Toward Wireless Human-Machine Collaboration in the 6G Era

1 T o ward W ireless Human-Machine Collaboration in the 6G Era Gaoyang Pang, W anchun Liu, Senior Member , IEEE, Chentao Y ue, Daniel E. Que vedo, F ellow , IEEE, Karl H. Johansson, F ellow , IEEE, Branka V ucetic, Life F ellow , IEEE, and Y onghui Li, F ellow , IEEE Abstract —The next industrial re volution, Industry 5.0, will be driven by advanced technologies that f oster human-machine col- laboration (HMC). It will leverage human creativity , judgment, and dexterity with the machine’s str ength, precision, and speed to impro ve productivity , quality of life, and sustainability . Wir eless communications, empower ed by the emerging capabilities of sixth-generation (6G) wir eless networks, will play a central r ole in enabling flexible, scalable, and low-cost deployment of geograph- ically distrib uted HMC systems. In this article, we first intr oduce the generic architectur e and key components of wireless HMC (WHMC). W e then present the network topologies of WHMC and highlight impactful applications across various industry sectors. Driv en by the prospective applications, we elaborate on new perf ormance metrics that r esearchers and practitioners may consider during the exploration and implementation of WHMC and discuss new design methodologies. W e then summarize the communication requirements and review promising state-of-the- art technologies that can support WHMC. Finally , we present a proof-of-concept case study and identify several open challenges. I . I N T RO D U C T I O N T HE next industrial revolution, Industry 5.0, en visions a future of unprecedented human-machine collaboration (HMC) [1], where increasingly powerful and precise machines work in harmony with the unique creativity of humans. Unlike previous industrial paradigms, Industry 5.0 places humans back at the center of the production process, harnessing their creati vity , intuition, and problem-solving skills while delegating repetiti ve, monotonous, hazardous, and heavy-duty tasks to machines. This division of labor enables humans to focus on more stimulating and value-adding activities that are difficult to automate. In this paradigm, humans act as guides, quality assurance experts, and decision-makers, collaborating with machines either by teaching and cooperating side by side or by remotely controlling an actuator as an extension of their own body . T o enable timely and reliable collaborations between geographically distributed humans and machines, wireless HMC (WHMC) has emerged as a critical enabler , providing the required flexibility and scalability . The fifth generation (5G) network is enabling the “factory of the future” through digitization and HMC [2]. Nev er- theless, WHMC applications in dif ferent vertical industries G. Pang, W . Liu, C. Y ue, D. Que vedo, B. V ucetic, and Y . Li are with the School of Electrical and Computer Engineering, The Univ ersity of Sydney , Sydne y , NSW 2006, Australia (e-mail: { gaoyang.pang; wanchun.liu; chentao.yue; daniel.quev edo; branka.vucetic; yonghui.li } @sydney .edu.au). K. H. Johansson is with the Division of Decision and Control Systems, School of Electrical Engineering and Computer Science, and Digital Futures, KTH Royal Institute of T echnology , 100 44 Stockholm, Sweden (e-mail: kallej@kth.se). hav e unique generic requirements that go beyond ultra-reliable and low-latenc y communication (URLLC), such as global connectivity , high mobility , and low jitter . These requirements hav e not been fully met in 5G networks, whose capabilities remain insufficient to realize the full vision of WHMC. The sixth generation (6G) wireless network, ho wev er , will introduce advanced networking technologies and mark a new era of connecti vity and communication capabilities [3]. By seamlessly integrating terrestrial and non-terrestrial networks, 6G will extend accessibility for both humans and machines, facilitating remote control and monitoring of dynamic physical processes. Moreover , 6G will support the transformation of human-machine interfaces (HMIs), allowing real-time, high- bandwidth, multi-modal data exchange between humans and machines. This allows the encoding of human intuitive control and provides the remote human operator with an immersiv e experience during collaboration. Furthermore, artificial intelli- gence (AI) will be seamlessly embedded into 6G, empowering machine intelligence to unlock the full potential of WHMC. Despite the promise of 6G, reliable WHMC cannot be sim- ply achieved by connecting humans, machines, and networks without considering their interdependencies [4]. It requires a new networked system topology with coupled wireless human and automated control loops. Thus, WHMC lies at the inter- section of several engineering and scientific disciplines, such as human psychophysics, behavioral economics, mechanical design, communication, and control engineering. Con ventional siloed research efforts in these fields are not adequate since the dynamics and interdependencies among human, communi- cation, and control systems demand a more holistic