1、Preliminary Study of Advanced Technologies towards 6G Era:QITs 2023Executive SummaryOver the past year,the latest developments in the 6th generation(6G)communication systems research have been reported all over the world.6G is emerging as an important direction for research and development in the fi

2、eld of communications.As one enabling technology for 6G,quantum information technologies(QITs)continue to attract interest from academia and industry due to the expected information processing capabilities beyond their classical counterparts.In the 6G era,the importance of cybersecurity in mobile co

3、mmunications is expected to rise exponentially.Chapter 2 focuses on quantum secure communication aiming at safeguarding critical information by applying quantum mechanisms.The introduction starts with key technologies including quantum key distribution(QKD)and quantum random number generator(QRNG),f

4、ollowed by state-of-the-art standardization activities for quantum key distribution networks(QKDN)all over the world.Regarding the implications for 6G,China Unicom has built a quantum key cloud platform in Xiongan New Area and carried out a wide range of quantum encryption technology research and ap

5、plication demonstrations.Thus,two of the representative application scenarios,namely,quantum encrypted call and quantum public network cluster intercom will also be introduced in this chapter.To satisfy the dramatically increased communication system performance and rich diversity of innovative serv

6、ices expected by 6G,the emerging quantum machine learning(QML)has attracted significant attention due to its information processing paradigm by combining the established benefits of quantum mechanisms and machine learning.Considering quantum-enhanced reinforcement learning has the potential to revol

7、utionize the field of artificial intelligence(AI),chapter 3 gets insight into the research of quantum-enhanced machine learning by analyzing representative works in detail from two aspects.One is to study how to speed up the reinforcement learning(RL)by applying the quantum mechanism.The other shows

8、 an experiment performed to reconstruct an unknown photonic quantum state with a limited amount of copies,for which the performance in terms of fidelities can be improved with the assistance of the semi-quantum reinforcement learning approach.前 言过去一年中，有关第六代通信系统(6G)研究的最新进展在全球范围内被广泛报道。6G 正逐渐成为通信领域的重要研

9、发方向。作为 6G 的使能技术之一，量子信息技术(QITs)因其超越经典信息技术的信息处理能力预期，在学术界和工业界开始受到青睐。在 6G 时代，网络安全在移动通信中的重要性预计将呈指数级增长。本白皮书在第 2 章重点介绍旨在通过应用量子机制保护关键信息的量子安全通信。该章节首先介绍了量子密钥分发(QKD)和量子随机数生成器(QRNG)等关键技术，接着全面回顾了全球量子密钥分发网络(QKDN)的最新标准化活动。关于量子信息技术的 6G 应用，中国联通在雄安新区建设量子密钥云平台，并开展了广泛的量子加密技术研究和应用示范。本章也将分享了其中两个具有代表性的应用场景，即，量子加密通话和量子公网集群

10、对讲。为满足 6G 所期望的大幅提高的通信系统性能和丰富多样的创新服务，新兴的量子机器学习(QML)因其信息处理范式融合了量子机制和机器学习的技术优势而备受关注。考虑到量子增强强化学习具有彻底改变人工智能(AI)领域的潜力，本白皮书第 3 章从两个方面分析了相关代表性工作，以探索量子增强机器学习的研究。其一研究如何通过应用量子方法来加速强化学习(RL)。其二展示了用有限数量的副本重建未知光子量子态的实验，在半量子强化学习方法的帮助下，可以提高保真度方面的性能。IntroductionQuantum Secure CommunicationKey Technologies Overall Pic

11、tureQuantum Key DistributionQuantum Random Number GeneratorStandardization Activities for QKDNITU-TITU-T Study Group 13ITU-T Study Group 17ITU-T Study Group 11ITU-T FG-QIT4NETSI ISG-QKDISO/IEC JTC 1/SC 27Implications for 6GApplication Scenario 1:Quantum encrypted callApplication Scenario 2:Quantum p

