motion prediction model.In this paper,a hybrid prediction model built on long short-term memory(LSTM)neural network using surface electromyography(sEMG)is applied to predict the elbow motion of the users in advance.This model includes two sub-models:a back-propagation neural network and an LSTM network.The former extracts a preliminary prediction of the elbow motion,and the latter corrects this prediction to increase accuracy.The proposed model takes time series data as input,which includes the sEMG signals measured by electrodes and the continuous angles from inertial measurement units.The offline and online tests were carried out to verify the established hybrid model.Finally,average root mean square errors of 3.52°and 4.18°were reached respectively for offline and online tests,and the correlation coefficients for both were above 0.98.展开更多
Ship motion attitude is influenced by dynamic marine conditions,presenting significant challenges in developing effective prediction networks.Contemporary prediction networks demonstrate limitations in hidden feature ...Ship motion attitude is influenced by dynamic marine conditions,presenting significant challenges in developing effective prediction networks.Contemporary prediction networks demonstrate limitations in hidden feature extraction,long-term dependency maintenance,and frequency characteristic incorporation.This paper presents an enhanced model integrating the informer network with a Time Convolutional Network(TCN)and a Frequency-Enhanced Channel Attention Mechanism(FECAM).The model employs a TCN for multi-feature extraction and applies Dimension-Segment-Wise(DSW)embedding for comprehensive multi-dimensional sequence analysis.Furthermore,it incorporates discrete cosine transform within the FECAM module for thorough data frequency analysis.The model integrates these components with the informer model for multivariate prediction.This approach maintains the informer model's capabilities in long-term multivariate prediction while enhancing feature extraction and local frequency information capture from ship motion attitude data,thus improving long-term multivariate prediction accuracy.Experimental results indicate that the proposed model outperforms traditional ship motion attitude prediction methods in forecasting future motion,reducing attitude prediction errors,and improving prediction accuracy.展开更多
Respiratory motion-induced vertebral movements can adversely impact intraoperative spine surgery,resulting in inaccurate positional information of the target region and unexpected damage during the operation.In this p...Respiratory motion-induced vertebral movements can adversely impact intraoperative spine surgery,resulting in inaccurate positional information of the target region and unexpected damage during the operation.In this paper,we propose a novel deep learning architecture for respiratory motion prediction,which can adapt to different patients.The proposed method utilizes an LSTM-AE with attention mechanism network that can be trained using few-shot datasets during operation.To ensure real-time performance,a dimension reduction method based on the respiration-induced physical movement of spine vertebral bodies is introduced.The experiment collected data from prone-positioned patients under general anaesthesia to validate the prediction accuracy and time efficiency of the LSTM-AE-based motion prediction method.The experimental results demonstrate that the presented method(RMSE:4.39%)outperforms other methods in terms of accuracy within a learning time of 2 min.The maximum predictive errors under the latency of 333 ms with respect to the x,y,and z axes of the optical camera system were 0.13,0.07,and 0.10 mm,respectively,within a motion range of 2 mm.展开更多
Accurate prediction of shipmotion is very important for ensuringmarine safety,weapon control,and aircraft carrier landing,etc.Ship motion is a complex time-varying nonlinear process which is affected by many factors.T...Accurate prediction of shipmotion is very important for ensuringmarine safety,weapon control,and aircraft carrier landing,etc.Ship motion is a complex time-varying nonlinear process which is affected by many factors.Time series analysis method and many machine learning methods such as neural networks,support vector machines regression(SVR)have been widely used in ship motion predictions.However,these single models have certain limitations,so this paper adopts amulti-model prediction method.First,ensemble empirical mode decomposition(EEMD)is used to remove noise in ship motion data.Then the randomforest(RF)prediction model optimized by genetic algorithm(GA),back propagation neural network(BPNN)prediction model and SVR prediction model are respectively established,and the final prediction results are obtained by results of three models.And the weights coefficients are determined by the correlation coefficients,reducing the risk of prediction and improving the reliability.The experimental results show that the proposed combined model EEMD-GARF-BPNN-SVR is superior to the single predictive model and more reliable.The mean absolute percentage error(MAPE)of the proposed model is 0.84%,but the results of the single models are greater than 1%.展开更多