approach. Fully harnessing the potential of WHMC requires the con ver - gence of these fields, which presents formidable challenges. Ke y barriers include the dev elopment of new fundamental mathematical models, performance metrics, and design tech- nologies consistent with a unified WHMC architecture. In this paper , we take an initial step toward addressing these challenges, aiming to bridge existing gaps for the large- scale rollout of WHMC. W e aim to motiv ate researchers to treat the wireless, human, and control systems as a united system, enabling fully integrated designs supporting a wide range of emerging applications. Specifically , we propose a general frame work of WHMC to establish a comprehensi ve understanding of the concept. W e dev elop new design method- ologies tailored to WHMC systems, addressing their unique challenges and constraints. W e then identify and discuss key design requirements and state-of-the-art technologies, pro- viding researchers with insights into the ev olving WHMC 2 research landscape. W e also present a proof-of-concept case study to demonstrate and validate the effecti veness of the proposed WHMC framew ork. Finally , we highlight key open challenges in WHMC, as well as broader issues extending beyond technology . I I . E N V I S I O N E D W H M C A. WHMC and Its Cor e Components An illustrative example of WHMC in advanced manufac- turing is collaborativ e engine assembly , as shown in Fig. 1. In such a scenario, actuators, such as robot arms, ex ecute the instructions of the human operator in a remote site (Site A), and the machine control commands generated by a centralized automated controller within the factory . Similarly , in collaborativ e dri ving for logistics operations outside the factory , vehicle actuators follow the maneuvering intentions of a remote human operator (Site B) while concurrently responding to steering and motion control commands issued by on-board autonomous controllers. These examples rev eal the following core components of WHMC. 1) Human: Remote operators make critical decisions and ex ecute collaborati ve tasks using their creativity , judgment, and dexterity . 2) Interface: HMIs deployed at remote sites capture human instructions and replicate the HMC environment, enabling the human operator to both perform and monitor the collaboration task effecti vely . 3) Sensor: Distributed sensors in the HMC en vironment (e.g., factory floor) capture real-time state information of the WHMC system, including v ariables to be controlled (e.g., positions), the status of actuators, task execution progress, and en vironmental dynamics. 4) Actuator: Actuators are distrib uted in the HMC en viron- ment to implement control actions issued by both the remote human operator and the autonomous machine controller . 5) Contr oller: This component provides the computational control logic for autonomous actuators. It integrates sensor data to generate machine control commands that ensure the collaboration goals are achiev ed. 6) Network: The communication netw ork bridges the remote site and the HMC en vironment, ensuring reliable and timely exchange of data and control signals between the two domains. T ogether, these core components form the generic WHMC architecture, which can be flexibly adapted to div erse Industry 5.0 applications. B. Network T opologies of WHMC In WHMC, the human operator , the machine controller, and the task are the three critical entities. Their wireless connection topologies are important in three aspects. 1) Agent autonomy: Agent autonomy describes the roles of a human operator and a machine controller in a giv en collaboration task, as shown in Fig. 2. In a human-machine symbiosis system, both agents’ actions are applied to the actuators simultaneously . In the machine-dominated system, the machine controller is supervised by the human operator , Fig. 1. Illustration of WHMC in adv anced manufacturing with abstracted key components. Fig. 2. WHMC topology at a glance of agent autonomy between human H, and machine M, for a collaboration task T . and only machine actions driv e the actuators. In the human- dominated system, the human operator makes decisions based on the recommendation from the machine controller , and only human actions driv e the actuators. Ignorance, aw areness, and trustworthiness are defined to describe the information av ailable to these two agents for decision-making. Without information exchange (Ignorance), the agents ignore each other . A wareness allo ws the observ ation of the state of their counterpart. When both the state and action of the counterpart are known, an agent can estimate its trust in collaboration. 2) Agent multiplicity: Analogous to multiple channel ac- cess, agent multiplicity describes human-actuator associations in multi-agent WHMC systems. In orthogonal human control, each human operator manages a dif ferent collaboration task. In non-orthogonal human control, at least one collaboration task is shared by multiple human operators who are not permitted to handle multiple tasks simultaneously . Non-orthogonal hu- man coordination control describes the case when the human operator is able to coordinate multiple tasks concurrently . Considering the capabilities of the machine agent in fast response and multi-tasking, the machine controller is assumed to be centralized and al ways capable of handling multiple tasks in the above three cases. 