12、ublic network cluster intercomQuantum-Enhanced Reinforcement LearningReconstruction of a Photonic Qubit State with Reinforcement Learning01 02 04 0505 050507090961617181921Table of ContentsExecutive Summary 前言 1 22.1 2.1.12.1.22.1.32.22.2.12.2.1.12.2.1.22.2.1.32.2.1.42.2.22.2.32.32.3.12.3

13、.23.13.2Reference Acknowledgement24 25Quantum Machine Learning(QML)3191.IntroductionThe scope of this annually revised white paper is to introduce quantum information technologies(QITs)with the aim of taking advantage of their powerful information processing capabilities to fulfill stringent demands

14、 of communication and computing envisaged by 6G systems.The version of 2023 will further introduce two benefits expected from QITs to communication and computing systems,i.e.,quantum secure communication and quantum machine learning.Chapter 2.Quantum Secure CommunicationIn 6G era,the importance of c

15、ybersecurity in mobile communications is expected to rise exponentially.Chapter 2 focuses on quantum secure communication aiming at safeguarding critical information by applying quantum mechanisms.Chapter 2 starts with key technologies including quantum key distribution(QKD)and quantum random number

16、 generator(QRNG),followed by state-of-the-art standardization activities for quantum key distribution networks(QKDN)all over the world.Regarding the implications for 6G,two novel application scenarios are introduced,namely,quantum encrypted call and quantum public network cluster intercom.Chapter 3.

17、Quantum Machine Learning(QML)To satisfy the dramatically increased communication system performance and rich diversity of innovative services expected by 6G,the emerging QML has attracted significant attention due to its information processing paradigm by combining the established benefits of quantu

18、m mechanisms and machine learning.Considering quantum-enhanced reinforcement learning has the potential to revolutionize the field of artificial intelligence(AI),chapter 3 gets insight into the research of quantum-enhanced machine learning by analyzing representative works in detail from two aspects

19、.One is to study how to speed up reinforcement learning by applying the quantum approach.The other shows an experiment performed to reconstruct an unknown photonic quantum state with a limited amount of copies,for which the performance in terms of fidelities can be improved with the assistance of th

20、e semi-quantum reinforcement learning approach.-04-Preliminary Study of Advanced Technologies towards 6G Era:QITs 20232.Quantum Secure Communication2.1 Key Technologies2.1.1 Overall PictureQuantum communication is a new and rapidly developing communication technology that has become a hot topic in f

21、rontier science and technology,the security of which is guaranteed by quantum mechanics.Quantum key distribution(QKD)is the most maturely developed quantum communication technology,using quantum superposition states or entanglement to distribute qubits,with unconditional security at the theoretical

22、level.The Quantum Random Number Generator(QRNG)is known to the general public as a relatively mature product with the help of QKD.QRNG is a system for generating true random numbers based on the principles of quantum physics or quantum effects and has important applications in areas such as practica

23、l quantum communication systems.2.1.2 Quantum Key DistributionTo date,there are many different protocols for QKD,all of them can be divided into two main categories:prepare-and-measure(PM)protocol,entanglement-based(EB)protocol.For the former,the transmitter generates a random bit sequence and then

24、encodes them on quantum states,which are subsequently sent to receiver to measure.For the latter,one party prepares enough entangled states for distributing over the channel to the other party,and then purifies and measures the entangled states to obtain the secure keys.Because the PM scheme is easi

25、er to implement,it is often used to structure practical systems,in which the Prepared by both parties and Measured by center scheme is mostly adopted.In addition,it can be further divided into two types:DV-QKD and CV-QKD.The successful demonstration of the BB84 QKD in 1989 proved the theoretical unc

26、onditional security of QKD.Although QKD is theoretically unconditionally secure,imperfections in the practical devices can expose the system to threats.Therefore,Decoy-state protocol and Measurement-device-independent(MDI)QKD are shown in Table 2-1,which address the vulnerabilities of weakly coheren

27、t sources and detector devices respectively,enable the unconditional security of QKD to be guaranteed in the non-perfect device case.This is a major Preliminary Study of Advanced Technologies towards 6G Era:QITs 2023-05-Table 2 1 The stages of development of the QDK protocoladvance in the protocoliz

