graphic

Pai Zheng, Guest Editor

This Special Issue serves as a bridge between the ASME Journal of Manufacturing Science and Engineering (JMSE) and the global community of manufacturing researchers. Its primary objective is to curate a collection of high-level scientific articles that push the boundaries of knowledge in the realm of Human–Robot Collaboration (HRC) for forward-looking, human-centric smart manufacturing. It encourages researchers to present their innovative methodologies, tools, systems, and practical case studies, fostering advancements that integrate cognitive computing, mixed reality, and advanced data analytics. By emphasizing proactive teamwork and seamless interaction, this initiative aims to narrow the gap between human operators and industrial robots. Contributions are sought in areas such as cognitive HRC systems, safety considerations, adaptive motion planning, human intention prediction, and semantic knowledge representation—key components in achieving efficient and effective collaboration within the manufacturing industry. Beyond its scientific impact, this Special Issue also seeks to unite leading scientific communities worldwide.

A dedicated team of Guest Editors has been assembled to curate a diverse selection of compelling articles for this Special Issue. Led by Dr. Pai Zheng (The Hong Kong Polytechnic University, Hong Kong SAR, China), the team includes esteemed members: Prof. Jinsong Bao (Donghua University, China), Dr. Tao Peng (Zhejiang University, China), Dr. Xi Vincent Wang (KTH Royal Institute of Technology in Stockholm, Sweden), Prof. Lihui Wang (KTH Royal Institute of Technology in Stockholm, Sweden), and Prof. Aydin Nassehi (University of Bristol, UK).

The collection of articles spans from May to September 2023, and each article has undergone a rigorous peer review process in collaboration with our associate editors, a hallmark of JMSE’s commitment to excellence. We are thrilled to showcase 11 exceptional articles in this Special Issue, representing contributions from researchers in China, the USA, France, Italy, and the Republic of Korea, thus bringing together leading scientific communities from around the world.

While this collection is not exhaustive, it serves as a snapshot of the current landscape of advanced manufacturing research on a global scale. These articles cover a wide array of topics, including human intention prediction, adaptive robot motion planning, cognitive systems, and implementations in manufacturing. Together, they shed light on the intricate dynamics of HRC systems within manufacturing tasks, with the overarching goal of enabling flexible and resilient manufacturing. The main features of these articles are briefly stated below:

Human intention prediction in HRC is a key context-awareness capability for HRC systems [1], which contains the perception of human motion, action, status, etc.

Xu et al. [2] discussed the importance of seamless turn-taking in HRC to enhance production efficiency. It highlighted the challenge of robots adapting to different human operating habits and varying proficiency levels, for early predictions of turn-taking actions. The paper proposed a method using spiking neural networks for early turn-taking prediction in HRC assembly tasks. It incorporated dynamic motion primitives to establish trajectory templates for different operating speeds, adjusting to human uncertainty. The method’s effectiveness was demonstrated through a gear assembly case, showing reduced turn-taking recognition time in HRC.

Flowers et al. [3] coordinated human motions and robot movements in HRC effectively. The Spatio-Temporal Avoidance of Predictions-Prediction and Planning Framework (STAP-PPF) was introduced to predict multi-step human motions based on the objects manipulated by humans, allowing proactive determination of time-optimal robot paths while considering anticipated human motion and robot speed restrictions. It continuously updated predictions and adapted robot paths in real time to mitigate delays and human discomfort. Results indicated that STAP-PPF generated shorter robot trajectories, adapted well to real-time human motion, and maintained improved robot–human separation during close interactions.

To enhance both human well-being and robotic flexibility within HRC, Fan et al. [4] modeled a Human Digital Twin (HDT), which contains 3D human posture, action intention, and ergonomic risk. The HDT was a centralized digital representation of various human data integrated into the cyber-physical production system to optimize performance and efficiency in HRC. It was then used to adaptively optimize robotic motion trajectories in dynamic HRC applications. Experiments and case studies were conducted to demonstrate the effectiveness of this approach.

Adaptive motion planning in HRC is the prerequisite to ensure human safety and allow fluent collaboration in direct and indirect contact between humans and robots in close proximity [5].

Xiao et al. [6] used Multi-Agent Reinforcement Learning (MARL) to plan HRC disassembly tasks of electric vehicle batteries. Among the system, a 2D planar disassembly trajectory is established using the Q-learning algorithm to determine optimal disassembly paths. Then, a standard trajectory was matched to complete the disassembly task sequence. The feasibility of this method was verified through disassembly operations on a battery module case.

