IEEE TRANSACTIONS ON INDUSTRIAL INFORMATICS, VOL. 19, NO. 9, SEPTEMBER 2023 9255 Impacts of Wireless on Robot Control: The Network Hardware-in-the-Loop Simulation Framework and Real-Life Comparisons Honghao Lv , Student Member, IEEE , Zhibo Pang , Senior Member, IEEE , Koushik Bhimavarapu , Member, IEEE , and Geng Yang , Member, IEEE Abstract — As many robot applications become more re- liant on wireless communications, wireless network latency and reliability have a growing impact on robot control. This article proposes a network hardware-in-the-loop (N- HiL) simulation framework to evaluate the impacts of wire- less on robot control more efficiently and accurately, and then improve the design by employing correlation analysis between communication and control performances. The N- HiL method provides communication and robot developers with more trustworthy network conditions, while the huge efforts and costs of building and testing the entire physical robot system in real life are eliminated. These benefits are showcased in two representative latency-sensitive applica- tions: 1) safe multirobot coordination for mobile robots, and 2) human-motion-based teleoperation for manipula- tors. Moreover, we deliver a preliminary assessment of two new-generation wireless technologies, the Wi-Fi6 and 5G, for those applications, which has demonstrated the effec- tiveness of the N-HiL method as well as the attractiveness of the wireless technologies. Index Terms — 5G, hardware-in-the-loop, multirobot coor- dination, teleoperation, Wi-Fi 6, wireless control. Manuscript received 18 October 2022; accepted 25 November 2022. Date of publication 8 December 2022; date of current version 24 July 2023. This work was supported in part by the Swedish Foundation for Strategic Research under Grant APR20-0023; in part by the National Natural Science Foundation of China under Grant 51975513; in part by the Natural Science Foundation of Zhejiang Province, China under Grant LR20E050003; in part by the Major Research Plan of National Natural Science Foundation of China under Grant 51890884; and in part by the Major Research Plan of Ningbo Innovation 2025 under Grant 2020Z022. The work of Honghao Lv was supported by the China Scholarship Council. Paper no. TII-22-4339. (Corresponding author: Zhibo Pang.) Honghao Lv is with the State Key Laboratory of Fluid Power and Mechatronic Systems, School of Mechanical Engineering, Zhejiang Uni- versity, Hangzhou 310027, China, and also with the Department of Intelligent Systems, Royal Institute of Technology, 114 28 Stockholm, Sweden (e-mail: lvhonghao@zju.edu.cn). Zhibo Pang is with the Department of Automation Technology, ABB Corporate Research, 722 26 Vasteras, Sweden, and also with the De- partment of Intelligent Systems, Royal Institute of Technology, 114 28 Stockholm, Sweden (e-mail: zhibo@kth.se). Koushik Bhimavarapu is with the Department of Automation Tech- nology, ABB Corporate Research, 722 26 Vasteras, Sweden (e-mail: koushik.bhimavarapu@se.abb.com). Geng Yang is with the State Key Laboratory of Fluid Power and Mechatronic Systems, School of Mechanical Engineering, Zhejiang Uni- versity, Hangzhou 310027, China (e-mail: yanggeng@zju.edu.cn). Color versions of one or more figures in this article are available at https://doi.org/10.1109/TII.2022.3227639. Digital Object Identifier 10.1109/TII.2022.3227639 I. I NTRODUCTION T HE growing wireless network technologies and the cloud and edge computing toward Industrial 4.0 enable various wireless-network-controlled robot platforms that are practically applied to industrial production [1], [2]. Intuitively, wireless networks enable higher flexibility of the robotic system and simplify design and installation processes, and diminish mainte- nance needs, especially for mobile robots [3], [4]. And wireless networks make it possible to implement a remote operation, because the controllers and actuators are deployed over a wire- less link, such as the teleoperation of the manipulators [5], [6]. However, wireless communication imposes challenges in time-critical robot control scenarios, where wireless networks involved extra latency compared to wired connection [7]. To ensure the robustness and reliability of the robotic system, simulation and testing for the communication are required dur- ing the design stage and deployment process of an industrial robotic platform [8]. Introducing latency or other communi- cation characteristics into the controller codes to imitate the profiles of the real network is a common solution for the sim- ulation [9]. However, the