28、ation of QKD.Twin-Field(TF)QKD protocol,for the first time,breaks the PLOB bound without a quantum repeater,becoming a widely recognized technical solution for ultra-long-range QKD.With a new transmission distance of 833 km in 2022 2-1,TF-QKD is a step closer to bringing the 1,000 km quantum communi

29、cation.Compared with DV-QKD,CV-QKD has the capability of Mbit/s high-speed key formation at short and medium transmission distances,which is suitable for high-speed metropolitan area network applications.The development of the CV-QKD system architecture is divided into three stages,with the random l

30、ocal oscillation(RLO),local local oscillation(LLO)and discrete modulated digital system,among which the discrete modulated digital system is expected to become the mainstream commercial solution for CV-QKD in the future.In 2022,the LLO-CV-QKD system demonstrated in the metropolitan area was reported

31、 with a secure key rate of 21.53 Mbit/s at 25 km distance 2-2,realizing LLO-CV-QKD with ultra-high secure key rate and laying a solid foundation for CV-QKD with even higher secure key rate.In the various QKD protocols mentioned above,none of the device security risks have been completely avoided,alt

32、hough MDI-QKD has addressed the flaw of attacker-controlled probes.The ideal solution would be to apply an entanglement-based DI-QKD system that deals with the security vulnerabilities that allow an attacker to control all devices and can reach an upper limit of information-theoretic security at the

33、 physical level.In 2022,the British,German and Chinese research teams,simultaneously reported three experimental advances in DI-QKD proof-of-principle,enabling 3.32bit/s in a DI-QKD system based on the E91 protocol 2-3,a predictive entanglement-based DD-QKD with a BER of 0.078 2-4,and a 200m fiber D

34、I-QKD based on polarization entangled photons 2-5.It is important to note that these techniques are theoretically validated and are currently difficult to industrialize due to the very strong capabilities of the hypothetical attacker.-06-Preliminary Study of Advanced Technologies towards 6G Era:QITs

35、 2023The satellite transmission QKD system is also a major development technology for quantum communication,and its main goal is to conduct satellite-ground high-speed QKD experiments with the help of a satellite platform,and to proceed with wide-area quantum key network experiments.Satellite QKD ne

36、tworks have unique advantages.On the one hand,compared with optical fiber transmission QKD,satellite transmission QKD has lower loss and can significantly increase the transmission distance.On the other hand,satellites can be used as repeaters,which can effectively improve the application scope and

37、security of QKD.In recent years,countries have attached great importance to the development of satellite QKD and have carried out a series of experiments on satellite quantum networks.In 2022,Chinas Mozi satellite has reached the current farthest QKD of 1200 km 2-6 and launched the worlds first QKD

38、micro-nano-satellite Jinan-1 2-7.As QKD is on the commercialization track,integrated photonics provides a powerful,miniaturized and cost-effective platform to implement QKD transmitter and receiver devices.The design of integrated QKD systems requires the selection of different optical designs as we

39、ll as material platforms depending on the requirements of the application.Silicon-based platforms offer proven processing platforms but require the use of hybrid integrated laser sources;InP platforms allow monolithic integration of lasers and high-speed phase modulators,but device size as well as c

40、ost aspects still need to be improved.Future developments in full-chip QKD tend to use not just one of these materials,but a combination of several materials to design devices suitable for the system,thereby replacing a large number of high-performance discrete devices,reducing device cost and size,

41、improving system integration,and further promoting the large-scale commercialization of quantum communication systems.2.1.3 Quantum Random Number GeneratorQuantum Random Number Generator(QRNG)is a system that generates true random numbers based on the principles of quantum physics with the character

42、istics of unpredictability,irreducibility,and unbiasedness,which is a vital device in quantum communication systems and can be applied in QKD systems.In the QRNG system,the corresponding quantum state needs to be prepared first.Afterward,the quantum state is measured and the raw data is obtained.The

43、 quantum randomness-07-Preliminary Study of Advanced Technologies towards 6G Era:QITs 2023contained in the raw data can be quantified by modeling as well as by calculation.Based on the results of the quantization analysis,the raw data are post-processed to obtain the final true random number.QRNGs a

