Jiwen Zhang

Dr. Jiwen Zhang


Research Focus & Metrics

Academic Career Length

10Years (2017-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
28 July 2020 in Journal of Mechanisms and Robotics
Verified
Journal Article
Closed access
PDF file
MHMing Hao
Ming Hao
JZJiwen Zhang
Jiwen Zhang
KCken chen
ken chen
HAHaruhiko Harry Asada
Haruhiko Harry Asada
CFChenlong Fu
Chenlong Fu
Walking with load carriage is a common requirement for individuals in many situations. Legged exoskeletons can transfer the load weight to the ground with rigid-leg structures, thus reducing the load weight borne by the human user. However, the inertia of paralleled structures and the mechanical joint tend to disturb natural motions of human limbs, leading to high-energy consumption. Different from exoskeletons, Supernumerary Robotic Limbs (SuperLimbs) are kinematically independent of the human limbs, thus avoiding the physical interference with the human limbs. In this paper, a SuperLimb system is proposed to assist the human walking with load carriage. The system has two rigid robotic limbs, and each robotic limb has four degrees-of-freedom (DOFs). The SuperLimbs can transfer the load weight to the ground through the rigid structures, thus reducing the weight borne by the human user. A hybrid control strategy is presented to assist the human as well as avoid disturbing user’s natural motions. Motions of the SuperLimb system are generated autonomously to follow the gait of the human user. The gait synchronization is controlled by a finite state machine, which uses inertial sensors to detect the human gait. Human walking experiments are conducted to verify this concept. Experiments indicate that the SuperLimbs can follow the human gait as well as distribute the load weight. Results show that our SuperLimb system can reduce 85.7% of load weight borne by the human when both robotic limbs support and 55.8% load weight on average. This study may inspire the design of other wearable robots and may provide efficient solutions for human loaded walking.

23 November 2019 in Sensors
Verified
Journal Article
Open access
PDF file
ZWZe Wu
Ze Wu
JZJiwen Zhang
Jiwen Zhang
KCken chen
ken chen
CFChenlong Fu
Chenlong Fu
Currently, with the satisfaction of people’s material life, sports, like yoga and tai chi, have become essential activities in people’s daily life. For most yoga amateurs, they could only learn yoga by self-study, like mechanically imitating from yoga video. They could not know whether they performed standardly without feedback and guidance. In this paper, we proposed a full-body posture modeling and quantitative evaluation method to recognize and evaluate yoga postures to provide guidance to the learner. Back propagation artificial neural network (BP-ANN) was adopted as the first classifier to divide yoga postures into different categories, and fuzzy C-means (FCM) was utilized as the second classifier to classify the postures in a category. The posture data on each body part was regarded as a multidimensional Gaussian variable to build a Bayesian network. The conditional probability of the Gaussian variable corresponding to each body part relative to the Gaussian variable corresponding to the connected body part was used as criterion to quantitatively evaluate the standard degree of body parts. The angular differences between nonstandard parts and the standard model could be calculated to provide guidance with an easily-accepted language, such as “lift up your left arm”, “straighten your right forearm”. To evaluate our method, a wearable device with 11 inertial measurement units (IMUs) fixed onto the body was designed to measure yoga posture data with quaternion format, and the posture database with a total of 211,643 data frames and 1831 posture instances was collected from 11 subjects. Both the posture recognition test and evaluation test were conducted. In the recognition test, 30% data was randomly picked from the database to train BP-ANN and FCM classifiers, and the recognition accuracy of the remaining 70% data was 95.39%, which is highly competitive with previous posture recognition approaches. In the evaluation test, 30% data were picked randomly from subject three, subject four, and subject six, to train the Bayesian network. The probabilities of nonstandard parts were almost all smaller than 0.3, while the probabilities of standard parts were almost all greater than 0.5, and thus the nonstandard parts of body posture could be effectively separated and picked for guidance. We also tested separately the trainers’ yoga posture performance in the condition of without and with guidance provided by our proposed method. The results showed that with guidance, the joint angle errors significantly decreased.

24 June 2024 in IEEE Transactions on Instrumentation and Measurement
Verified
Journal Article
Closed access
JXJing Xu
Jing Xu
ZLZhihu Li
Zhihu Li
HLHe Liu
He Liu
RCRui Chen
Rui Chen
DWDan Wu
Dan Wu
GWGuolei Wang
Guolei Wang
Sinusoidal fringe projection profilometry (FPP) is a promising 3D measurement technique. However, the accuracy of sinusoidal FPP will be seriously reduced due to the impact of the defocusing effect. Different from existing methods with constant fringe shapes, this paper proposes a defocus-driven variable-shape method that adapts the fringe shape to the defocusing degree to improve accuracy. The defocusing effect reduces the modulation of each harmonic in the fringe, especially the high-order harmonics. Taking advantage of this property, high-order harmonics can be introduced to the fringe to improve the fundamental harmonic modulation, thereby achieving a higher signal-to-noise ratio (SNR) without causing significant harmonic errors. With the increase of defocusing degree, more high-order harmonics can be introduced to the fringe, leading to a varying optimal fringe shape. This paper employs a parameterized fringe shape and an error optimization method to determine the optimal fringe shape for each defocusing degree. Then, the variable-shape fringe is generated according to the defocusing distribution of the measurement target. The experimental results demonstrate that the variable-shape fringe can achieve higher accuracy than the traditional methods with constant fringe shapes.