practical communication condition, particularly the long-term stable performance of a wireless network, is extremely difficult to be modeled and simulated [10]. Li and Savkin [11] simulated unmanned aerial vehicle (UAV) navigation based on the wireless sensor network using the professional simulation software V-REP in the industrial in- ternet of things applications. However, a stable and invulnerable network condition is assumed in the simulation environments. Despite the simulation method of [11] is able to do long-term stability tests but has no generalized capabilities to evaluate different network conditions and communication uncertainty. The model-based simulation methods are commonly used for validating the auto guide vehicles (AGVs) coordination, such as in [12] and [13], while both cannot be used to evaluate the communication uncertainty. Hardware-in-loop (HiL) simulation refers to the simulation technology, which simulates one part of the whole system with computer modeling while using physical modeling or an actual system for the other part [14]. In the past few years, research efforts have been made in HiL simulation for robot system design and optimization [15], [16]. Zhou et al. [14] proposed an HiL simulation method for underwater acoustic This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see http://creativecommons.org/licenses/by/4.0/

9256 IEEE TRANSACTIONS ON INDUSTRIAL INFORMATICS, VOL. 19, NO. 9, SEPTEMBER 2023 TABLE I C OMPARISONS B ETWEEN THE P ROPOSED A PPROACH AND O THER R ELATED S IMULATION W ORKS FOR R OBOT C ONTROL Fig. 1. Diagram of the proposed network hardware-in-the-loop framework. communication, but without applying it to robots or other devices. Marco et al. proposed an on-ground HiL simulation method for prevalidation and optimization of on-orbit robotic missions. However, the technical requirements of the proposed approach are the real-time computations and the negligible latency given by Assumption 1 in [16]. An HiL simulation system for manipulator control in [17] was built to improve the system stability by deploying the self-designed PID controller. Lamping et al. [18] worked on a multiagent UAV system based on robot operation system (ROS), and they experimented with control and supervision algorithms on multiple UAVs using HiL simulation. The majority of these HiL simulation studies for robotic applications aimed to provide a flexible and easy-to-use interface for a controller design [19]. Markus et al. [20] from ABB and NOKIA investigated the capabilities of 5G and LTE for supporting industrial robotic applications using the traffic- model-based communication simulation methods. However, the simulation results are constrained to the specific scenario de- fined in this article and do not have the ability to evaluate the impact of network uncertainty. One common constraint of HiL simulation for robotics, mechatronics, and control is the limited modeling of the communication condition, especially for the growing wireless network controlled robots [21]. Lv et al. [22] presented an HiL simulation framework and conducted a preliminary study on wireless robot control using Wi-Fi 6. But it lacks the systematic evaluation of the network performance, and the simultaneous correlation between communication and controlisnotelaborated.Thesummarizedcomparisonsarelisted in Table I . To cover the gap in the literature, a network hardware-in-the- loop(N-HiL)simulationframeworkshowninFig. 1 isdeveloped in this article for evaluating the impacts of the wireless network on robot control systems, considering the scalability of the net- work interface, and the friendliness to both the communication and robot developers in the field. The major contributions can be summarized as follows: 1) A novel N-HiL simulation framework is proposed to evaluate the impacts of wireless on robot control, which provides more trustworthy results, while the huge efforts and costs of testing the entire physical robot system in real life are eliminated. 2) Two typical latency-sensitive robot control cases, safe multirobot coordination and human-motion-based tele- operation, are investigated, and the performance compar- isons of Wi-Fi 6 and 5G wireless networks are conducted. 3) The correlation between the communication and control is investigated using the proposed N-HiL method, and a selection rule is given for choosing the wireless networks