Human motion prediction is a critical issue in human-robot collaboration(HRC)tasks.In order to reduce the local error caused by the limitation of the capture range and sampling frequency of the depth sensor,a hybrid h...Human motion prediction is a critical issue in human-robot collaboration(HRC)tasks.In order to reduce the local error caused by the limitation of the capture range and sampling frequency of the depth sensor,a hybrid human motion prediction algorithm,optimized sliding window polynomial fitting and recursive least squares(OSWPF-RLS)was proposed.The OSWPF-RLS algorithm uses the human body joint data obtained under the HRC task as input,and uses recursive least squares(RLS)to predict the human movement trajectories within the time window.Then,the optimized sliding window polynomial fitting(OSWPF)is used to calculate the multi-step prediction value,and the increment of multi-step prediction value was appropriately constrained.Experimental results show that compared with the existing benchmark algorithms,the OSWPF-RLS algorithm improved the multi-step prediction accuracy of human motion and enhanced the ability to respond to different human movements.展开更多
Ensuring accurate parameter identification and diving motion prediction of marine crafts is essential for safe navigation,optimized operational efficiency,and the advancement of marine exploration.Addressing this,this...Ensuring accurate parameter identification and diving motion prediction of marine crafts is essential for safe navigation,optimized operational efficiency,and the advancement of marine exploration.Addressing this,this paper proposes an instrumental variable-based least squares(IVLS)algorithm.Firstly,aiming to balance complexity with accuracy,a decoupled diving model is constructed,incorporating nonlinear actuator characteristics,inertia coefficients,and damping coefficients.Secondly,a discrete parameter identification matrix is designed based on this dedicated model,and then a IVLS algorithm is innovatively derived to reject measurement noise.Furthermore,the stability of the proposed algorithm is validated from a probabilistic point of view,providing a solid theoretical foundation.Finally,performance evaluation is conducted using four depth control datasets obtained from a piston-driven profiling float in Qiandao Lake,with desired depths of 30 m,40 m,50 m,and 60 m.Based on the diving dynamics identification results,the IVLS algorithm consistently shows superior performance when compared to recursive weighted least squares algorithm and least squares support vector machine algorithm across all depths,as evidenced by lower average absolute error(AVGAE),root mean square error(RMSE),and maximum absolute error values and higher determination coefficient(R2).Specifically,for desired depth of 60 m,the IVLS algorithm achieved an AVGAE of 0.553 m and RMSE of 0.655 m,significantly outperforming LSSVM with AVGAE and RMSE values of 8.782 m and 11.117 m,respectively.Moreover,the IVLS algorithm demonstrates a remarkable generalization capability with R2 values consistently above 0.95,indicating its robustness in handling varied diving dynamics.展开更多
Previous studies revealed that the error of pole coordinate prediction will significantly increase for a prediction period longer than 100 days, and this is mainly caused by short period oscillations. Empirical mode d...Previous studies revealed that the error of pole coordinate prediction will significantly increase for a prediction period longer than 100 days, and this is mainly caused by short period oscillations. Empirical mode decomposition （EMD）, which is increasingly popular and has advantages over classical wavelet decomposition, can be used to remove short period variations from observed time series of pole co- ordinates. A hybrid model combing EMD and extreme learning machine （ELM）, where high frequency signals are removed and processed time series is then modeled and predicted, is summarized in this paper. The prediction performance of the hybrid model is compared with that of the ELM-only method created from original time series. The results show that the proposed hybrid model outperforms the pure ELM method for both short-term and long-term prediction of pole coordinates. The improvement of prediction accuracy up to 360 days in the future is found to be 24.91% and 26.79% on average in terms of mean absolute error （MAE） for the xp and yp components of pole coordinates, respectively.展开更多
基金supported by the National Key R&D Program of China(No.2018YFB1305200)the Natural Science Foundation of Zhejiang Province(No.LGG21F030011)。
文摘Research on human motion prediction has made significant progress due to its importance in the development of various artificial intelligence applications.However,effectively capturing spatio-temporal features for smoother and more precise human motion prediction remains a challenge.To address these issues,a robust human motion prediction method via integration of spatial and temporal cues(RISTC)has been proposed.This method captures sufficient spatio-temporal correlation of the observable sequence of human poses by utilizing the spatio-temporal mixed feature extractor(MFE).In multi-layer MFEs,the channel-graph united attention blocks extract the augmented spatial features of the human poses in the channel and spatial dimension.Additionally,multi-scale temporal blocks have been designed to effectively capture complicated and highly dynamic temporal information.Our experiments on the Human3.6M and Carnegie Mellon University motion capture(CMU Mocap)datasets show that the proposed network yields higher prediction accuracy than the state-of-the-art methods.