3) Agent geography: Agent geography describes the agent- task distance of various WHMC systems across different alti- tudes. Short-distance WHMC systems within a single wireless 3 cov erage area can be easily supported by advanced wire- less technology . For mid- and long-distance WHMC, packets often traverse fronthaul/midhaul/backhaul (x-haul) segments and metro/core transport, so end-to-end latency and jitter are shaped by transport switching, queuing, and path diversity , not only by the air interface. This makes transport design a first- order factor in sustaining stable WHMC across sites. C. Applications Empower ed by WHMC WHMC holds immense promise across v arious domains. One major area is adv anced manufacturing as presented in Section II-A. Besides, WHMC also enables synergistic inter- actions in following applications. 1) Agricultur e: Drone-assisted crop harvesting is a rep- resentativ e application of WHMC. Ground harvesting robots face challenges in safely grasping the objects randomly dis- tributed in three-dimensional (3D) agricultural en vironments due to blind spots. Grasping in visible objects requires coop- eration with external devices that provide the global position feedback of objects. WHMC enables a human operator to con- trol a drone to deliver a bird’ s eye vie w for object localization to close the vision gap of ground robots. 2) T ransportation: In autonomous driving, human assis- tance remains vital in unseen situations or challenging cases. A remote human operator may support the self-driving vehicle’ s decisions at an intersection. Through a dri ving-simulation HMI, the operator changes the vehicle’ s steering direction based on the dynamic traffic conditions. The local machine controller inside the vehicle maintains a minimal safe distance to pedestrians and other vehicles. 3) Healthcar e: T elesurgery is a cutting-edge application of WHMC, which enables highly qualified and experienced surgeons to perform critical sur geries remotely . The surgeons’ actions are captured and transmitted by HMIs, and then used to control a robotic device in the operating room to replicate the surgeons’ operations. The local machine controller inside the operating room provides assistive operations, such as bleeding control, blood/oxygen circulation, and handover of sur gical instruments. D. P erformance Metrics of WHMC In conv entional cyber-physical control systems (CPCSs), the performance is characterized by the control cost, which is defined as a function of plant state and control inputs. The cost function design depends on the specific goals and constraints of the CPCS. For example, in production-line automation, the cost function quantifies mission completion time. In autonomous vehicle control, the cost function is related to the safety margin to obstacles. In contrast to tra- ditional CPCSs, WHMC integrates human operators, wireless communication networks, and control systems into a tightly coupled frame work characterized by strong dynamics and interdependencies. This integration results in a novel topology featuring intertwined wireless human and autonomous control loops. Consequently , WHMC in volv es three interconnected dynamic layers: 1) human state dynamics influencing the operator’ s control decisions, 2) control system dynamics gov- erning machine behavior , and 3) wireless channel dynamics affecting communication reliability and latency . Therefore, the performance metrics of WHMC are inherently more complex and less tractable than those of conv entional CPCSs. WHMC introduces a multitude of ne w performance metrics for its design and implementation. In WHMC, the quality of collaboration (QoC) is multi-goal-oriented and depends on div erse factors related to the three-lev el dynamics with the full consideration of task performance and human wellness. W e link QoC to measurable metrics across the three coupled layers. At the task layer, QoC reflects control/mission out- comes (e.g., tracking error , completion time). At the human layer , QoC captures perceptual fluency and workload (e.g., delay/jitter-induced discontinuity , interaction smoothness, in- tervention frequency). At the network layer , QoC is primarily shaped by latency , jitter , reliability , and sustained capacity av ailability . For instance, immersi ve teleoperation is typically jitter-limited (fluency), while supervisory control is more outage- and av ailability-limited (safety and continuity). E. Design Methodologies of WHMC Once QoC performance metrics are well established, new design methodologies are necessary to optimize WHMC. Human-type communications (HTC) and machine-type com- munications (MTC) will naturally coexist in WHMC. They hav e significantly different communication requirements, data behaviors, and data structures, which should be considered during design. Herein, we propose a WHMC optimization methodology , as shown in Fig. 3. 