9258 IEEE TRANSACTIONS ON INDUSTRIAL INFORMATICS, VOL. 19, NO. 9, SEPTEMBER 2023 Fig. 3. N-HiL simulation framework for safe multirobot coordination. deploy the test on the proposed N-HiL framework since it is a key to pretest for avoiding collision and ensuring safety [26]. Moreover, the manipulator teleoperation is chosen as the sec- ond test case, which is another typical communication-sensitive robot application [27]. In order to achieve accurate and smooth control of a multijoint manipulator remotely, the high-frequency discrete joint commands need to be accurately transferred from the remote side to the robot controller [28]. The packet loss and latency introduced by the communication will change the motion status suddenly, which will greatly affect the control performance. B. Case Study 1: Safe Multirobot Coordination In this case, a centralized multirobot coordination platform from Orebro University is deployed, in which robots are driven by generic second-order dynamics and coordinated by heuristic robot ordering policies [12]. As shown in Fig. 3 , the motion planner is used to calculate the path and trajectory of each robot. The modeling method for the coordination avoids the need for a priority discretization of the environment and allows robots to “follow each other” through CSs (overlap of multirobot trajectory envelops). The coordinator decides the robot order in the CS to avoid collisions and ensure safety. The simulated robot forward model is driven by the motion command from the con- troller, and the real-time multirobot motion status is displayed in a graphical user interface. The modeling method we used in this article has been validated formally and experimentally in simulation and with real robots [29]. The robot trajectory tracker tracks the robot trajectories and sends the real-time robot status to the controller. The controller and the robot forward model, as the simulated robot system, are achieved by Java, running on Ubuntu 20.04 PC. Here, two agents are designed to forward the robot command and update the real-time robot status, as illustrated in Fig. 3 . Agent 1 is responsible to forward the robot motion command from the controller to the simulated robot model, which is done over two independent UDP sockets. Agent 2 is responsible to forward the current robot status from the driven robot model to the controller. These two agents are programmed by Java and deployed on a Windows 10 PC. The network hardware interface Fig. 4. N-HiL simulation framework for human-motion-based teleoper- ation. is deployed between the simulated robot and the agents. To more closely match the real AGVs, the proposed N-HiL framework has the feasibility to add more simulated agents with more APs connections and add more sniffer channels to measure the wireless network performance. C. Case Study 2: Human-Motion-Based Teleoperation In this case, a human–robot motion transfer teleoperation system is deployed to evaluate the impacts of wireless using the N-HiL framework. Here, the captured motion data from the operator are used to control the YuMi robot. The external guide motion (EGM) interface from ABB provides a motion control interface with high frequency (up to 250 Hz), which enables real-time human–robot motion mapping. As shown in Fig. 4 , the software Axis Neuron captures human motion and sends the data to the robot controller via ROS serial protocol from Windows 10 to Ubuntu 20.04. The controller is responsible for processing and converting the human motion data to robot tool center point data using the designed mapping strategy [30]. The robot tool center point data are used to calculate the real-time joint configuration by the inverse kinematic solver. The motion controller program is connected to the motion server running on the simulated robot model via TCP protocol. The motion server is responsible to control normal motion like linear motion and joint motion using the given robot target. The UDP User Communication device is defined for the task of each arm, which receives the joint values at a high rate. The EGM motion is activated/deactivated by the motion server. When the EGM motion is activated, the generated joint value queue will be sent using Google Protocol Buffers (Protobuf) based on the UDP protocol. In this case, the communication between the human motion capture part on the Windows 10 system and the ubuntu 20.04 is connected by Ethernet. The network hardware interface is injected between the controller running on a ubuntu 20.04 PC and the simulated YuMi robot in RobotStudio on a Windows 10 PC. RobotStudio is a mature commercial simulation software provided by ABB and has been used for many years in the industry. It is used in this article for simulating the YuMi robot provides the same configuration of