基金supported by the Urgent Need for Overseas Talent Project of Jiangxi Province(Grant No.20223BCJ25040)the Thousand Talents Plan of Jiangxi Province(Grant No.jxsg2023101085)+3 种基金the National Natural Science Foundation of China(Grant No.62106093)the Natural Science Foundation of Jiangxi(Grant Nos.20224BAB212011,20232BAB212008,20242BAB25078,and 20232BAB202051)The Youth Talent Cultivation Innovation Fund Project of Nanchang University(Grant No.XX202506030015)funded by Princess Nourah bint Abdulrahman University Researchers Supporting Project number(PNURSP2025R759),Princess Nourah bint Abdulrahman University,Riyadh,Saudi Arabia.
文摘Predicting human motion based on historical motion sequences is a fundamental problem in computer vision,which is at the core of many applications.Existing approaches primarily focus on encoding spatial dependencies among human joints while ignoring the temporal cues and the complex relationships across non-consecutive frames.These limitations hinder the model’s ability to generate accurate predictions over longer time horizons and in scenarios with complex motion patterns.To address the above problems,we proposed a novel multi-level spatial and temporal learning model,which consists of a Cross Spatial Dependencies Encoding Module(CSM)and a Dynamic Temporal Connection Encoding Module(DTM).Specifically,the CSM is designed to capture complementary local and global spatial dependent information at both the joint level and the joint pair level.We further present DTM to encode diverse temporal evolution contexts and compress motion features to a deep level,enabling the model to capture both short-term and long-term dependencies efficiently.Extensive experiments conducted on the Human 3.6M and CMU Mocap datasets demonstrate that our model achieves state-of-the-art performance in both short-term and long-term predictions,outperforming existing methods by up to 20.3% in accuracy.Furthermore,ablation studies confirm the significant contributions of the CSM and DTM in enhancing prediction accuracy.
基金partly supported by the National Natural Science Foundation of China(61903034,U1913203,61973034,91120003)the Program for Changjiang Scholars and Innovative Research Team in University(IRT-16R06,T2014224)+1 种基金China Postdoctoral Science Foundation funded project(2019TQ0035)Beijing Institute of Technology Research Fund Program for Young Scholars。
文摘Intersections are quite important and complex traffic scenarios,where the future motion of surrounding vehicles is an indispensable reference factor for the decision-making or path planning of autonomous vehicles.Considering that the motion trajectory of a vehicle at an intersection partly obeys the statistical law of historical data once its driving intention is determined,this paper proposes a long short-term memory based(LSTM-based)framework that combines intention prediction and trajectory prediction together.First,we build an intersection prior trajectories model(IPTM)by clustering and statistically analyzing a large number of prior traffic flow trajectories.The prior trajectories model with fitted probabilistic density is used to approximate the distribution of the predicted trajectory,and also serves as a reference for credibility evaluation.Second,we conduct the intention prediction through another LSTM model and regard it as a crucial cue for a trajectory forecast at the early stage.Furthermore,the predicted intention is also a key that is associated with the prior trajectories model.The proposed framework is validated on two publically released datasets,next generation simulation(NGSIM)and INTERACTION.Compared with other prediction methods,our framework is able to sample a trajectory from the estimated distribution,with its accuracy improved by about 20%.Finally,the credibility evaluation,which is based on the prior trajectories model,makes the framework more practical in the real-world applications.