1) QoC-awar e codesign of HTC and MTC: QoC-aw are communications are essential to WHMC to guarantee application-lev el performance. Their design in volves the co- ordination and synchronization of HTC and MTC for the human control loop and the machine control loop, respecti vely , which must be jointly optimized to guarantee QoC. Given the multi-goal-oriented nature of QoC, the distinct communication requirements of HTC and MTC are supported by tailoring wireless links, while mitigating their inter-link interference. The design of each link is adaptive to dynamic human control behaviors, collaborative control tasks, and time-varying wire- less en vironments using historical data, such as channel/human states and human/machine control actions. The design also requires a comprehensive in vestigation of the relationship be- tween QoC and the communication protocol, frame or packet structure, and transmission strategy . In addition, power and bandwidth allocation among multiple WHMC systems must reflect heterogeneous QoC-aware communication demands. 2) QoC-oriented contr ol-communication codesign: Unlike traditional communication-oriented systems, a WHMC system emphasizes the goal-oriented codesign of communication and control systems with explicit awareness of human in volvement. Most existing communication architectures are agnostic to collaborativ e control objecti ves. Their design focuses primarily on maximizing communication performance (e.g., throughput, reliability) rather than optimizing control performance or hu- man satisfaction. Understanding collaborative control dynam- ics and human characteristics offers an untapped opportunity 4 Fig. 3. The design methodology for WHMC, where multi-goal-oriented QoC bridges the fundamental design and the desired collaboration goals. to relax the stringent communication requirements. The key is to prioritize transmission packets based on their contribution to QoC. This approach contrasts sharply with current systems (e.g., 5G), where all packets are treated with equal priority and only over -the-air transmission errors are considered. 3) F eedback via re war d shaping and incentive mechanism: Optimizing WHMC performance requires instantaneous feed- back that captures the synergistic interplay among humans, machine controllers, and wireless networks. When human control actions are not observable by machines, i.e., no trustworthiness in Fig. 2, the resulting human inputs can unintentionally disrupt machine operations. T o address this, the desired QoC objectives can be embedded into the joint design of these entities, enabling a unified feedback framew ork. QoC can be interpreted as instantaneous feedback signals (i.e., rew ards for machines and incentiv es for humans) deriv ed from the combined ef fects of wireless communication design, human control behavior , and machine control strategies on WHMC performance. Furthermore, WHMC naturally supports human–machine co-adaptation and co-ev olution, allo wing both agents to estimate and anticipate each other’ s performance, thereby enhancing overall collaboration effecti veness. I I I . D E S I G N R E Q U I R E M E N T S A N D B U I L D I N G B L O C K S A. K ey Communication Design Considerations W ireless communication, as a key entity , plays a piv otal role in supporting v arious WHMC applications. The detailed ser- vice requirements for CPCSs and for professional multimedia (video, imaging, and audio) applications hav e been extensiv ely summarized in 3GPP TS 22.104 and 3GPP TS 22.263, re- spectiv ely . Howe ver , defining service requirements for diverse WHMC applications is significantly more challenging. Con ventional 5G service categories, i.e., URLLC, enhanced mobile broadband (eMBB), massiv e machine-type commu- nications (mMTC), address low latency , high throughput, and massi ve connecti vity separately . WHMC is generationally agnostic and can be instantiated on current 5G networks (e.g., 5G standalone with slicing) for near-term deployments, while benefiting from future 6G enhancements (e.g. high-rate sensing, AI-native slicing, and extended coverage) for broader WHMC applications. Here, we identify fiv e key considera- tions for WHMC communication. They form the fundamental communication requirements that enable effecti ve WHMC. 1) Accessibility: Accessibility is the ability of a WHMC system to establish and maintain connectivity wherever hu- mans and machines operate. As human acti vity increasingly extends to space, aerial, mariti me, and remote regions, WHMC is essential in seamless collaboration. W ireless communica- tions are critical to bridging these physical distances and ensuring global connectivity . Long-distance WHMC commu- nications across different altitudes challenge 5G terrestrial net- works, which cannot provide global connectivity . For instance, remote industrial or disaster-response missions may require connectivity spanning hundreds of kilometers via drones or satellites, far beyond terrestrial 5G capabilities. 