LV et al.: IMPACTS OF WIRELESS ON ROBOT CONTROL 9259 Fig. 5. Hardware setup for the experiments of the above two case studies. The different hardware settings in the red dotted box are the options for the two case studies, respectively. Case1: safe multirobot coordination, Case2: human-motion-based teleoperation. the EGM interface and the IRC5 controller behavior compared to the real robot. IV. W IRELESS FOR S AFE M ULTIROBOT C OORDINATION To validate the reliability of the proposed N-HiL framework compared to the previous methods, comparative experiments be- tween the statistical modeling method and the proposed N-HiL method are conducted. The statistical modeling model used and compared to the N-HiL simulation results in the experimental Section IV is following the simulation methods from [29]. And the experiments and evaluations under the wireless network condition are carried out, choosing the Ethernet wired network performance as the baseline. The hardware setup for the safe multirobot coordination case is demonstrated in Fig. 5 with the case1 option. One NUC10 PC running the Open JDK with Intel(R) Core(TM) i5-10210U CPU @ 1.60 GHz × 4 processor is set as the simulated agent. An HP Laptop PC running the Ubuntu OS with Intel(R) Core(TM) i7-7500U CPU @ 2.70 GHz × 4 processor is set to simulate the safe multirobot coordination platform. The ULT PC is another HP Laptop PC to sniff the network latency from the ET2000 probe. During the simulation for the safe multirobot coordination case, the maximum CPU usage case is 5.6%. A. Experiment Design and Environment Two practical AGV use cases from the real industry appli- cations, harbor, and warehouse are chosen for the reliability validations of the proposed N-HiL framework. Harbor AGV is with a bigger size and higher moving speed, while the warehouse is with a smaller size and lower moving speed [31]. Seven robots are simulated for each case and the map with several obstacles is set for each case. The collision counts are logged corresponding to the timestamp for each test. And the collision rate P collision is calculated based on the total number of critical sections (CS) in the test. Fig. 6 shows the collision occurrences under 5G network conditions for the harbor AGV and warehouse AGV multirobot coordination cases. For the simulation using the statistical mod- eling that used a Bernoulli distribution, the packet loss rate (PLR) of 0, 0.1, and the packet latency of 0, 10, 50, and 100 ms Fig. 6. Collision occurrences under 5G network conditions for the harbor AGV and warehouse AGV multirobot coordination. Fig. 7. N-HiL simulation results compared with the statistical modeling for the harbor AGV safe coordination case. are set as the condition variables, respectively, to conduct the simulation for both habor AGV and warehouse AGV cases. For the simulation used in the proposed N-HiL framework, the wired network Ethernet, the wireless network Wi-Fi 6, and 5G are tested to evaluate the impacts of wireless on the safe multirobot coordination. Furthermore, the experiments under the Wi-Fi 6 network condition for safe multirobot coordination are tested in both Wi-Fi 6 @short and Wi-Fi 6 @long conditions. We clarified that the N-HiL simulation experiments were conducted in our laboratories with multiple functional areas for daily tests and experiments. Most of the laboratories are related to the industrial field with various industrial devices and equipment, such as metal control cabinets, motors, programmable logical controller automation equipment, heavy-duty autonomous mobile robots, manipulators, etc., where the environment is similar to the industrial field. B. Results and Analysis The experiment results of the harbor AGV case and the warehouse AGV case are listed in Figs. 7 and 8 . From the results of the experiment, using the statistical modeling methodology,