Waseem et al. [7] focused on the real-time scheduling for these robots in the dynamic and uncertain shop floor environment. The problem was framed as a Markov Decision Process (MDP), and the Q-learning algorithm was employed to determine an optimal policy for the robot’s movements, considering various product types. The proposed approach was validated through a numerical case study, which demonstrates a notable improvement in production throughput (approximately 23%) compared to alternative policies.

Bai et al. [8] addressed the issue of calibration errors and multiple disturbances in processing trajectory for large-scale machining features. Their paper introduced a practical path- tracking synchronous control algorithm. The proposed path-tracking controller corrected paths dynamically while following pre-planned paths. Meanwhile, cross-coupled technology was integrated into the control algorithm, using real-time feedback from a highly repeatable 3D visual measurement instrument to ensure that tracking and synchronous errors of the dual manipulators converge to zero.

Cognitive HRC systems focuses on knowledge representation learning in HRC scenarios, enhancing robot learning capability and improving human well-being among teamwork [9].

Park et al. [10] discussed the development of autonomous robotic operations generated from human demonstration. The platform employed the YOLOv5 neural network model for rapid object localization without the need for prior CAD models or extensive datasets during training. After a simple human demonstration to identify the target object for picking and placing, an Iterative Closest Points (ICP) algorithm estimated the target object’s pose based on depth data. Experimental results with four object types and four human demonstrations showed that the platform can recognize the target object and estimate its pose within 0.5 s, achieving a 95.6% success rate.

Verna et al. [11] aimed to align with Industry 5.0 and ensure human well-being in HRC. A tool named “Human-Robot Collaboration Quality and Well-Being Assessment Tool” (HRC-QWAT) was introduced. This tool evaluated human well-being, specifically in terms of the stress response. The paper presents a case study of collaborative human–robot assembly to illustrate the tool’s applicability. The results demonstrate that the HRC-QWAT can effectively evaluate both production quality and human well-being, offering a valuable resource for companies to monitor and enhance manufacturing processes.

HRC systems, and implementations in manufacturing: HRC emerges as a promising solution to enhance manufacturing performance by leveraging the strengths of both human cognitive flexibility and adaptability, and robots’ precision, strength, and repeatability, of which cutting-edge works have been reported on its practical values.

Gao et al. [12] discussed the role of HRC in the task of surface defect inspection. Robots could assist in reducing workload, human collaboration was essential for rechecking uncertain defects. However, a bottleneck in this collaboration is the lack of a method to determine which samples should be rechecked. To address this issue and enable HRC-based surface defect inspection, the paper introduced a two-stage Transformer model with focal loss. This method divided the inspection process into detection and recognition, establishing collaboration rules for workers to recheck defects. The proposed approach allowed a robot to effectively collaborate with workers to improve surface quality, with experimental results demonstrating significant accuracy improvements (1.70–4.18%).

Gao et al. [13] introduced a dual-metric neural network with attention guidance. It employed an attention-guided recognition network incorporating channel attention and position attention modules to efficiently learn representative defect features from small samples. Then, to address defects with confusing surface images, a dual-metric function was introduced to control the distance between samples in the feature space, enhancing classification boundaries between intra-class and inter-class samples. Experimental results using a fabric defect dataset demonstrated that this approach surpassed other methods in terms of accuracy, recall, precision, F1-score, and few-shot accuracy.

Wang et al. [14] addressed the issue of limited availability of small-scale data in the human cyber-physical system. The paper utilized a Data Augmentation-Gradient Boosting Decision Tree (DA-GBDT) model and an adaptive data augmentation rate selection algorithm to balance training time and prediction accuracy. Experimental results on automobile covering products demonstrated that this method effectively reduced the average prediction error compared to conventional quality prediction methods. The predicted quality information could guide product optimization decisions in smart manufacturing systems.

Finally, the guest editors extend their heartfelt gratitude to all the authors for their valuable contributions to this special issue. They also appreciate the thoughtful feedback and recommendations provided by the anonymous reviewers. It is envisioned that this special issue will combine advanced technologies with human-centered design principles to create a collaborative and adaptive production environment for enhancing productivity, safety, and operator well-being.

Conflict of Interest

There are no conflicts of interest.

Data Availability Statement

No data, models, or code were generated or used for this paper.