44、re divided into two main categories:discrete and continuous,depending on the random source used.The discrete QRNG mainly uses signals such as single photon sources and entangled photon pairs as carriers of random variables.The scheme is simple in principle and has obvious quantum uncertainty,but the

45、 random number generation rate of this scheme is low,which is mainly limited by the linewidth of the random source and the detection efficiency of the single-photon detector.The continuous QRNG uses the true randomness of the spontaneous radiation photon phase to convert the random fluctuation phase

46、 into light intensity,which is then captured and quantized by a high-speed analog-to-digital converter to obtain the raw quantum random number.This scheme is not restricted by the saturation count rate of single-photon detectors and substantially increases the generation rate of raw random numbers.C

47、urrently,the development direction of QRNG technology is focused on increasing the generation rate of quantum random numbers,miniaturization of quantum random number generating devices,and reducing the cost of quantum random number generators.The random number generation rate is the most important m

48、etric for QRNG.In 2022,Ghent University,together with the Technical University of Denmark and the Politecnico di Bari in Italy,experimentally demonstrated an ultra-fast generation rate of 100Gbit/s 2-8,raising the new record for vacuum quantum random number generation by an order of magnitude.Beside

49、s,QRNG chips with stable performance,low cost,and high volume production have become an urgent requirement for cryptographic systems.Many companies and research institutes are conducting miniaturization and chip-based research,and a variety of technology solutions and device forms are becoming comme

50、rcially available for QRNG products,with the highest random number generation rates increasing to 10Gbit/s.Koreas SKT and Samsung launched GalaxyQuantum3 smartphone to promote chip-based QRNG in mobile terminal authentication and information encryption applications.In the future,QRNG is expected to

51、enter the consumer market rapidly as the QRNG chip-based technology matures and cost-effectiveness is realized.-08-Preliminary Study of Advanced Technologies towards 6G Era:QITs 20232.2.1 ITU-TITU-T was the first SDO to standardize QKD as a network.In July 2018,ITU-T SG13 initiated the first work it

52、em(i.e.,Y.3800)on QKD and brought in the concept of Quantum Key Distribution Network(QKDN)firstly.Afterwards,there are more than 40 work items conducted by 4 different groups in ITU-T under the umbrella of QKDN,which can be divided into 4 branches as follows:Study Group 13(Q16/13 and Q6/13):focus on

53、 network aspects of QKDN Study Group 17(Q15/17,formerly Q4/17):focus on security aspect of QKDN Study Group 11(Q2/11):focus on QKDN high layer protocols and signaling Focus Group on Quantum information technology for Networks(FG-QIT4N):to study the implications of QITs for both quantum and ICT netwo

54、rkFigure 2 1 QKDN standardization timeline2.2 Standardization Activities for QKDNQKD and its networking technologies have attracted a lot of interest in multiple SDOs,e.g.,ISO,IEC,ITU,IEEE,IETF,ETSI,as shown in Figure 2-1.The status of Quantum Key Distribution Networks(QKDN)standardization in differ

55、ent SDOs will be briefly reviewed in the following sub-clauses.-09-Preliminary Study of Advanced Technologies towards 6G Era:QITs 20232.2.1.1 ITU-T Study Group 13A landscape diagram for the QKDN standardization work in SG13 is as illustrated in Figure 2-2.SG13 has the following work items on QKDN as

56、 listed in Table 2-2.Figure 2 2:QKDN standardization landscape in ITU-T SG13Table 2 1 The stages of development of the QDK protocol-10-Preliminary Study of Advanced Technologies towards 6G Era:QITs 20232.2.1.2 ITU-T Study Group 17A landscape diagram for the QKDN standardization work in SG17 is illus

57、trated in Figure 2-3.SG17 has the following work items on QKDN as listed in Table 2-3.-11-Preliminary Study of Advanced Technologies towards 6G Era:QITs 2023Figure 2 3:QKDN standardization work items in SG17Table 2 3 QKDN related work items in ITU-T SG17-12-Preliminary Study of Advanced Technologies