LV et al.: IMPACTS OF WIRELESS ON ROBOT CONTROL 9261 Fig. 9. Network performance analysis of Wi-Fi 6 and 5G network conditions recorded in the experiments. (a)–(d) are the PDF and CCDF curves of the network downstream latency; (e)–(h) are the PDF and CCDF of the network upstream latency. The desired joint position (the commands sent from the re- mote side) values are defined as J d , while the measured joint values (feedback of the real joint states) are defined as J r . J d and J r are time-series signals, which are logged corresponding to the system timestamp with a sample rate of 500 Hz. The timestamps between the peaks of J d and J r are compared and subtracted as the motion delay τ by (1), and the motion delay sequence τ of all periods in one test is given by (2) ˆ τ i = arg max t ∈ N i J r ( t ) − arg max t ∈ N i J d ( t ) (1) ˆ T = ˆ τ 1 ˆ τ 2 · · · ˆ τ P (2) where i = { 1 , 2 , 3 , . . . , P , } and N i denotes the i th period of the repeated motion. The cross-correlation between J d and J r is calculated follow- ing (3) and the average motion delay ¯ T is given by (4) ˆ R J d J r ( m ) = ∑ N − m − 1 n = 0 J d n ∗ + mJ r n ∗ , m ≥ 0 , ˆ R ∗ JrJd ( − m ) , m < 0 . (3) ¯ T = arg max m ∈ [ − T 2 , T 2 ] ˆ R JdJr ( m ) . (4) Fig. 10. Intuitive comparison for motion delay τ and peak amplitude error ε peak under Ethernet, Wi-Fi 6, and 5G network conditions, using raw data and the filtered data. In order to evaluate the remote-control performance of the teleoperation, the absolute errors between the peaks of desired joint values J d and the measure joint values J r are calculated as the peak amplitude error ε peak by (5), and the peak amplitude error sequence E peak is defined by (6) ˆ ε peak i = max t ∈ N i J r ( t ) − max t ∈ N i J d ( t ) (5) ˆ E peak = ˆ ε peak 1 ˆ ε peak 2 · · · ˆ ε peak P . (6) Further, the average peak amplitude error E peak are calculated as E peak = mean ˆ E peak = ∑ P 1 ˆ ε peak p P . (7) Corresponding to the network analysis, the above evaluation metrics of robot motion, the robot motion delay τ , and peak amplitude error ε peak , are also statistically analyzed toward the network latency L . More intuitive results of the motion delay τ and the peak amplitude error ε peak from the robot motion perspective is shown in Fig. 10 . The robot motion indicators under the Ethernet condition are chosen as the baseline, and the robot motion performance using the raw data and the filtered data is analyzed under four wireless network conditions. From Fig. 10(a) and (b) , the wireless network leads to a larger ε peak , which means decreasing the accuracy of robot control. Further, it is clear in Fig. 10(c) and (d) that the filter reduces the ε peak , improving the accuracy, but introducing an additional delay. In order to prove the validity of the simulated YuMi robot model in Robot Studio, we implemented the same control framework on the real YuMi robot using Ethernet and compared the key metrics of the control performance with the N-HiL simulation experiments. The results on the real YuMi robot show the average motion delay ¯ T of 0.3174 s and the average peak amplitude error E peak

9262 IEEE TRANSACTIONS ON INDUSTRIAL INFORMATICS, VOL. 19, NO. 9, SEPTEMBER 2023 Fig. 11. Statistical analysis of robot motion performance under the four wireless conditions compared to Ethernet, using the raw data: (a)–(d) are the PDF curves of the motion delay; (e) is the CCDF comparison of the motion delay; (f)–(i) are the PDF curves of the peak amplitude error; (j) is the CCDF comparison of the peak amplitude error. of 1.9755°, which is consistent with the N-HiL simulation results under the Ethernet condition. C. Impacts of Wireless to Control Performance For the motion delay τ of the robot, the PDF curves un- der the wireless network conditions compared to the Ethernet condition are shown in Fig. 11(a) – (d) . It is clearly shown that the motion delay under the 5G network condition and the three Wi-Fi 6 network conditions are almost at the same level while using the raw data. For the peak amplitude error ε peak , the corresponding PDF curves are listed in Fig. 11(f) – (i) . Fig. 12. Statistical analysis of robot motion performance under the four wireless conditions compared to Ethernet, interacting with the control strategy by deploying a filter on robot controller: (a)–(d) are the PDF curves of the motion delay; (e) is the CCDF comparison of the motion delay; (f)–(i) are the PDF curves of the peak amplitude error; and (j) is the CCDF comparison of the peak amplitude error. Furthermore, the CCDF curves of motion delay and peak am- plitude error are calculated and plotted in Fig. 11(e) and (j) . For the motion delay, the righter the CCDF curve, the greater the motion delay. For the peak amplitude error, the righter CCDF curve means the smaller error (negative values). In Fig. 11(e) , the motion delays under the Wi-Fi 6 condition are smaller than that under the 5G network condition, which shows the advantage of Wi-Fi 6 compared to 5G in this situation for the high response. In Fig. 11(j) , the errors under wireless network conditions are obviously larger compared to the wired Ethernet condition, and the 5G has smaller errors compared to all Wi-Fi 6 conditions.

