3GPP Release 18: 5G-Advanced RAN Features(RAN1 to RAN5)
- Venkateshu
- Nov 28
- 9 min read
Introduction
3GPP Release 18 marks the first 5G-Advanced release, focusing on AI/ML integration, extreme performance for XR/Industrial IoT, mobile IAB, enhanced positioning, and spectrum efficiency up to 71 GHz. RAN workgroups (RAN1–RAN5) deliver these through physical layer evolution, protocol optimizations, architecture updates, RF requirements, and testing frameworks.
Release 18 builds on Rel-17's foundations with AI/ML for RAN optimization, centimeter-level positioning, mobile IAB for dynamic backhaul, XR/multisensory support, and duplex/spectrum innovations. RAN1 drives PHY/AI enhancements, RAN2 refines L1/L2 mobility and XR, RAN3 advances architecture for AI/ML and slicing, RAN4 specifies RF for new bands/devices, and RAN5 ensures conformance testing. These enable new verticals like FRMCS rail, RedCap evolution, and NTN expansions while boosting eMBB/URLLC/mMTC.
RAN1: Physical Layer and AI/ML Innovations
RAN1 leads PHY advancements, AI/ML framework, and new capabilities like smart repeaters and duplex evolution.
Key Features
MIMO evolution (multi-panel UL rank-8, MU-MIMO up to 24 DMRS ports, multi-TRP TCI framework).
How it works: Extends Type-I/II CSI reporting with unified TCI framework across multi-TRP panels. gNB schedules up to 24 DMRS ports for MU-MIMO (vs 12 in Rel-17), enabling rank-8 UL per UE. DCI indicates joint TCI states; UE applies phase/precoding across panels.
Problem solved: Rel-17 multi-TRP lacked unified signaling, causing 20-30% spectral efficiency loss in dense deployments. Rank limitation capped UL throughput at 4-6 layers per UE. Result: 40% UL capacity gain in stadiums/festivals
AI/ML for CSI feedback compression, beam management, and positioning.
How it works: Neural network compresses Type-II CSI (32 ports → 8 coefficients) using codebook trained offline. gNB deploys model via RRC; UE reports compressed feedback. Beam prediction uses historical L1-RSRP patterns to pre-position beams before handover.
Problem solved: CSI overhead consumed 15-20% DL resources; beam management failed 25% during high-mobility (e.g., highways). Result: 50% CSI overhead reduction, 30% handover success rate improvement.
Coverage enhancements (UL full-power transmission, low-power wake-up signals).
How it works: gNB signals UE to apply full PA power across all UL layers (no power backoff per layer). Separate low-power wake-up receiver (duty-cycled, -110dBm sensitivity) receives WUS (Wake-up Signal) before main DRX cycle. WUS carries 1-bit indication (monitor PDCCH or sleep).
Problem solved: Rel-17 UL coverage limited by per-layer power backoff (3dB loss in rank-4 MIMO); main receiver consumed 50% UE battery during DRX monitoring. Result: +3dB UL coverage extension, 40% battery saving for IoT/video streaming.
Sidelink CA in ITS bands, dynamic spectrum sharing (DSS) with LTE CRS.
How it works: Sidelink supports CA across n47 (5.9GHz ITS) + FR1; autonomous resource selection with inter-UE coordination Type 2c. NTN IoT disables HARQ (open-loop repetition only) due to >500ms RTT; pre-compensates Doppler in DMRS.
Problem solved: Rel-17 sidelink single-carrier only (50% throughput loss); NTN IoT HARQ timeouts caused 30% packet loss. Result: 2x sidelink throughput in V2X platoons, 95% NTN IoT reliability
XR/multisensory communications (high-reliable low-latency support).
How it works: New QoS flow with <1ms latency budget, multisensory packet marking (video+ haptic+audio streams). gNB prioritizes via pre-emption; UE reports pose/motion data for predictive scheduling.
Problem solved: Rel-17 XR support was unicast-only; haptic feedback latency exceeded 20ms (unusable for teleoperation). Result: End-to-end <5ms for AR/VR+haptics in industrial remote control.
NTN enhancements (UL coverage for smartphones, HARQ disabling for IoT).
How it works: Release 18 improves uplink coverage for smartphones in Non-Terrestrial Networks (NTN) by optimizing the physical layer transmission, allowing higher transmission power and better link budget management adapted to satellite channels. For IoT devices over NTN, HARQ feedback is disabled, shifting to an open-loop repetition scheme because the long satellite RTT makes traditional HARQ feedback inefficient.