基金funded by the National Natural Science Foundation of China(Nos.42174011 and 41874001)Jiangxi Province Graduate Student Innovation Fund(No.YC2021-S614)+2 种基金Jiangxi Provincial Natural Science Foundation(No.20202BABL212015)the East China University of Technology Ph.D.Project(No.DNBK2019181)the Key Laboratory for Digital Land and Resources of Jiangxi Province,East China University of Technology(No.DLLJ202109)
文摘After the first Earth Orientation Parameters Prediction Comparison Campaign(1 st EOP PCC),the traditional method using least-squares extrapolation and autoregressive(LS+AR)models was considered as one of the polar motion prediction methods with higher accuracy.The traditional method predicts individual polar motion series separately,which has a single input data and limited improvement in prediction accuracy.To address this problem,this paper proposes a new method for predicting polar motion by combining the difference between polar motion series.The X,Y,and Y-X series were predicted separately using LS+AR models.Then,the new forecast value of X series is obtained by combining the forecast value of Y series with that of Y-X series;the new forecast value of Y series is obtained by combining the forecast value of X series with that of Y-X series.The hindcast experimental comparison results from January 1,2011 to April 4,2021 show that the new method achieves a maximum improvement of 12.95%and 14.96%over the traditional method in the X and Y directions,respectively.The new method has obvious advantages compared with the differential method.This study tests the stability and superiority of the new method and provides a new idea for the research of polar motion prediction.
文摘A new method improves prediction of the motion of a hybrid monohull in regular waves. Stem section hydrodynamic coefficients of a hybrid monohull with harmonic oscillation were computed using the Reynolds Averaged Navier-Stokes Equations （RANSE）. The governing equations were solved using the finite volume method. The VOF method was used for free surface treatment, and RNGK-ε turbulence model was employed in viscous flow calculation. The whole computational domain was divided into many blocks each with structured grids, and the dynamic process was treated with moving grids. Using a 2-D strip method and 2.5D theory with the correction hydrodynamic coefficients allows consideration of the viscous effect when predicting longitudinal motion of a hybrid monohull in regular waves. The method is effective at predicting motion of a hybrid monohull, showing that the viscous effect on a semi-submerged body cannot be ignored.
基金supported by the National Natural Science Foundation of China(Grant No.41704015,42271436)the Shandong Natural Science Foundation of China(Grant No.ZR2017MD032,ZR2021MD030)+1 种基金a Project of Shandong Province Higher Education Science and Technology Program(Grant No.J17KA077)Talent introduction plan for Youth Innovation Team in universities of Shandong Province(innovation team of satellite positioning and navigation).
文摘High-precision polar motion(PM) prediction is of important significance in astronomy, geodesy, aviation,hydrographic mapping, interstellar navigation, and so on. SSA can effectively extract the trend and period terms of PM,in the process of achieving high-precision medium-and long-term polar motion prediction,it is necessary to solve the end effect problem and overfitting problem of SSA forecasting method;therefore, ARMA was applied to decreasethe end effect, and a suitable combination of reconstructed components was determined to avoid the high variance reaction of SSA overfitting. Based on the decomposition and reconstruction of the PM by SSA, the reconstructed components are determined to participate in the SSA iterative fitting model according to the variance contribution rate. The combination of the reconstructed components representing the polar motion period term and the trend term is determined according to the correlation analysis of the selected reconstructed components. After the above work, the principal component prediction sequence is obtained by fitting the period term and the trend term to convergence, respectively, and then, the SSA end effect is modified, and the residual term is predicted based on ARMA. The test results show that he prediction accuracy of SSA + ARMA at the front of the X and Y directions are improved by 96.90% and 97.53% compared with those of SSA, respectively,and the forecast accuracy of 365 days are improved by 37.93% and 19.53% in the X and Y directions compared with those of Bulletin A, respectively.
文摘A lightweight multi-layer residual temporal convolutional network model(RTCN)is proposed to target the highly complex kinematics and temporal correlation of human motion.RTCN uses 1-D convolution to efficiently obtain the spatial structure information of human motion and extract the correlation in the time series of human motion.The residual structure is applied to the proposed network model to alleviate the problem of gradient disappearance in the deep network.Experiments on the Human 3.6M dataset demonstrate that the proposed method effectively reduces the errors of motion prediction compared with previous methods,especially of long-term prediction.