9264 IEEE TRANSACTIONS ON INDUSTRIAL INFORMATICS, VOL. 19, NO. 9, SEPTEMBER 2023 network with a LHST profile. In this combination, the la- tency of the wireless network is longer but more bounded (i.e., smaller tail), which is beneficial to reducing control error,andthefilterincontrolcanfurtherreducethecontrol error caused by the network latency. Since it is only verified in limited test cases, these recommen- dations are preliminary. In the future, we will further investigate some data-driven and even reinforcement learning methods so that the robot control system can “negotiate” with the commu- nication infrastructure to achieve the best tradeoffs, especially when there are multiple objectives to be optimized. Moreover, current tuning on the robot local controller only involved a low-pass filter, which is just a small step towards the “control communication co-design.” In future article, we will employ more advanced control methods, such as the time-domain pas- sivity approaches in the N-HiL framework. VI. C ONCLUSION In this article, we propose a novel N-HiL simulation frame- work to effectively and efficiently investigate the impacts of wireless on robot control under real network conditions. Two representative robot applications, which are sensitive to com- munication latency, safe multirobot coordination, and human- motion-based teleoperation, are investigated using the proposed N-HiL framework. The Wi-Fi 6 and 5G wireless networks are tested and compared, drawing the network profiles of SHLT and LHST with the CCDF curves. Results show that the proposed N-HiL framework provides more trustworthy results compared to the statistical modeling method while eliminating the huge efforts and costs of building and testing the entire physical robot system in real life. From the statistical analysis of the wireless latency and the robot motion performances, a preliminary answer is drawn on how the wireless networks affect robot control. Furthermore, the interplay between the wireless and the control strategy is investigated by comparing the robot control performances under the wireless with different latency profiles. The experiment results will be helpful for both robot developers and communication developers to find the best tradeoffs between costs and performances. R EFERENCES [1] W. Dai, H. Nishi, V. Vyatkin, V. Huang, Y. Shi, and X. Guan, “Industrial edge computing: Enabling embedded intelligence,” IEEE Ind. Electron. Mag. , vol. 13, no. 4, pp. 48–56, Dec. 2019. [2] A. Wichmann, T. Korkmaz, and A. S. Tosun, “Robot control strategies for task allocation with connectivity constraints in wireless sensor and robot networks,” IEEE Trans. Mobile Comput. , vol. 17, no. 6, pp. 1429–1441, Jun. 2018. [3] J. Gielis and A. Prorok, “Improving 802.11p for delivery of safety-critical navigation information in robot-to-robot communication networks,” IEEE Commun. Mag. , vol. 59, no. 1, pp. 16–21, Jan. 2021. [4] D. Baumann, F. Mager, U. Wetzker, L. Thiele, M. Zimmerling, and S. Trimpe, “Wireless control for smart manufacturing: Recent approaches and open challenges,” Proc. IEEE , vol. 109, no. 4, pp. 441–467, Apr. 2021. [5] C. L. Fall et al., “A multimodal adaptive wireless control interface for people with upper-body disabilities,” IEEE Trans. Biomed. Circuits. Syst. , vol. 12, no. 3, pp. 564–575, Jun. 2018. [6] Y. Maddahi, S. Liao, W.-K. Fung, E. Hossain, and N. Sepehri, “Selec- tion of network parameters in wireless control of bilateral teleoperated manipulators,” IEEE Trans. Ind. Inform. , vol. 11, no. 6, pp. 1445–1456, Dec. 2015. [7] S. Manfredi, E. Natalizio, C. Pascariello, and N. R. Zema, “Stability and convergence of a message-loss-tolerant rendezvous algorithm for wireless networked robot systems,” IEEE Trans. Control Netw. Syst. , vol. 7, no. 3, pp. 1103–1114, Sep. 2020. [8] A. Kadian et al., “Sim2Real predictivity: Does evaluation in simulation predictreal-worldperformance?,” IEEERobot.Automat.Lett. ,vol.5,no.4, pp. 6670–6677, Oct. 2020. [9] Y.-C. Liu, T.-C. Lin, and M.