58、 towards 6G Era:QITs 20232.2.1.3 ITU-T Study Group 11A landscape diagram for the QKDN standardization work in SG11 is illustrated in Figure 2-4.SG11 has the following work items on QKDN protocols,as listed in Table 2-4.2.2.1.4 ITU-T FG-QIT4NFG-QIT4N has the following work items on QKDN as listed in

59、Table 2-5.Figure 2 4:QKDN standardization work items in SG11Table 2 4 QKDN related work items in ITU-T SG11Table 2 4 QKDN related work items in ITU-T SG11-13-Preliminary Study of Advanced Technologies towards 6G Era:QITs 2023Table 2 6:QKD related work items in ETSI2.2.2 ETSI ISG-QKDETSI initiated th

60、e industry specification group(ISG)on QKD in 2008.ETSI ISG-QKD has published nine specifications on QKD until 2019 and have several work items ongoing as listed in Table 2-6.The previous work mainly focused on QKD link-level issues,including QKD optical components,modules,internal and application in

61、terfaces,practical security,etc.Note that ETSI has also initiated the study of QKD network architectures recently and the specification of QKD security certification based on common criteria.-14-Preliminary Study of Advanced Technologies towards 6G Era:QITs 2023-15-Preliminary Study of Advanced Tech

62、nologies towards 6G Era:QITs 20232.2.3 ISO/IEC JTC 1/SC 27ISO/IEC JTC 1/SC 27 initiated the study period Security requirements,test and evaluation methods for quantum key distribution in 2017.In 2019,the study period was completed,and a new work item ISO/IEC 23837(Part 1&2)was established as listed

63、in Table 2-7.2.3 Implications for 6GThe Quantum key cloud platform obtains quantum keys from QKD or QRNG,and stores and manages the key safely.Through the security mechanism,the quantum keys can be distributed to the user security terminal and provide high-level security protection to the user secur

64、ity terminal even in the face of challenges of quantum computing.The Quantum key cloud platform can provide quantum encryption services for the government,enterprises,and individuals to protect the storage and transmission of data safely.At present,China Unicom has built a Quantum key cloud platform

65、 in Xiongan New Area and carried out quantum encryption technology research and application demonstration,such as quantum encrypted call,quantum public network cluster intercom,quantum video conference,and quantum UAV patrol.The following describes two of the representative application scenarios,nam

66、ely,quantum encrypted call and quantum public network cluster intercom.Table 2 7:QKD related works items in ISO/IEC JTC1-16-Preliminary Study of Advanced Technologies towards 6G Era:QITs 20232.3.1 Application Scenario 1:Quantum encrypted callFigure 2 5 System diagram of Quantum encrypted callIn the

67、application scenario of quantum encrypted call,the security terminal uses the pre-charged quantum keys as“the identity authentication keys”and“the basic encryption keys”.When making a call or sending a message,the Quantum key cloud platform selects a set of quantum keys as“the session keys”,and encr

68、ypts them with“the basic encryption keys”and sends them to the security terminal.The security terminal decrypts“the session keys”with“the basic encryption keys”,and then“the session keys”can be used to protect the voice and data stream of the security terminal.In addition,we have developed a special

69、 App,through which the quantum security terminal can realize encrypted transmission of text,voice,pictures,files,and other contents.And the App supports the function of burn after reading.-17-Preliminary Study of Advanced Technologies towards 6G Era:QITs 20232.3.2 Application Scenario 2:Quantum publ

70、ic network cluster intercomFigure 2 6 System diagram of Quantum public network cluster intercomIn the application scenario of quantum public network cluster intercom,the terminal and the Command and dispatching platform integrate the quantum SDK,which can obtain the quantum keys from the Quantum key

71、 cloud platform and perform quantum encryption to ensure the security of the cluster voice,video,image and other service data and operational signaling when transmitted over the public network.If the eavesdropper uses terminals C and D without integrating quantum encryption function to illegally ent

72、er the cluster intercom system of terminals A and B,which integrating quantum encryption function,when terminals A and B send voice and video messages,terminals C and D cannot crack the received quantum encrypted information,that is,they cannot receive voice or video information normally.However,ter

