Multi-Connectivity in 5G New Radio: Configuration Algorithms and Performance Evaluation
Author
Lopez Lechuga, Melisa
Term
4. term
Education
Publication year
2018
Submitted on
2018-05-03
Pages
63
Abstract
5G New Radio (NR) er udviklet til at understøtte mange forskellige behov, blandt andet Ultra-Reliable Low-Latency Communications (URLLC), hvor meget små beskeder skal nå frem på cirka 1 ms med en succesrate på 99,999%. En foreslået metode er dual connectivity med dataduplikering: den samme pakke sendes samtidig via to forskellige basestationer, så mindst én kopi med stor sandsynlighed når frem. Dette projekt bruger systemniveau-simuleringer til at undersøge duplikering på PDCP-laget (Packet Data Convergence Protocol) i forbindelse med dual connectivity. Hver brugerterminal (UE) sender to kopier af hver URLLC-pakke over sine to forbindelser. Det simulerede net er heterogent: 21 makroceller (store dækningsområder) og inden i hver en klynge af fire picoceller (små celler). Først optimerer vi scenariet for enkeltforbindelse og finder, at det kan bære op til 8 Mbit/s URLLC-trafik, mens kravene til latenstid og pålidelighed opfyldes. Når vi derefter aktiverer dual connectivity kun for URLLC-trafik i denne lav-interferens-situation, giver det ingen målbar gevinst. Dernæst tilføjer vi tung baggrundstrafik af typen enhanced Mobile Broadband (eMBB) med fulde buffere (altså altid data at sende). I dette blandede scenarie kan dual connectivity sænke latenstiden, men gevinsten er meget følsom over for de konkrete netforhold. Til sidst vurderer vi en optimering: hvis UE'en modtager pakken via den ene forbindelse, annullerer nettet udsendelsen af duplikatet på den anden (pakken kasseres, før den sendes). Denne cancel-on-success-tilgang forbedrer latenstiden, især ved høj belastning, fordi ekstra kø- og bufferforsinkelse undgås.
5G New Radio (NR) is designed to support many different needs, including Ultra-Reliable Low-Latency Communications (URLLC), where very small messages must arrive in about 1 ms with a 99.999% success rate. One proposed way to achieve this is dual connectivity with data duplication: the same packet is sent at the same time through two different base stations, so at least one copy is likely to get through. This project uses system-level simulations to study duplication at the PDCP (Packet Data Convergence Protocol) layer for dual connectivity. Each user device (UE) sends two copies of every URLLC packet over its two links. The simulated network is heterogeneous: 21 macro cells (large coverage areas) and, within each, a cluster of four pico cells (small cells). We first optimize the scenario for single connectivity and find it can carry up to 8 Mbps of URLLC traffic while meeting the latency and reliability targets. When we then enable dual connectivity only for URLLC traffic in this low-interference setting, it provides no measurable gain. Next, we add heavy background enhanced Mobile Broadband (eMBB) traffic with full buffers (i.e., there is always data to send). In this mixed scenario, dual connectivity can reduce latency, but the size of the benefit is highly sensitive to the exact network conditions. Finally, we evaluate an optimization: if the UE successfully receives the packet over one link, the network cancels transmission of the duplicate on the other link (discarding it before it is sent). This cancel-on-success approach improves latency, especially at high load, by avoiding extra queueing/buffering delay.
[This abstract was generated with the help of AI]
Keywords
Documents