基金This work was supported in part by the Key Program of NSFC(Grant No.U1908214)Program for Innovative Research Team in University of Liaoning Province(LT2020015)+1 种基金the Support Plan for Key Field Innovation Team of Dalian(2021RT06)the Science and Technology Innovation Fund of Dalian(Grant No.2020JJ25CY001).
文摘In recent years,human motion prediction has become an active research topic in computer vision.However,owing to the complexity and stochastic nature of human motion,it remains a challenging problem.In previous works,human motion prediction has always been treated as a typical inter-sequence problem,and most works have aimed to capture the temporal dependence between successive frames.However,although these approaches focused on the effects of the temporal dimension,they rarely considered the correlation between different joints in space.Thus,the spatio-temporal coupling of human joints is considered,to propose a novel spatio-temporal network based on a transformer and a gragh convolutional network(GCN)(STTG-Net).The temporal transformer is used to capture the global temporal dependencies,and the spatial GCN module is used to establish local spatial correlations between the joints for each frame.To overcome the problems of error accumulation and discontinuity in the motion prediction,a revision method based on fusion strategy is also proposed,in which the current prediction frame is fused with the previous frame.The experimental results show that the proposed prediction method has less prediction error and the prediction motion is smoother than previous prediction methods.The effectiveness of the proposed method is also demonstrated comparing it with the state-of-the-art method on the Human3.6 M dataset.
基金European Commission,Joint Research Center,Grant/Award Number:HUMAINTMinisterio de Ciencia e Innovación,Grant/Award Number:PID2020‐114924RB‐I00Comunidad de Madrid,Grant/Award Number:S2018/EMT‐4362 SEGVAUTO 4.0‐CM。
文摘Predicting the motion of other road agents enables autonomous vehicles to perform safe and efficient path planning.This task is very complex,as the behaviour of road agents depends on many factors and the number of possible future trajectories can be consid-erable(multi-modal).Most prior approaches proposed to address multi-modal motion prediction are based on complex machine learning systems that have limited interpret-ability.Moreover,the metrics used in current benchmarks do not evaluate all aspects of the problem,such as the diversity and admissibility of the output.The authors aim to advance towards the design of trustworthy motion prediction systems,based on some of the re-quirements for the design of Trustworthy Artificial Intelligence.The focus is on evaluation criteria,robustness,and interpretability of outputs.First,the evaluation metrics are comprehensively analysed,the main gaps of current benchmarks are identified,and a new holistic evaluation framework is proposed.Then,a method for the assessment of spatial and temporal robustness is introduced by simulating noise in the perception system.To enhance the interpretability of the outputs and generate more balanced results in the proposed evaluation framework,an intent prediction layer that can be attached to multi-modal motion prediction models is proposed.The effectiveness of this approach is assessed through a survey that explores different elements in the visualisation of the multi-modal trajectories and intentions.The proposed approach and findings make a significant contribution to the development of trustworthy motion prediction systems for autono-mous vehicles,advancing the field towards greater safety and reliability.
基金financially supported by the National Key Research and Development Program of China (Grant No.2022YFE010700)the National Natural Science Foundation of China (Grant No.52171259)+1 种基金the High-Tech Ship Research Project of Ministry of Industry and Information Technology (Grant No.[2021]342)Foundation of State Key Laboratory of Ocean Engineering in Shanghai Jiao Tong University (Grant No.GKZD010086-2)。
文摘During ship operations,frequent heave movements can pose significant challenges to the overall safety of the ship and completion of cargo loading.The existing heave compensation systems suffer from issues such as dead zones and control system time lags,which necessitate the development of reasonable prediction models for ship heave movements.In this paper,a novel model based on a time graph convolutional neural network algorithm and particle swarm optimization algorithm(PSO-TGCN)is proposed for the first time to predict the multipoint heave movements of ships under different sea conditions.To enhance the dataset's suitability for training and reduce interference,various filter algorithms are employed to optimize the dataset.The training process utilizes simulated heave data under different sea conditions and measured heave data from multiple points.The results show that the PSO-TGCN model predicts the ship swaying motion in different sea states after 2 s with 84.7%accuracy,while predicting the swaying motion in three different positions.By performing a comparative study,it was also found that the present method achieves better performance that other popular methods.This model can provide technical support for intelligent ship control,improve the control accuracy of intelligent ships,and promote the development of intelligent ships.