73、minals A and B can receive the messages normally,which were sent by terminals C and D.-18-Preliminary Study of Advanced Technologies towards 6G Era:QITs 20233.Quantum Machine Learning(QML)It is highly expected that the 6th generation(6G)communication systems will lay a foundation of pervasive digiti

74、zation,ubiquitous connection and full intelligence.To satisfy the dramatically increased communication system performance and rich diversity of innovative services,Quantum Machine Learning(QML)is emerged due to its information processing capability beyond its classical counterpart,which is achieved

75、by combining the established benefits of quantum mechanism and machine learning.In the white paper of version 2021,we introduced the concepts and basic paradigms of QML on a high level.Wherein quantum-enhanced machined learning can be further categorized according to the three branches of ML(i.e.,su

76、pervised learning,unsupervised learning,and reinforcement learning).In particular,quantum-enhanced reinforcement learning has a potential to revolutionize the field of artificial intelligence(AI).In this following,we will get insight into the research of quantum-enhanced machine learning by analyzin

77、g two representative works in detail.The first work in 3-1 gains speed-up of reinforcement learning by probing the environment in superpositions and provides a general method of quantum improvements in the three paradigms of machine learning.The second work in 3-2 shows an experiment performed to re

78、construct an unknown photonic quantum state with a limited amount of copies,for which the performance in terms of fidelities can be improved when assisted by semi-quantum reinforcement learning approach.3.1 Quantum-Enhanced Reinforcement LearningReinforcement Learning(RL)3-3 is an area of machine le

79、arning concerned with how intelligent agents react in an environment with a target of maximizing the reward.The focus of RL is on finding a balance between exploration(of uncharted territory)and exploitation(of current knowledge)3-4.As compared to supervised learning,labeled training data is not req

80、uired for reinforcement learning.However,partially supervised RL algorithms can combine the advantages of supervised and RL algorithms.One powerful feature of RL is suitable for dealing with large environments.Reinforcement learning is typically used for solving control and classification problems.C

81、onventional and notable RL algorithms such as Q-learning and multi-armed bandit take as an input the current state of the network and enable the prediction of the next state.A-19-Preliminary Study of Advanced Technologies towards 6G Era:QITs 2023promising application of RL in communication contribut

82、es to scheduling parameters optimization across various layers.Additionally,deep learning can be combined with RL to facilitate learning long-term temporal dependence sequences in such a way that the accumulation of errors wont grow very fast 3-5.In quantum-enhanced reinforcement learning,a quantum

83、agent interacts with a classical or quantum environment and occasionally receives rewards for its actions,which allows the agent to learn what to do in order to gain more rewards.There are various ways of achieving quantum speedup.For example,in 3-6 a quantum agent which has quantum processing capab

84、ility is provided in achieving a quadratic speed-up for active learning.Alternatively,the work in 3-1 gains speed-up by probing the environment in superpositions.Furthermore,a general method of quantum improvements in the three paradigms of machine learning is provided in 3-1.This section will intro

85、duce the major work in 3-1.The QML can be represented by an agent-environment paradigm,where a learning agent A interacts with interacts with an unknown environment E via the exchange of messages,interchangeably issued by A(called actions)and E(called percepts).For reinforcement learning,the percept

86、 space also contains the reward.In the quantum extension,both A and E are quantum systems,where the sets of actions and percepts become Hilbert spaces and form orthonormal bases.The agent and the environment act on a common communication register RC(capable of representing both percepts and actions)

87、.The agent(environment)is described as a sequence of completely positive trace-preserving maps M_tA(M_ -one for each time-step-that acts on the register RC,but also a private register RA(E)that constitutes the internal memory of the agent(environment),as illustrated in Figure 3-1.The central object

88、characterizing an interaction is,for the quantum case,generated by performing periodic measurements on RC in the classical(often called computational)basis.The generalization of this process for the quantum case is a tested interaction:we define the tester as a sequence of controlled maps of the for

89、m .The history,relative to a given tester,is defined to be the state of the register RT.Based on this,the work establishes a schema for quantum improvements in RL agent.The kernel idea is to obtain a useful property of the environment and identify settings where quantum-20-Preliminary Study of Advan

