Xiaohang Shi

Mr. Xiaohang Shi


Research Focus & Metrics

Academic Career Length

3Years (2024-Present)

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

Representative Publications

View All Publications
20 April 2025 in Journal of Robotics and Mechatronics
Verified
Journal Article
Open access
XSXiaohang Shi
Xiaohang Shi
YYYuji Yamakawa
Yuji Yamakawa
Manipulating deformable linear objects (DLOs) presents significant challenges, particularly in industrial applications. However, most previous research has primarily focused on the manipulation of single DLOs, leaving the problem of picking up stacked DLOs without causing collisions largely unexplored. This task is common in industrial production processes, such as the sorting of pipes or cables. This paper introduces a novel strategy to address this issue using a simple setup with only a depth sensor. The proposed approach begins by acquiring a point cloud of the scene. The supervoxel method is then employed to segment the points into several clusters. Based on these clusters, a graph is constructed and pruned by graph theory to search for DLO axes. The axes belonging to the same DLO are subsequently connected, and occlusion relationships are identified. Finally, the best target, optimal picking point, and motion planning strategy are evaluated. Experimental results demonstrate that this method is highly adaptable to complex scenarios and can accommodate DLOs of varying diameters.

7 October 2024 in IEEE Access
Verified
Journal Article
Open access
XSXiaohang Shi
Xiaohang Shi
QGQitong Guo
Qitong Guo
KMKenichi Murakami
Kenichi Murakami
YYYuji Yamakawa
Yuji Yamakawa
With the increasing application of robots in industrial manufacturing, the 3D vision perception is becoming more and more important. However, many researchers prefer to focus on the accuracy of 3D recognition rather than the speed. Faster speed can accelerate industrial production a lot. Meanwhile, the cost is also an important factor in reality. Since these, we propose the 2D-Edge-based 4-points congruent set (4PCS) algorithm. This algorithm utilizes the affine invariance and the projection invariance to find the correspondence between 3D model and 2D depth image, resulting in a significant increase in speed and keeping the accuracy. Its necessary hardware are just a consumer-level CPU and a consumer-level depth camera. On the Sileana dataset, the average recognition time is 0.5ms per image.

1 May 2025 in Optics Letters
Verified
Journal Article
Closed access
XSXiaohang Shi
Xiaohang Shi
QGQitong Guo
Qitong Guo
RJRuoyu Jia
Ruoyu Jia
CYChunxin Yang
Chunxin Yang
KMKenichi Murakami
Kenichi Murakami
YYYuji Yamakawa
Yuji Yamakawa
We proposed an ultra-low-latency projector-camera system for 3D imaging based on the one-dimensional de Bruijn pattern. The feature of this pattern is that four selected different 5-bit de Bruijn sequences form a 7-bit approximate de Bruijn sequence, thus increasing the number of codes while maintaining the short code length, which achieves the balance between precision and decoding speed. The one-dimension property adapts well to GPU acceleration. We measured various moving and deformable objects with different materials, colors, and contrasts, and the results performed very well. For the planar surface, the proposed system acquires 120.0K points per image, and the latency is 0.059 ms with 0.80-mm precision on average.

4 March 2026 in Robotics
Verified
Journal Article
Open access
PDF file
CYChunxin Yang
Chunxin Yang
RJRuoyu Jia
Ruoyu Jia
QGQitong Guo
Qitong Guo
XSXiaohang Shi
Xiaohang Shi
MHMasahiro Hirano
Masahiro Hirano
YYYuji Yamakawa
Yuji Yamakawa
Most existing research in human pose estimation focuses on predicting joint positions, paying limited attention to recovering the full 6D human pose, which comprises both 3D joint positions and bone orientations. Position-only methods treat joints as independent points, often resulting in structurally implausible poses and increased sensitivity to depth ambiguities—cases where poses share nearly identical joint positions but differ significantly in limb orientations. Incorporating bone orientation information helps enforce geometric consistency, yielding more anatomically plausible skeletal structures. Additionally, many state-of-the-art methods rely on large, computationally expensive models, which limit their applicability in real-time scenarios, such as human–robot collaboration. In this work, we propose STAG-Net, a novel 2D-to-6D lifting network that integrates Graph Convolutional Networks (GCNs), attention mechanisms, and Temporal Convolutional Networks (TCNs). By simultaneously learning joint positions and bone orientations, STAG-Net promotes geometrically consistent skeletal structures while remaining lightweight and computationally efficient. On the Human3.6M benchmark, STAG-Net achieves an MPJPE of 41.8 mm using 243 input frames. In addition, we introduce a lightweight single-frame variant, STG-Net, which achieves 50.8 mm MPJPE while operating in real time at 60 FPS using a single RGB camera. Extensive experiments on multiple large-scale datasets demonstrate the effectiveness and efficiency of the proposed approach.

25 March 2026 in IEEE Robotics and Automation Letters
Verified
Journal Article
Closed access
QGQitong Guo
Qitong Guo
XSXiaohang Shi
Xiaohang Shi
RJRuoyu Jia
Ruoyu Jia
CYChunxin Yang
Chunxin Yang
KMKenichi Murakami
Kenichi Murakami
YYYuji Yamakawa
Yuji Yamakawa
Although many impressive advances have been reported in the table tennis robots field using reinforcement learning method, challenges related to policy complexity and adaptability continue to hinder large-scale deployment and practical applications. In this work, building upon extensive prior studies, we propose several techniques and integrate them into a unified framework that reduces the difficulty of training and deployment while enhancing the human player enjoyment. Specifically, a recursively nested design is introduced in the hierarchical decision-making system, which fully separates high-level decision-making from low-level execution, eliminating the need to train multiple agents at the execution layer. The multi-objective problem is also studied by introducing a tolerance-based weight modulation mechanism, which can balance the landing accuracy with other objectives and adjust the playing strategy to meet diverse goals. Comprehensive experiments demonstrate the effectiveness of the proposed framework and techniques. Notably, the robot achieved up to 18 consecutive rallies with human player, which, to the best of our knowledge, is the longest rally attained with a fixed-base collaborative robot arm. The proposed methods can be readily extended to other racket-sport robots or similar tasks.

9 March 2026 in IEEE Robotics and Automation Letters
Verified
Journal Article
Closed access
XSXiaohang Shi
Xiaohang Shi
QGQitong Guo
Qitong Guo
RJRuoyu Jia
Ruoyu Jia
CYChunxin Yang
Chunxin Yang
KMKenichi Murakami
Kenichi Murakami
YYYuji Yamakawa
Yuji Yamakawa
Driven by significant advancements in structure, sensors, and control algorithms, multi-fingered hand systems have received increasing attention from both academia and industry. While existing research primarily focuses on the interaction static rigid objects manipulation, real-world scenarios often involve dynamic tasks and non-rigid materials. To address these challenges, we developed a multi-fingered manipulation system designed to catch dynamic, non-rigid objects with small-deformation. We decomposed this objective into three components: teleoperation, reinforcement learning-based control, and deformation feedback. This paper presents a 3D vision-based framework that integrates these elements. Experimental results demonstrate that the framework can catch a paper sheet with 5.40 mm deformation and a paper balloon with 12.19 mm deformation. Regarding system latency, the latencies for teleoperation, RL-based control, and deformation feedback are 1.29 ms, 0.65 ms, and 0.47 ms, respectively.