基金supported by the Beijing Municipal Science and Technology Commission and Zhongguancun Science Park Management Committee,No.Z221100002722020National Nature Science Foundation of China,No.62072045Innovation Transfer Fund of Peking University Third Hospital,No.BYSYZHKC2021110。
文摘Recently,human motion prediction has gained significant attention and achieved notable success.However,current methods primarily rely on training and testing with ideal datasets,overlooking the impact of variations in the viewing distance and viewing angle,which are commonly encountered in practical scenarios.In this study,we address the issue of model invariance by ensuring robust performance despite variations in view distances and angles.To achieve this,we employed Riemannian geometry methods to constrain the learning process of neural networks,enabling the prediction of invariances using a simple network.Furthermore,this enhances the application of motion prediction in various scenarios.Our framework uses Riemannian geometry to encode motion into a novel motion space to achieve prediction with an invariant viewing distance and angle using a simple network.Specifically,the specified path transport square-root velocity function is proposed to aid in removing the view-angle equivalence class and encode motion sequences into a flattened space.Motion coding by the geometry method linearizes the optimization problem in a non-flattened space and effectively extracts motion information,allowing the proposed method to achieve competitive performance using a simple network.Experimental results on Human 3.6M and CMU MoCap demonstrate that the proposed framework has competitive performance and invariance to the viewing distance and viewing angle.
基金supported by the National Natural Science Foundation of China(Nos.62403171,T2388101)。
文摘Accurate motion prediction of free-tumbling satellites is crucial for the success of capture operations.This paper proposes a two-step method to estimate the motion states and parameters of such satellites,thereby enabling precise long-term motion prediction.This paper begins with a measurement of the system's degree of observability,quantified through the Empirical Observability Gramian(EOG).Based on this measurement,a batch processing algorithm is first employed to estimate the satellite's constant parameters offline.Subsequently,an online filtering algorithm,utilizing a minimal state set,fine-tunes these parameters and estimates the motion states in real time.This integrated approach significantly enhances both convergence properties and estimation accuracy,particularly for systems with poor observability.Utilizing the predicted long-term motion of the satellite,a composite evaluation metric is formulated to identify the optimal capture point and moment.The base pose of the space robot is then adjusted to ensure that the optimal capture point lies within the manipulator's dexterous workspace,which is determined through a pre-constructed capability map.The effectiveness of the proposed method is demonstrated through both simulation and experimental results.
基金supported by the National Key Research and Development Program of China under Grant No.2023YFC3805200the National Natural Science Foundation of China under Grant Nos.52178495 and 52308513the Postdoctoral Fellowship Program of CPSF under Grant No.GZB20230487。
文摘The traditional approach to probabilistic seismic hazard analysis(PSHA)relies on ground motion records,which restricts its application in regions with sparse seismic records or low seismicity.Recently,the 3D physics-based simulation(PBS)has been recognized as a more effective tool,which offers the flexibility to generate time histories of simulated ground motions.The PBS methods are essential for obtaining ground motion parameters and compensating for lack of records.In this study,building on the theoretical framework of the China Probabilistic Seismic Hazard Analysis(CPSHA)method,we integrated the hierarchical potential focal region model from the fifth-generation seismic ground motion parameters zonation map of China and an appropriate seismicity model reflecting spatial distribution characteristics.Ground motion parameters at the target scale were calculated using PBS for near-field seismic simulations and ground motion prediction equations(GMPEs)for far-field seismic predictions,accounting for the uncertainties in ground motion attenuation from both methods to compute the seismic hazard of each site.In this manner,we established a comprehensive regional probabilistic seismic hazard analysis method combining PBS and GMPE.Using Tianjin as a case study,a probabilistic seismic hazard analysis was conducted with this method,providing seismic hazard curves for specific sites within each administrative region and zoning maps,which were then compared with the results of the fifth-generation zonation maps.The results indicate that the calculated seismic hazard values are generally consistent with the fifth-generation map at the lower limit,while the upper limit is slightly higher due to the near-fault effect.