90、ced Technologies towards 6G Era:QITs 2023technique provably helps to improve.Then,construct an improved agent that uses the properties from the previous points.Figure 3 1 Tested agent-environment interaction 3-13.2 Reconstruction of a Photonic Qubit State with Reinforcement LearningExtracting inform

91、ation from an unknown quantum state is an important task in quantum information.The general way by measuring the averages of a set of observables becomes difficult,even unfeasible in large target system,due to its relying on enough number of copies of the target state.The work in 3-2 shows an experi

92、ment performed to reconstruct an unknown photonic quantum state with a limited amount of copies.In particularly,a semi-quantum reinforcement learning approach is employed to adapt one qubit state,an agent,to an unknown quantum state,an environment,by successive single-shot measurements and feedback,

93、in order to achieve maximum overlap.This section will introduce major work in 3-2.The semi-quantum reinforcement learning algorithm outputted in 3-7 is applied to the work of 3-2 where the system involves three parts,namely,the environment(E),the register(R),and the agent(A).Regarding the experiment

94、al setup,the procedures can be described as follows with the illustration in Figure 3-2.The photon pairs are generated from periodically poled KTP(PPKTP)crystal by the spontaneous parametric down-conversion(SPDC)process,and work as environment system and register system,respectively.On the environme

95、nt route,photons are randomly prepared and go through a unitary gate which is determined by the last iteration.On the register route,the chopper works as a switch for creating a pseudo-on-demand-21-Preliminary Study of Advanced Technologies towards 6G Era:QITs 2023single photon source assisted by po

96、st-selection.Then,the photons are prepared at|H(H and V represent the horizontal and vertical polarizations,respectively)and interact with the environment by a controlled-NOT(CNOT)gate.After the interaction,measure the register qubit on the basis and obtain the outcome 0 or 1,also referred to as rei

97、nforcement signal,which is used to decide reward or punishment for the next iteration.If state is detected,the outcome is 0,while the outcome of 1 corresponds to a detected state .The outcome in the current iteration is input to a computer to control the action on the agent system as well as the env

98、ironment system by classical communications.If the outcome is 0,nothing is done,while a unitary operation is performed corresponding to an outcome of 1.The angle range for the rotation is modified by a reward function,which is decided by the outcome of the last step and contains a parameter controll

99、ing the reward/punishment ratios.Figure 3 2 Framework of experiment system 3-2-22-Preliminary Study of Advanced Technologies towards 6G Era:QITs 2023The experimental results show that,when assisted by such a quantum machine learning technique,fidelities of the deterministic single-photon agent state

100、s can achieve over 88%under a proper reward/punishment ratio()within 50 iterations(k),as illustrated in Figure 3-3.Moreover,as per the figure,a larger provides better results,at the cost of longer duration to achiever convergence.This is aligned with the characteristic in RL where a larger produces

101、larger exploration implying more information is extracted and thus enhancing the final fidelity,while the final fidelity is achieved later than with a smaller due to more fluctuations caused.Figure 3 3 Experimental results for the semi-quantum reinforcement learning process 3-2-23-Preliminary Study

102、of Advanced Technologies towards 6G Era:QITs 2023Reference2-1 Wang,S.,Yin,ZQ.,He,DY.et al.(2022-01-17).Twin-field quantum key distribution over 830-km fibre.Nature Photonics.16,154161.2-2 Wang,H.,Li,Y.,Pi,Y.et al.(2022-06-25).Sub-Gbps key rate four-state continuous-variable quantum key distribution

103、within metropolitan area.Communications Physics.5,162.2-3 Nadlinger,D.P.,Drmota,P.,Nichol,B.C.et al.(2022-07-27).Experimental quantum key distribution certified by Bells theorem.Nature 607,682686.2-4 Zhang,W.,van Leent,T.,Redeker,K.et al.(2022-07-27).A device-independent quantum key distribution sys

