Research On Random Access Technologies For Ultra-Reliable And Low-Latency Communications

With the wide adoption of the Internet of Things(IoT)and Machine Learning(ML),the Fifth Generation(5G)and Beyond 5G(B5G)are expected to cater to different Quality of Service(Qo S)requirements for diverse applications.The 5G New Radio(NR)ushered in a paradigm shift by introducing massive Machine Type Communications(m MTC)and Ultra-Reliable and Low Latency Communications(URLLC)requirement categories for MTC.Among them,the URLLC service is critical and essential for the most demanding mission-critical use-cases,including factory automation,automation vehicles,remote control,and virtual/augmented reality(VR/AR),with stringent latencyreliability requirements.While 5G NR and other wireless systems have enabled URLLC services under certain scenarios,the true vision of Io T connectivity as required by the diverse range of MTC applications is yet to be realized.One of the main challenges is how to provide intelligent radio access for critical Io T devices.To address the challenge,this thesis focuses on the system design and performance enhancement of the Radio Access Network(RAN)for URLLC service.The study is conducted from two aspects,including the network analysis of the grant-free procedure,and the dynamic optimization of uplink transmission.First,considering the URLLC requirements in critical IoT networks,grant-free(GF)random access,also known as configuration grant(CG),is proposed as a romising URLLC solution for 5G NR.A novel spatio-temporal mathematical model was presented to analyze three different contentionbased GF schemes,including the Reactive,the K-repetition,and the Proactive schemes,respectively.In this model,the reliability and latency performances of a randomly chosen Io T device are characterized and analyzed jointly by defining the latent access failure probability.Then,the exact closed-form expression for the latent access failure probability is derived under three different contention-based GF schemes,which established the theoretical formulations and comparative insights for GF access designs.The results show that the GF schemes are expected to meet the URLLC requirements,but the performances are affected by network parameters such as network densities and repetition values.Under shorter latency constraints,the Proactive scheme provides the lowest latent access failure probability,whereas,under longer latency constraints,the K-repetition scheme achieves the lowest latent access failure probability,which depends on K.If K is overestimated,the Proactive scheme provides lower latent access failure probability than the Krepetition scheme.The analytical model proposed in this paper can give the performance insights of the three GF schemes directly and can also be applied for the reliability and latency performance evaluation of other types of GF schemes in the cellular-based networks.Second,considering the massive URLLC(mURLLC)service,non-orthogonal multiple access(NOMA)has been proposed to synergize with GF,and a signature-based GF-NOMA scheme was designed,where the resource configuration of the GF-NOMA scheme for m URLLC service is formulated as a Partially Observable Markov Decision Process(POMDP)problem with enormous state and action spaces.To address this problem,Deep Reinforcement Learning(DRL)-based uplink resource configuration approaches are developed to optimize the number of successfully served user equipments(UEs)under the latency constraint via adaptively configuring the uplink resources,including the repetition values and the contention-transmission unit(CTU)numbers,for the Krepetition scheme and the Proactive scheme,respectively.The results show the superior performance of the proposed approaches over the conventional load estimation-based approach(LE-URC)in heavy traffic and demonstrate its capability in dynamically configuring in long term for m URLLC service.In addition,with our learning optimization,the Proactive scheme always outperforms the Krepetition scheme in terms of the number of successfully served UEs,especially under the high backlog traffic scenario.Finally,considering the limitation of the waiting latency of the arrival data packets on the m URLLC service,i.e.,UE must wait till the next CG period to transmit the packets when the UE uplink data arrives after the starting slot offset of the CG,which will add to the latency much more than the latency constraint(1ms).To mitigate this problem,this thesis proposed a Multiple Configured-Grants GF-NOMA(MCG-GF-NOMA)transmission scheme to support different starting offsets of the resources with respect to uplink packet arrival time.Then the latency(including waiting latency)and reliability performances for each CG are characterized and analyzed and a Cooperative Multi-Agent based DDQN(CMA-DDQN)algorithm is proposed to optimize the allocations of resources among MCGs to maximize the number of successful transmissions under the CTU resource constraint,the latency constraint,and the starting slot constraint.The results have shown that with this scheme,UE can choose any of these grants for transmission as soon as the data arrives to further reduce waiting latency and cope with the extreme m URLLC services under the complex environment.The number of successfully served UEs in the MCG-GF-NOMA system is circa four times more than that in the GF-NOMA system,and the latency of successfully served UEs in the MCG-GF-NOMA system is circa half of that in the GF-NOMA system in high traffic scenario.