基金the National Key Research and Development Program of China(No.2020YFC2007500)the Science and Technology Commission of Shanghai Municipality(No.20DZ2220400)。
文摘In the face of the large number of people with motor function disabilities,rehabilitation robots have attracted more and more attention.In order to promote the active participation of the user's motion intention in the assisted rehabilitation process of the robots,it is crucial to establish the human motion prediction model.In this paper,a hybrid prediction model built on long short-term memory(LSTM)neural network using surface electromyography(sEMG)is applied to predict the elbow motion of the users in advance.This model includes two sub-models:a back-propagation neural network and an LSTM network.The former extracts a preliminary prediction of the elbow motion,and the latter corrects this prediction to increase accuracy.The proposed model takes time series data as input,which includes the sEMG signals measured by electrodes and the continuous angles from inertial measurement units.The offline and online tests were carried out to verify the established hybrid model.Finally,average root mean square errors of 3.52°and 4.18°were reached respectively for offline and online tests,and the correlation coefficients for both were above 0.98.
文摘Ship motion attitude is influenced by dynamic marine conditions,presenting significant challenges in developing effective prediction networks.Contemporary prediction networks demonstrate limitations in hidden feature extraction,long-term dependency maintenance,and frequency characteristic incorporation.This paper presents an enhanced model integrating the informer network with a Time Convolutional Network(TCN)and a Frequency-Enhanced Channel Attention Mechanism(FECAM).The model employs a TCN for multi-feature extraction and applies Dimension-Segment-Wise(DSW)embedding for comprehensive multi-dimensional sequence analysis.Furthermore,it incorporates discrete cosine transform within the FECAM module for thorough data frequency analysis.The model integrates these components with the informer model for multivariate prediction.This approach maintains the informer model's capabilities in long-term multivariate prediction while enhancing feature extraction and local frequency information capture from ship motion attitude data,thus improving long-term multivariate prediction accuracy.Experimental results indicate that the proposed model outperforms traditional ship motion attitude prediction methods in forecasting future motion,reducing attitude prediction errors,and improving prediction accuracy.
基金supported in part by the National Key R&D Program of China under grant 2022YFC240014102the Beijing Municipal Natural Science Foundation-Haidian Primitive Innovation Joint Fund Project under grant L212005the National Natural Science Foundation of China under grants 62073043 and 62003045.
文摘Respiratory motion-induced vertebral movements can adversely impact intraoperative spine surgery,resulting in inaccurate positional information of the target region and unexpected damage during the operation.In this paper,we propose a novel deep learning architecture for respiratory motion prediction,which can adapt to different patients.The proposed method utilizes an LSTM-AE with attention mechanism network that can be trained using few-shot datasets during operation.To ensure real-time performance,a dimension reduction method based on the respiration-induced physical movement of spine vertebral bodies is introduced.The experiment collected data from prone-positioned patients under general anaesthesia to validate the prediction accuracy and time efficiency of the LSTM-AE-based motion prediction method.The experimental results demonstrate that the presented method(RMSE:4.39%)outperforms other methods in terms of accuracy within a learning time of 2 min.The maximum predictive errors under the latency of 333 ms with respect to the x,y,and z axes of the optical camera system were 0.13,0.07,and 0.10 mm,respectively,within a motion range of 2 mm.
文摘Accurate prediction of shipmotion is very important for ensuringmarine safety,weapon control,and aircraft carrier landing,etc.Ship motion is a complex time-varying nonlinear process which is affected by many factors.Time series analysis method and many machine learning methods such as neural networks,support vector machines regression(SVR)have been widely used in ship motion predictions.However,these single models have certain limitations,so this paper adopts amulti-model prediction method.First,ensemble empirical mode decomposition(EEMD)is used to remove noise in ship motion data.Then the randomforest(RF)prediction model optimized by genetic algorithm(GA),back propagation neural network(BPNN)prediction model and SVR prediction model are respectively established,and the final prediction results are obtained by results of three models.And the weights coefficients are determined by the correlation coefficients,reducing the risk of prediction and improving the reliability.The experimental results show that the proposed combined model EEMD-GARF-BPNN-SVR is superior to the single predictive model and more reliable.The mean absolute percentage error(MAPE)of the proposed model is 0.84%,but the results of the single models are greater than 1%.