2) T ranspar ency: T ransparency describes how well the human operator perceiv es the environment and the slaved machine as an extension of the self, and how well the slav ed machine and human operator comprehend each other . Achieving this requires URLLC with lo w-jitter , high-fidelity and real-time communications, especially for geographically separated entities, whereby transmissions may traverse be- yond the wireless access segment. WHMC requires transport- aware URLLC. Flexible x-haul combined with slice-aware orchestration can reserve deterministic paths for control loops while scaling XR/multimedia streams. In practice, immersive mission-critical teleoperation (e.g., telesurgery or remote robot control) demands end-to-end latencies of a few milliseconds with jitter below 1 ms, exceeding current 5G URLLC perfor- mance. 3) Scalability: Scalability determines a WHMC system’ s ability to accommodate vast numbers of agents and dynami- cally on-board ne w participants. Network topology may e volv e rapidly as agents join or leav e the collaboration. For instance, in a rescue operation in volving human-assisted drone swarms through VR/AR interfaces, hundreds of drones may require si- multaneous wireless control loops for collision avoidance and intra-swarm visual coordination, i.e., non-orthogonal human coordination control in Section II-B2. Large-scale WHMC will demand wireless networks capable of integrating both high data rates and massive connectivity , going beyond 5G’ s separately enhanced eMBB and mMTC categories. 4) Resilience: Resilience is vital for WHMC applications in volving high-mobility agents, such as remote driving and autonomous fleet management, which impose stringent con- tinuity and reliability requirements. Frequent handovers in high-speed scenarios can cause service interruptions, under- mining control stability . For example, a 120 km/h vehi- cle may experience a handov er e very few seconds; e ven a millisecond-scale disruption can destabilize the control loop. Maintaining reliable URLLC during such transitions demands near-seamless mobility management, exceeding the resilience currently achiev able with 5G networks. 5) Sustainability: Since many de vices in WHMC, espe- cially sensors in Internet of Things (IoT), wearable interfaces, 5 mobile manipulators, and even satellites, are battery-powered, enhancing energy ef ficiency is important. Inefficient energy management and transmission rate control can lead to energy starvation, which may disable core components of WHMC in emergencies. For example, industrial IoT sensors may run for years on a single battery , and wearable HMIs often operate on milliwatt-le vel power budgets. Meeting these longevity requirements calls for ultra-energy-efficient communication strategies, which is a key focus of green 6G design. B. Pr omising Communication T echnologies The aforementioned gaps call for the dev elopment of ad- vanced and deeply integrated communication technologies. Sev eral emerging 6G technologies and multiple complemen- tary technologies can be jointly applied to support WHMC. 1) Human-bond communications and beyond: Immersive user experiences enhance human control-oriented decision- making by providing realistic feedback, improving situational awareness, reducing cogniti ve load, and increasing engage- ment. Human bond communication (HBC) aspires to achiev e this goal by integrating, digitizing, transmitting, and replicat- ing the five human senses [5]. Effecti ve communication should also capture human well-being and functioning to build trust and foster reliable collaboration. This requires the communica- tion of human control intentions, behavioral patterns, psycho- logical states (e.g., emotions), physical states (e.g., fatigue), and physiological states (e.g., heart rate). Dev eloping this communication system requires communication technologies beyond HBC, such as holographic communication [6] and affecti ve communication [7]. 2) Semantic and goal-oriented communication: WHMC in- volv es multi-modal human interaction, including text, speech, image, and body motion/gesture, requiring devices with similar interaction capabilities. This motiv ates the development of semantic communication systems, which can extract and trans- mit the semantic meaning of complex information rather than its raw bits in a noisy channel [8]. Semantic communication relies primarily on a shared kno wledge base that is understand- able to both human operators and machines. Communication efficienc y and reliability can be further improved by exploiting shared knowledge for semantic inference. Recent advances in AI technologies hav e boosted the potential of semantic communications in future 6G networks. 3) Inte grated sensing and communication: The prolifera- tion of wireless devices in large-scale WHMC can lead to sev ere spectrum congestion. Integrated sensing and commu- nication (ISA C) is promising to address this issue [9]. ISA C allows hardware and spectrum sharing between sensing and communication functions, which reduces the ov erall hardware cost and improv es spectrum efficiency . Empowered by ISA C, location-aware communication services, such as mobility man- agement, become more accessible, increasing connectivity for WHMC. Combined with IoT , ISAC also supports human activity recognition, vital signal monitoring, and spatial-a ware computing, advancing HMIs towards HBC and beyond. 