12 April 2017 in Robotica
Verified
Journal Article
Closed access
PDF file
JZJiwen Zhang
Jiwen Zhang
ZXZeyang Xia
Zeyang Xia
LLLi Liu
Li Liu
KCken chen
ken chen
SUMMARY: Stability, high response quality and rapidity are required for reactive omnidirectional walking in humanoids. Early schemes focused on generating gaits for predefined footstep locations and suffered from the risk of falling over because they lacked the ability to suitably adapt foot placement. Later methods combining stride adaptation and center of mass (COM) trajectory modification experienced difficulties related to increasing computing loads and an unwanted bias from the desired commands. In this paper, a hierarchical planning framework is proposed in which the footstep adaption task is separated from that of COM trajectory generation. A novel omnidirectional vehicle model and the inequalities deduced therefrom are adopted to describe the inter-pace connection relationship. A constrained nonlinear optimization problem is formulated and solved based on these inequalities to generate the optimal strides. A black-box optimization problem is then constructed and solved to determine the model constants using a surrogate-model-based approach. A simulation-based verification of the method and its implementation on a physical robot with a strictly limited computing capacity are reported. The proposed method is found to offer improved response quality while maintaining rapidity and stability, to reduce the online computing load required for reactive walking and to eliminate unnecessary bias from walking intentions.

15 April 2025 in IEEE Transactions on Instrumentation and Measurement
Verified
Journal Article
Closed access
BWBin Wang
Bin Wang
ZLZhaowei Liang
Zhaowei Liang
JZJiwen Zhang
Jiwen Zhang
DWDan Wu
Dan Wu
Thin parts are important industrial components in 3C precision assembly tasks. The six-degree-of-freedom(6-D) pose measurement of thin parts is crucial during the alignment process for robotic precision assembly. Due to their small thickness and weak texture, image geometric features from two surfaces couple together, reducing the pose measurement precision. Furthermore, multiple irregular shapes of thin parts bring significant challenges for high-precision pose measurement. In this study, a general two-stage 6-D pose measurement framework based on monocular vision for thin parts with arbitrary irregular cross-sectional shape is proposed. In the first stage, the coarse pose is predicted by a deep neural network with the input of the outer contour of the part in image. The network is rapidly trained exclusively in simulation and subsequently transferred directly for real application. In the second stage, an index named model reprojection contour mismatch (MRCM) is proposed to evaluate the estimated pose, and then two-step iterative optimization based on the MRCM is executed to obtain the fine pose. The secondary optimization strategy is deigned to further improve the precision and make the method applicable to parts with higher thickness. Both the applicability and precision of the proposed method are validated through adequate simulations and experiments. The proposed method demonstrates significantly higher precision in 6-D pose measurement compared to existing methods, while maintaining acceptable time efficiency.

22 May 2023 in IEEE Transactions on Instrumentation and Measurement
Verified
Journal Article
Closed access
ZLZhihu Li
Zhihu Li
CHChangping Hu
Changping Hu
JZJiwen Zhang
Jiwen Zhang
GWGuolei Wang
Guolei Wang
DWDan Wu
Dan Wu
JXJing Xu
Jing Xu
Sinusoidal patterns are widely employed in fringe projection profilometry for high-accuracy measurements. However, in practice, random factors such as camera noise will seriously affect the accuracy of sinusoidal patterns. This paper presents a smoothed-truncated-sine (STS) pattern design method that can achieve higher accuracy than sinusoidal patterns under the same noise. By insensitive harmonic modulation, the STS pattern can achieve a higher signal-to-noise ratio (SNR) to reduce the random error and maintain the amplitude of sensitive harmonics low to restrain the systematic error. First, the error mechanism is analyzed, and the error model is established for the phase shifting algorithm, taking into account the resistance of the phase shifting algorithm to part of high-order harmonic errors. Second, a STS pattern is proposed and optimized based on the error model. The parameters of the STS pattern are optimized to achieve a high SNR and a small phase error, resulting in improved performance of the multiple-step phase shifting algorithm, especially when the number of steps is large. Simulations and experiments demonstrate that the proposed method achieves higher accuracy than sinusoidal patterns and that the maximum error reduction ratio exceeds 20% if the number of steps is sufficiently large in sinusoidal fringe projection profilometry.