104、tem for distant users.Nature 607,687691.2-5 Wen-Zhao Liu,Yu-Zhe Zhang,Yi-Zheng Zhen,Ming-Han Li,Yang Liu,Jingyun Fan,Feihu Xu,Qiang Zhang,and Jian-Wei Pan.(2022-07-27).Toward a Photonic Demonstration of Device-Independent Quantum Key Distribution.Physical Review Letters.129,050502.2-6 Bo Li,Yuan Cao

105、,Yu-Huai Li,Wen-Qi Cai,Wei-Yue Liu,Ji-Gang Ren,Sheng-Kai Liao,Hui-Nan Wu,Shuang-Lin Li,Li Li,Nai-Le Liu,Chao-Yang Lu,Juan Yin,Yu-Ao Chen,Cheng-Zhi Peng,and Jian-Wei Pan.(2022-04-46).Quantum State Transfer over 1200 km Assisted by Prior Distributed Entanglement.Physical Review Letters.128,170501.2-7

106、http:/ https:/arxiv.org/abs/2209.04339v13-1 Dunjko,Vedran;Taylor,Jacob M.;Briegel,Hans J.(2016-09-20).Quantum-Enhanced Machine Learning.Physical Review Letters.117(13):130501.3-2 Yu,Shang;Albarran-Arriagada,F.;Retamal,J.C.;Wang,Yi-Tao;Liu,Wei;Ke,Zhi-Jin;Meng,Yu;Li,Zhi-Peng;Tang,Jian-Shun(2018-08-28)

107、.Reconstruction of a Photonic Qubit State with Quantum Reinforcement Learning.Advanced Quantum Technologies.2(78):1800074.3-3 https:/en.wikipedia.org/wiki/Reinforcement_learning3-4 Leslie Pack Kaelbling,Michael L.Littman,and Andrew W.Moore.1996.Reinforcement learning:a survey,in Journal of Artificia

108、l Intelligence Research.4,1(Jnauary 1996),237285.3-5 U.Challita,L.Dong and W.Saad,Proactive Resource Management for LTE in Unlicensed Spectrum:A Deep Learning Perspective,in IEEE Transactions on Wireless Communications,vol.17,no.7,pp.4674-4689,July 2018,doi:10.1109/TWC.2018.2829773.3-6 Paparo,Giusep

109、pe Davide;Dunjko,Vedran;Makmal,Adi;Martin-Delgado,Miguel Angel;Briegel,Hans J.(2014).Quantum Speedup for Active Learning Agents.Physical Review X.4(3):031002.3-7 F.Albarr_an-Arriagada,J.C.Retamal,E.Solano,and L.Lamata,Phys.Rev.A 2018 98,042315-24-Preliminary Study of Advanced Technologies towards 6G

110、 Era:QITs 2023AcknowledgementGrateful thanks to the following contributors for their wonderful work on this whitepaper:Editors:Chih-Lin I(China Mobile),Xin GUO(Lenovo)Contributors:China Information and Communication Technologies Group Corporation(CICT)&National Information Optoelectronics Innovation

111、 Center(NOEIC)University of Science and Technology Beijing(USTB)China Unicom Lenovo Yanxin HAN,Xin HUA,Xi XIAOZhangchao MA,Jianquan WANG,Lei SUN Zhongyan DU,Chao LENG,Changlian MA Xin GUO-25-Preliminary Study of Advanced Technologies towards 6G Era:QITs 2023 Yanxin HAN,Xin HUA,Xi XIAOZhangchao MA,Ji

112、anquan WANG,Lei SUN Zhongyan DU,Chao LENG,Changlian MA Xin GUOFuTURE FORUM is committed to cutting edge technologies study and applications.Controversies on some technical road-maps and methodologies may arise from time to time.FuTURE FORUM encourages open discussion and exchange of ideas at all lev

113、els.The White Paper released by FuTURE FORUM represents the opinions which were agreed upon by all participating organizations and were supported by the majority of FuTURE FORUM members.The opinions contained in the White Paper does not necessarily represent a unanimous agreement of all FuTURE FORUM members.FuTURE FORUM welcomes all experts and scholars active participation in follow-on working group meetings and workshops.we also highly appreciate your valuable contribution to the FuTURE White Paper series.