4) Other communication technologies: T erahertz (THz) communication provides terabit-per-second (Tbps) throughput [10] for data-intensi ve WHMC b ut is vulnerable to propagation en vironments. Reconfigurable intelligent surfaces (RISs) can shape reflections to control propagation environments [1]. Be- sides, embedding RIS nodes in non-terrestrial networks forms space–air–ground–sea integrated networks (SAGSIN), provid- ing flexible coverage [11]. Howe ver , massive HTC–MTC coexistence in SA GSIN strains signaling, interference, and energy . Rate-splitting multiple access (RSMA) mitigates this by splitting common/pri vate streams, enhancing spatial multi- plexing, connectivity , ef ficiency , and reliability [12]. C. Be yond Communication T echnologies While communication technologies are crucial, they alone cannot enable WHMC systems. Advancing this paradigm demands attention to societal and priv acy issues, as well as broader enabling technologies. Sensors, actuators, and HMIs serve as key interfacing technologies for seamless human–machine–en vironment interaction. Fle xible, stretchable sensors made from soft materials can conform to the body , robots, or surroundings for continuous contextual monitoring [13]. Collaborati ve robots (cobots) provide adaptable, portable, and cost-effecti ve actuation for collaborativ e, teleoperated, or autonomous WHMC tasks. Next-generation HMIs, such as XR, motion-capture, tactile, and brain–computer interfaces [7], offer higher control freedom than con ventional HMIs. Furthermore, WHMC relies on a programmable control plane that fuses communication and computation through SDN, NFV , and cloud–edge–end orchestration [14]. In current deployments, network slicing in the 5G standalone core can provide traffic isolation between (i) latency-critical control loops and (ii) high-rate sensing streams, while cross-domain orchestration binds RAN, transport, and edge resources into a per-task service chain. In addition, an O-RAN-style radio in- telligent controller illustrates how near-real-time RAN policies (e.g., scheduling) can be adapted for QoC-aware operation. Adaptiv e functional splits provide a practical knob to trade QoC vs. latency by shifting processing closer to the radio [15]. Finally , effecti ve control strategies remain the backbone of WHMC, which ensures safe and efficient cooperation under network imperfections. Shared autonomy frame work allocates control authority based on confidence, workload, and trust; haptic control and teleportation maintain stability via passi vity- based compensation and predictive feedback; communication- aware control (e.g., e vent/self-triggered and anytime control) prioritizes transmissions by criticality; and learning-enabled control (e.g., safe reinforcement learning) adapts to en viron- mental and human uncertainties. Integrating these control, computing, interfacing, and digital-twin technologies ensures that WHMC systems remain stable, responsive, and efficient. D. Deployment Roadmap In the near term, WHMC is feasible in factory settings using priv ate 5G/5G-Advanced with edge computing and traffic isolation for control loops. Early 6G deployments are expected to broaden WHMC by improving end-to-end latency/jitter control, integrating AI-assisted orchestration, and enabling richer XR-based interfaces. Later 6G ev olution can extend 6 WHMC to wide-area and multi-altitude scenarios by coupling non-terrestrial connectivity with transport-aw are slicing across RAN and optical x-haul. From a maturity perspectiv e, slicing and edge-enabled ser- vice chaining are av ailable in 5G standalone systems and can be used to separate control-loop traf fic from perception streams. In early 6G, tighter and more automated cross-domain orchestration, improved deterministic performance, and better support for adv anced XR are expected to become mainstream. T echnologies, such as semantic/goal-oriented communication, ISA C, and THz/RIS-based ubiquitous cov erage, remain sensi- tiv e to en vironments and standardization uncertainty , and are therefore positioned as longer-term accelerators. I V . I L L U S T R A T I V E C A S E S T U D Y A. Experiment Setup Motiv ated by typical manufacturing scenarios, we examine a wireless cart-pole control system, as shown in Fig. 4. It is a classic nonlinear and unstable control problem. The machine linearizes the cart-pole dynamics around the equilibrium point and uses a standard linear quadratic regulator to apply forces to the cart and balance the pole. A challenging scenario is introduced by adding a dynamic weight to the cart, which the machine cannot detect or compensate for . Howe ver , a human operator, who can observe this disturbance, interv enes by pressing a key to remov e the weight, prev enting system failure. The cart-pole system, machine