-T. Lin, “Indirect/direct learning coverage control for wireless sensor and mobile robot networks,” IEEE Trans. Control Syst. Technol. , vol. 30, no. 1, pp. 202–217, Jan. 2022. [10] J. D. Herath and A. Seetharam, “A Markovian model for analyzing opportunistic request routing in wireless cache networks,” IEEE Trans. Veh. Technol. , vol. 68, no. 1, pp. 812–821, Jan. 2019. [11] H. Li and A. V. Savkin, “Wireless sensor network based navigation of micro flying robots in the industrial Internet of Things,” IEEE Trans. Ind. Inform. , vol. 14, no. 8, pp. 3524–3533, Aug. 2018. [12] A. Mannucci, L. Pallottino, and F. Pecora, “Provably safe multi-robot coordination with unreliable communication,” IEEE Robot. Automat. Lett. , vol. 4, no. 4, pp. 3232–3239, Oct. 2019. [13] V. Digani, M. A. Hsieh, L. Sabattini, and C. Secchi, “Coordination of multiple AGVs: A quadratic optimization method,” Auton. Robots , vol. 43, no. 3, pp. 539–555, 2019. [14] L. Zhou, L. Liu, M. Li, and J. Wang, “HLSP/UWAC: A hardware-in-loop simulation platform for underwater acoustic communication,” in Proc. IEEE 6th Int. Conf. Wireless Commun. Netw. Mobile Comput. , 2010, pp. 1–4. [15] Y. Chen and D. J. Braun, “Hardware-in-the-loop iterative optimal feed- back control without model-based future prediction,” IEEE Trans. Robot. , vol. 35, no. 6, pp. 1419–1434, Dec. 2019. [16] M. D. Stefano, H. Mishra, A. M. Giordano, R. Lampariello, and C. Ott, “A relative dynamics formulation for hardware-in-the-loop simulation of on-orbit robotic missions,” IEEE Robot Autom. Lett. , vol. 6, no. 2, pp. 3569–3576, Apr. 2021. [17] J. Kumar and E. Amutha, “Control and tracking of robotic manipulator using PID controller and hardware in loop simulation,” in Proc. IEEE Int. Conf. Commun. Netw. Technol. , 2014, pp. 1–3. [18] A. P. Lamping, J. N. Ouwerkerk, and K. Cohen, “Multi-UAV control and supervision with ROS,” in Proc. Aviation Technol., Integr., Oper. Conf. , 2018, Art. no. 4245. [19] C. Qi, F. Gao, X. Zhao, Q. Wang, and Q. Sun, “Distortion compensation for a robotic hardware-in-the-loop contact simulator,” IEEE Trans. Control Syst. Technol. , vol. 26, no. 4, pp. 1170–1179, Jul. 2018. [20] M. Aleksy, F. Dai, N. Enayati, P. Rost, and G. Pocovi, “Utilizing 5G in industrial robotic applications,” in Proc. IEEE 7th Int. Conf. Future Internet Things Cloud , 2019, pp. 278–284. [21] S. Hayat, E. Yanmaz, and R. Muzaffar, “Survey on unmanned aerial vehicle networks for civil applications: A communications view- point,” IEEE Commun. Surv. Tut. , vol. 18, no. 4, pp. 2624–2661, Oct.–Dec. 2016. [22] H. Lv, Z. Pang, M. Xiao, and G. Yang, “Hardware-in-the-loop simulation for evaluating communication impacts on the wireless-network-controlled robots,” in Proc. IEEE IECON 48th Annual Conf. Ind. Electron. Soc. , 2022, pp. 1–6, doi: 10.1109/IECON49645.2022.9968471 . [23] P. Park, S. C. Ergen, C. Fischione, C. Lu, and K. H. Johansson, “Wireless network design for control systems: A survey,” IEEE Commun. Surv. Tut. , vol. 20, no. 2, pp. 978–1013, Apr.–Jun. 2018. [24] H. Andreasson et al., “Autonomous transport vehicles: Where we are and what is missing,” IEEE Robot. Automat. Mag. , vol. 22, no. 1, pp. 64–75, Mar. 2015. [25] L. Zhang, H. Gao, and O. Kaynak, “Network-induced constraints in networked control systems—A survey,” IEEE Trans. Ind. Inform. , vol. 9, no. 1, pp. 403–416, Feb. 2013. [26] K. Sakurama and T. Sugie, “Generalized coordination of multi-robot systems,” Found. Trends Syst. Control , vol. 9, no. 1, pp. 1–170, 2021. [27] R. Baranitha, R. Mohajerpoor, and R. Rakkiyappan, “Bilateral teleop- eration of single-master multislave systems with semi-markovian jump stochastic interval time-varying delayed communication channels,” IEEE Trans. Cybern. , vol. 51, no. 1, pp. 247–257, Jan. 2021. [28] S. Chen, Z. Wang, A. Chakraborty, M. Klecka, G. Saunders, and J. Wen, “Robotic deep rolling with iterative learning motion and force control,” IEEE Robot. Automat. Lett. , vol. 5, no. 4, pp. 5581–5588, Oct. 2020. [29] F. Pecora, H. Andreasson, M. Mansouri, and V. Petkov, “A loosely-coupled approach for multi-robot coordination, motion planning and control,” in Proc. Int. Conf. Automat. Plan. Scheduling , 2018, pp. 485–493.