References

1.
Li
,
S.
,
Zheng
,
P.
,
Fan
,
J.
, and
Wang
,
L.
,
2022
, “
Toward Proactive Human-Robot Collaborative Assembly: A Multimodal Transfer-Learning-Enabled Action Prediction Approach
,”
IEEE Trans. Ind. Electron.
,
69
(
8
), pp.
8579
8588
.
2.
Xu
,
W.
,
Feng
,
S.
,
Yao
,
B.
,
Ji
,
Z.
, and
Liu
,
Z.
,
2023
, “
Turn-Taking Prediction for Human–Robot Collaborative Assembly Considering Human Uncertainty
,”
ASME J. Manuf. Sci. Eng.
,
145
(
12
), p.
121007
.
3.
Flowers
,
J.
, and
Wiens
,
G.
,
2023
, “
A Spatio-Temporal Prediction and Planning Framework for Proactive Human–Robot Collaboration
,”
ASME J. Manuf. Sci. Eng.
,
145
(
12
), pp.
1
14
.
4.
Fan
,
J.
,
Zheng
,
P.
, and
Lee
,
C. K. M.
,
2023
, “
A Vision-Based Human Digital Twin Modelling Approach for Adaptive Human-Robot Collaboration
,”
ASME J. Manuf. Sci. Eng.
,
145
(
12
), p.
121002
.
5.
Li
,
S.
,
Zheng
,
P.
,
Liu
,
S.
,
Wang
,
Z.
,
Wang
,
X. V.
,
Zheng
,
L.
, and
Wang
,
L.
,
2023
, “
Proactive Human–Robot Collaboration: Mutual-Cognitive, Predictable, and Self-Organising Perspectives
,”
Rob. Comput. Integr. Manuf.
,
81
, pp.
1
30
.
6.
Xiao
,
J.
,
Gao
,
J.
,
Anwer
,
N.
, and
Eynard
,
B.
,
2023
, “
Multi-Agent Reinforcement Learning Method for Disassembly Sequential Task Optimization Based on Human-Robot Collaborative Disassembly in Electric Vehicle Battery Recycling
,”
ASME J. Manuf. Sci. Eng.
,
145
(
12
), p.
121001
.
7.
Waseem
,
M.
, and
Chang
,
Q.
,
2023
, “
Adaptive Mobile Robot Scheduling in Multiproduct Flexible Manufacturing Systems Using Reinforcement Learning
,”
ASME J. Manuf. Sci. Eng.
,
145
(
12
), p.
1211005
.
8.
Bai
,
Q.
,
Li
,
P.
,
Tian
,
W.
,
Shen
,
J.
,
Li
,
B.
, and
Hu
,
J.
,
2023
, “
Vision Guided Dynamic Synchronous Path Tracking Control of Dual Manipulator Cooperative System
,”
ASME J. Manuf. Sci. Eng.
,
145
(
12
), p.
121003
.
9.
Zheng
,
P.
,
Li
,
S.
,
Xia
,
L.
,
Wang
,
L.
, and
Nassehi
,
A.
,
2022
, “
A Visual Reasoning-Based Approach for Mutual-Cognitive Human-Robot Collaboration
,”
CIRP Ann.
,
77
(
1
), pp.
377
380
.
10.
Park
,
J.
,
Han
,
C.
,
Jun
,
M.
, and
Yun
,
H.
,
2023
, “
Autonomous Robotic Bin Picking Platform Generated From Human Demonstration and YOLOv5
,”
ASME J. Manuf. Sci. Eng.
,
145
(
12
), p.
121006
.
11.
Verna
,
E.
,
Puttero
,
S.
,
Genta
,
G.
, and
Galetto
,
M.
,
2023
, “
A Novel Diagnostic Tool for Human-Centric Quality Monitoring in Human-Robot Collaboration Manufacturing
,”
ASME J. Manuf. Sci. Eng.
,
145
(
12
), p.
121009
.
12.
Gao
,
Y.
,
Gao
,
L.
, and
Li
,
X.
,
2023
, “
A Two-Stage Focal Transformer for Human-Robot Collaboration-Based Surface Defect Inspection
,”
ASME J. Manuf. Sci. Eng.
,
145
(
12
), p.
121004
.
13.
Gao
,
P.
,
Wang
,
J.
,
Xia
,
M.
,
Qin
,
Z.
, and
Zhang
,
J.
,
2023
, “
Dual-Metric Neural Network With Attention Guidance for Surface Defect Few-Shot Detection in Smart Manufacturing
,”
ASME J. Manuf. Sci. Eng.
,
145
(
12
), pp.
1
12
.
14.
Wang
,
T.
,
Yang
,
C.
,
Bingtao
,
H.
,
Feng
,
Y.
,
Gao
,
X.
,
Yang
,
C.
, and
Tan
,
J.
,
2023
, “
Data Augmentation-Based Manufacturing Quality Prediction Approach in Human Cyber-Physical Systems
,”
ASME J. Manuf. Sci. Eng.
,
145
(
12
), p.
121008
.