controller , and human operator are spatially separated, communicating ov er a wire- less network. Control signals and system states are exchanged using short-packet transmissions to ensure low latency . The av erage channel gains follow the free-space path loss model. The time-varying channel gains are generated from Rayleigh fading channel models with a unit mean. Communication parameters are summarized in T able I. The mass of the cart, pole, and weight are 10 kg, 4 kg, and 5 kg, respecti vely . The pole length is 4 m. The initial pole angle is π /6. B. Results and Discussion The control performance is ev aluated at each time step using a quadratic cost function on the deviation of the pole angle, reflecting the system’ s ability to maintain balance. A lower cost indicates better control. Fig. 5(a) compares the accumulated control cost across three scenarios: machine-only control, human-only control, and WHMC (i.e., a machine- dominated system in Fig. 2). The WHMC scenario sho ws a slower increase in cost and reaches a significantly lower steady-state value, demonstrating that collaborativ e control yields superior performance. Fig. 5(b) presents the impact of signal-to-noise ratio (SNR) on control performance in the WHMC case. When the human operator is activ ely engaged, higher SNRs lead to improved control outcomes. Howe ver , when the operator is distracted (e.g., due to f atigue), increasing SNR does not enhance performance. This case study is a minimal proxy for WHMC, demonstrating the complementar- ity of human resilience and machine speed. The autonomous controller stabilizes fast dynamics, while the human operator provides sparse interventions when unmodeled disturbances Fig. 4. The wireless cart-pole system for the case study of WHMC. Fig. 5. Evaluation of WHMC: (a) Impacts of decision-makers on the control performance. (b) Impacts of communications on the control performance. T ABLE I S U MM A RY O F C OM M U N IC A T I ON P A R A M ET E R S O F TH E C A S E S T U DY Communication parameters V alue Free-space path loss model V alue Code rate [bps] 2 Antenna gain 4 Packet length [symbols] 500 Carrier frequency [MHz] 915 Min transmit power [dBm] 20 Distances to plant [m] 45-50 Background noise power [dBm] -70 Path loss exponent 2.9 occur . Communication quality affects performance. Effecti ve collaboration requires integrated optimization across control, communication, and human engagement, motiv ating the def- inition of QoC to jointly capture control outcomes, network key performance indicators (KPIs), and human state. V . O P E N C H A L L E N G E S A N D F U T U R E D I R E C T I O N S A. Fundamental Mathematical Modeling At present, there is a lack of robust theories and tools for designing flexible and scalable WHMC systems. Unlike con- ventional control systems in structured and repetiti ve settings, WHMC systems inv olve dynamic interactions between human and machine via wireless communication. This requires funda- mentally new mathematical models and analytical frameworks that can manage system complexity , capture essential collabo- rativ e human–machine dynamics, and remain mathematically tractable. Future research must focus on collaborative task abstraction, human behavior modeling, and wireless channel characterization in complex en vironments. B. T ractable P erformance and Stability Analysis There is a mismatch between traditional communication KPIs and the QoC requirements in WHMC. Pack ets differ in importance, deadlines, and reliability , so KPI gains alone 7 may waste resources without improving QoC. A tractable framew ork must jointly capture QoC and packet-lev el dynam- ics while linking them to control outcomes. Equally critical is stability . Transmission errors and delays can destabilize closed loops, yet formalizing stability is hard without faithful human behavior models. A unified framework that captures div erse performance goals and human effort remains elusive. Progress demands an interdisciplinary approach to characterize fundamental limits of WHMC. C. Optimal Design of Lar ge-Scale Systems The first step tow ard optimal design of large-scale WHMC systems is to formulate problems that account for the three- layer dynamics—control, communication, and human behav- ior . Howe ver , deriving closed-form expressions for objecti ve functions and stability constraints is challenging due to the analytical intractability of many performance metrics, leading to NP-hard mixed-integer nonlinear programs. Y et, current AI- based approaches typically treat machine controllers, human operators, and wireless networks in isolation, limiting their ef- fectiv eness. Dev eloping integrated AI framew orks for div erse WHMC applications is vital for scalability through improv ed model adaptation, inference efficienc y , and interpretability . D. Critical Issues be yond T echnology WHMC also brings critical considerations beyond technol- ogy . The deployment of wireless communication systems is closely tied to upstream industries, which may lag behind the rapid adv ancements in communication theories. Ethical and societal