LV et al.: IMPACTS OF WIRELESS ON ROBOT CONTROL 9265 [30] H. Lv et al., “GuLiM: A hybrid motion mapping technique for teleoper- ation of medical assistive robot in combating the COVID-19 pandemic,” IEEE Trans. Med. Robot. Bionics , vol. 4, no. 1, pp. 106–117, Feb. 2022. [31] J. Fu, J. Zhang, G. Ding, S. Qin, and H. Jiang, “Determination of vehicle requirements of AGV system based on discrete event simulation and response surface methodology,” Proc. Inst. Mech. Eng. B J. Eng. Manuf. , vol. 235, no. 9, pp. 1425–1436, 2021. Honghao Lv (Student Member, IEEE) received the B.E. degree in mechanical engineering from the China University of Mining and Technology, Xuzhou, China, in 2018. He is currently working toward the Ph.D. degree in mechatronic engi- neering with the School of Mechanical Engi- neering, Zhejiang University, Hangzhou, China. He was with the School of Electrical Engineer- ing and Computer Science, KTH Royal Institute of Technology, Stockholm, Sweden, as a joint Ph.D. Student, from 2021 to 2022, under the financial support from the China Scholarship Council. He developed dual-arm robotic teleoperation systems based on the motion capture techniques. He worked on mobile robotics, multirobot coordination, and network hardware-in-the-loop simulation as a Guest Researcher with the ABB Corporate Research, Västerås, Sweden. His research interests include dual-arm robotic teleoperation, human–robot intelligent inter- face, and safe interaction. Zhibo Pang (Senior Member, IEEE) received MBA degree in innovation and growth from the University of Turku, Turku, Finland, in 2012, and the Ph.D. degree in electronic and computer systems from the KTH Royal Institute of Tech- nology, Stockholm, Sweden, in 2013. He is currently a Senior Principal Scientist with the ABB Corporate Research Sweden, Västerås, Sweden, and an Adjunct Professor with the University of Sydney, Camperdown, NSW, Australia, and the KTH Royal Institute of Technology. He works on enabling technologies in electronics, com- munication, computing, control, artificial intelligence, and robotics for Industry4.0 and Healthcare4.0. Dr. Pang is a Member of IEEE IES Industry Activities Committee and a Co-Chair of the Technical Committee on Industrial Informatics. He is an Associate Editor of IEEE TII, IEEE JBHI, and IEEE JESTIE. He was General Chair of IEEE ES2017, General Co-Chair of IEEE WFCS2021, and Invited Speaker at the Gordon Research Conference AHI2018. He was the recipient of the “Inventor of the Year Award” by ABB Corporate Research Sweden, three times in 2016, 2018, and 2021, respectively. Koushik Bhimavarapu (Member, IEEE) re- ceived the B.Tech. degree in electronics and communication engineering from the Jawahar- lal Nehru Technological University, Hyderabad, India, in 2020, and the M.Sc. degree in electri- cal engineering with emphasis on telecommu- nication systems from the Blekinge Institute of Technology, Karlskrona, Sweden, in 2022. He is currently working as an Associate Sci- entist on industrial automation with the ABB Corporate Research Center, Västerås, Sweden. His research interests include industrial wireless communication net- works, internet of things, cloud computing, low-latency communication, and time-sensitive networking. Geng Yang (Member, IEEE) received the B.E. and M.Sc. degrees in instrument science and engineering from the College of Biomedical Engineering and Instrument Science, Zhejiang University (ZJU), Hangzhou, China, in 2003 and 2006, respectively, and the Ph.D. degree in elec- tronic and computer systems from the Depart- ment of Electronic and Computer Systems, the Royal Institute of Technology (KTH), Stockholm, Sweden, in 2013. From 2013 to 2015, he was a Postdoctoral Researcher with the iPack VINN Excellence Center, School of Informa- tion and Communication Technology, KTH. He is currently a Professor with the School of Mechanical Engineering, ZJU. His research interests include bioinspired flexible and stretchable sensors for robot sensing and human–robot interaction, focusing on key enabling technologies for Human-Cyber-Physical-System and Healthcare 4.0. Dr. Yang is the Associate Editor of IEEE J OURNAL OF B IOMEDICAL AND H EALTH I NFORMATICS and Bio-Design and Manufacturing .