基金supported by the National Natural Science Foundation of China(61701270)the Young Doctor Cooperation Foundation of Qilu University of Technology(Shandong Academy of Sciences)(2017BSHZ008)。
文摘Human motion prediction is a critical issue in human-robot collaboration(HRC)tasks.In order to reduce the local error caused by the limitation of the capture range and sampling frequency of the depth sensor,a hybrid human motion prediction algorithm,optimized sliding window polynomial fitting and recursive least squares(OSWPF-RLS)was proposed.The OSWPF-RLS algorithm uses the human body joint data obtained under the HRC task as input,and uses recursive least squares(RLS)to predict the human movement trajectories within the time window.Then,the optimized sliding window polynomial fitting(OSWPF)is used to calculate the multi-step prediction value,and the increment of multi-step prediction value was appropriately constrained.Experimental results show that compared with the existing benchmark algorithms,the OSWPF-RLS algorithm improved the multi-step prediction accuracy of human motion and enhanced the ability to respond to different human movements.
基金supported in part by the National Natural Sci-ence Foundation of China under Grant 42376187in part by the National Key R&D Program of China under Grant 2023YFC2812800,in part by the Natural Science Foundation of Shanghai under Grant 22ZR1434600+2 种基金in part by the Oceanic Interdisciplinary Program of Shanghai Jiao Tong University under Grant SL2022MS016in part by the Shanghai Jiao Tong University 2030 Initiative under Grant WH510244001in part by the Shanghai Underwater Robot En-gineering Technology Innovation Center under Grant 21DZ2221600.
文摘Ensuring accurate parameter identification and diving motion prediction of marine crafts is essential for safe navigation,optimized operational efficiency,and the advancement of marine exploration.Addressing this,this paper proposes an instrumental variable-based least squares(IVLS)algorithm.Firstly,aiming to balance complexity with accuracy,a decoupled diving model is constructed,incorporating nonlinear actuator characteristics,inertia coefficients,and damping coefficients.Secondly,a discrete parameter identification matrix is designed based on this dedicated model,and then a IVLS algorithm is innovatively derived to reject measurement noise.Furthermore,the stability of the proposed algorithm is validated from a probabilistic point of view,providing a solid theoretical foundation.Finally,performance evaluation is conducted using four depth control datasets obtained from a piston-driven profiling float in Qiandao Lake,with desired depths of 30 m,40 m,50 m,and 60 m.Based on the diving dynamics identification results,the IVLS algorithm consistently shows superior performance when compared to recursive weighted least squares algorithm and least squares support vector machine algorithm across all depths,as evidenced by lower average absolute error(AVGAE),root mean square error(RMSE),and maximum absolute error values and higher determination coefficient(R2).Specifically,for desired depth of 60 m,the IVLS algorithm achieved an AVGAE of 0.553 m and RMSE of 0.655 m,significantly outperforming LSSVM with AVGAE and RMSE values of 8.782 m and 11.117 m,respectively.Moreover,the IVLS algorithm demonstrates a remarkable generalization capability with R2 values consistently above 0.95,indicating its robustness in handling varied diving dynamics.
基金supported by Chinese Academy of Sciences(No.201491)“Light of West China” Program(201491)
文摘Previous studies revealed that the error of pole coordinate prediction will significantly increase for a prediction period longer than 100 days, and this is mainly caused by short period oscillations. Empirical mode decomposition （EMD）, which is increasingly popular and has advantages over classical wavelet decomposition, can be used to remove short period variations from observed time series of pole co- ordinates. A hybrid model combing EMD and extreme learning machine （ELM）, where high frequency signals are removed and processed time series is then modeled and predicted, is summarized in this paper. The prediction performance of the hybrid model is compared with that of the ELM-only method created from original time series. The results show that the proposed hybrid model outperforms the pure ELM method for both short-term and long-term prediction of pole coordinates. The improvement of prediction accuracy up to 360 days in the future is found to be 24.91% and 26.79% on average in terms of mean absolute error （MAE） for the xp and yp components of pole coordinates, respectively.