considerations are increasingly relev ant (e.g., potential health risks from dense network deployment and high-frequency transmissions), as well as priv acy and surveil- lance issues in a sensor-rich en vironment. It is crucial to balance technological innovations with socio-economic factors to ensure global accessibility , affordability , and trust. V I . C O N C L U S I O N This paper presents a forward-looking vision for wireless- empowered HMC in the 6G era. W e highlighted the distinct performance metrics of WHMC compared to traditional com- munication systems, emphasizing the importance of human satisfaction and task-le vel outcomes. A design methodology tailored to WHMC has been proposed, recognizing the need to integrate knowledge from control, communication, and human factors. Our proof-of-concept study demonstrates the tangible benefits of WHMC, and we have outlined ke y future direc- tions, including scalable system design, tractable performance and stability analysis, and ethical deployment. R E F E R E N C E S [1] M. Noor-A-Rahim et al., “T ow ard Industry 5.0: Intelligent reflecting surface in smart manufacturing, ” IEEE Commun. Mag. , vol. 60, no. 10, pp. 72–78, 2022. [2] R. Sabella et al., “Robotics and industrial automation enabled by 5G - Ericsson, ” Ericsson T echnol. Rev . , no. 2, pp. 1–12, 2018. [3] C. -X. W ang et al., “On the road to 6G: V isions, requirements, key technologies, and testbeds, ” IEEE Commun. Surveys Tuts. , vol. 25, no. 2, pp. 905–974, 2023. [4] G. Pang, W . Liu, D. Niyato, D. Quev edo, B. V ucetic and Y . Li, “W ireless human-machine collaboration in Industry 5.0, ” IEEE T rans. W irel. Commun. , accepted, 2025. [5] T . Iftikhar , H. A. Khattak, Z. Ameer, M. A. Shah, F . F . Qureshi and M. Z. Shakir, “Human bond communications: Architectures, challenges, and possibilities, ” IEEE Commun. Mag. , vol. 57, no. 2, pp. 19–25, 2019. [6] T . Gong et al., “Holographic MIMO communications: Theoretical foun- dations, enabling technologies, and future directions, ” IEEE Commun. Surveys T uts. , vol. 26, no. 1, pp. 196–257, 2023. [7] D. Wu, B. -L. Lu, B. Hu and Z. Zeng, “ Af fectiv e brain–computer interfaces (aBCIs): A tutorial, ” Proc. IEEE , vol. 111, no. 10, pp. 1314– 1332, 2023. [8] W . Y ang et al., “Semantic communications for future Internet: Fundamen- tals, applications, and challenges, ” IEEE Commun. Surve ys T uts. , vol. 25, no. 1, pp. 213–250, 2022. [9] F . Liu et al., “Integrated sensing and communications: T oward dual- functional wireless networks for 6G and beyond, ” IEEE J. Sel. Areas Commun. , vol. 40, no. 6, pp. 1728–1767, 2022. [10] Z. Chen et al., “T erahertz wireless communications for 2030 and beyond: A cutting-edge frontier, ” IEEE Commun. Mag. , vol. 59, no. 11, pp. 66–72, 2021. [11] M. M. Azari et al., “Evolution of non-terrestrial networks from 5G to 6G: A survey , ” IEEE Commun. Surveys T uts. , vol. 24, no. 4, pp. 2633– 2672, 2022. [12] B. Clerckx et al., “ A primer on rate-splitting multiple access: T utorial, myths, and frequently asked questions, ” IEEE J. Sel. Areas Commun. , vol. 41, no. 5, pp. 1265–1308, 2023. [13] G. Pang, G. Y ang and Z. Pang, “Re view of robot skin: A potential enabler for safe collaboration, immersive teleoperation, and affecti ve interaction of future collaborativ e robots, ” IEEE T rans. Med. Rob . Bio. , vol. 3, no. 3, pp. 681–700, Aug. 2021. [14] S. Duan et al., “Distributed artificial intelligence empowered by end- edge-cloud computing: A survey , ” IEEE Commun. Surve ys T uts. , vol. 25, no. 1, pp. 591–624, 2022. [15] I. T omkos et al., ”The “X-Factor” of 6G networks: Optical transport empowering 6G innovations, ” IT Pr of. , vol. 26, no. 1, pp. 32–39, 2024. Gaoyang Pang [M] is a Post-Doctoral Research Associate with the School of Electrical and Computer Engineering (ECE), The University of Sydney (USyd), Australia. W anchun Liu [SM] is a Senior Lecture and an ARC DECRA Fello w at USyd. Her research interests are networked robotics, industrial IoT , and HMC. Chentao Y ue [M] is an ARC DECRA Fellow at USyd. His research interests are in the areas of coding theory and semantic communications. Daniel E. Quevedo [F] is a Professor of cyber-physical systems at the ECE, USyd. His research interests are in network ed control systems, control of power converters, and cyber-physical systems security . Karl H. Johansson [F] is Swedish Research Council Distinguished Professor in electrical engineering and computer science with KTH Royal Institute of T echnology in Sweden and Founding Director of Digital Futures. Branka V ucetic [LF] is an Australian Laureate Fellow , a Professor of T elecommunications, and Director of the Centre for IoT and T elecommunica- tions at USyd. She is a Fellow of the Australian Academy of T echnological Sciences and Engineering and the Australian Academy of Science. Y onghui Li [F] is an ARC Industry Laureate Fellow , a Professor of T elecom- munications, and Director of W ireless Engineering Laboratory at the ECE, USyd. His research interests are millimeter wav e communications, machine- to-machine communications, and coding techniques.