Problem solved: Previously, smartphone uplink coverage over NTN was limited by insufficient power control and link margin, leading to poor connectivity. HARQ feedback caused throughput degradation and latency issues for IoT devices due to satellite delay. Disabling HARQ eliminates feedback delays and improves reliability for constrained IoT devices. This enables robust global IoT and smartphone connectivity beyond terrestrial network.
Applications
Dense urban XR (multi-TRP MIMO reduces latency <1ms for AR/VR).
Industrial automation (AI/ML beam prediction cuts handover failures by 30%).
V2X/high-mobility (sidelink CA boosts reliability in n47 ITS band).
Implementation Aspects
gNB PHY: Integrate AI models for CSI compression (e.g., neural networks predict Type-II CSI from Type-I, reducing overhead 50%). Deploy multi-TRP TCI via RRC/DCI, with 2 TAs for UL timing.
UE: Support low-power wake-up receiver (separate from main RF chain) for DRX-aligned signaling.
Example: Ericsson's Rel-18 trial used AI/ML for beam management, achieving 20% spectral efficiency gain in multi-TRP setup.
RAN2: L1/L2 Mobility and XR Protocol Enhancements
RAN2 focuses on MAC/RLC/PDCP/RRC procedures for mobility, XR, and power efficiency.
Key Features
L1/L2-centric inter-cell mobility (dynamic cell switching, L1 beam management).
How it works: UE measures L1-RSRP on SSB/CSI-RS during connected mode without RRC gaps. gNB triggers CHO (Conditional Handover) based on L1 thresholds; UE executes autonomously. L2 switching via MAC CE (no RRC involved).
Problem solved: RRC-based handover interruption was 50-100ms; high-speed rail (500km/h) saw 40% failure rate. Result: <5ms interruption, 95% success at 350km/h
XR enhancements (multisensory data, dual connectivity activation).
How it works: RRC configures XR QoS flow with pose/motion reporting (6DoF data every 5ms). Conditional PSCell activation: UE measures SCG L1-RSRP, triggers via MAC CE without RRC reconfiguration. Multisensory marking differentiates video/haptic/audio streams.
Problem solved: Rel-17 DC activation interruption >50ms broke XR sync; no multisensory QoS differentiation. Result: <10ms SCG activation, independent QoS per sensory stream (haptics prioritized).
Multicast evolution (MBS in RRC_INACTIVE, dynamic group management).
How it works: gNB configures MBS sessions via RRC; Inactive UEs join via group ID without state transition. Dynamic switching: unicast→multicast based on UE count threshold. HARQ combines multicast+unicast receptions.
Problem solved: Rel-17 MBS required RRC_CONNECTED (70% battery drain for IoT). Result: 70% energy saving for software updates, 90% capacity gain in stadiums.
RRC state optimizations (small data over Inactive, slicing-aware reselection).
How it works: SIB carries slice-specific RACH occasions/PRACH masks. UE in Idle/Inactive performs slice-aware reselection (highest priority S-NSSAI first). RRC_CONNECTED UEs report allowed NSSAI changes during handover.
Problem solved: Rel-17 slice-unaware access caused 25% URLLC UEs landing on eMBB slices. Result: 95% first-attempt slice attachment success.
Power saving (extended DRX, measurement gap reduction).
How it works: Extended DRX allows User Equipment (UE) to lengthen sleep periods by monitoring paging and control channels less frequently. Measurement gap reduction minimizes interruptions in data transmission caused by measurement requirements by optimizing or combining measurement gaps with other signaling events.
Problem solved: Energy consumption in UEs was high due to frequent control channel monitoring and measurement gaps causing the radio to frequently switch states. By extending DRX cycles and reducing measurement gaps, battery life improves significantly for all device categories, especially for IoT devices requiring prolonged operation.
Applications
High-speed rail (L1/L2 handover <5ms interruption via CHO/DAPS evolution).
Cloud gaming/AR (XR QoS flows with <10ms latency).
Massive IoT (MBS multicast cuts energy 70% for software updates).
Implementation Aspects
Stack changes: RRC signaling for L1 measurements (new report triggers on SSB/CSI-RS), CHO with MCG/SCG targets.
Example: Conditional PSCell addition in NR-DC; UE measures L1-RSRP, triggers activation without RRC gap, tested in lab with Keysight setups showing 50% faster SCG setup.
RAN3: Architecture for AI/ML and Mobile IAB
RAN3 specifies NG-RAN interfaces, IAB mobility, and AI/ML framework.
Key Features
AI/ML for NG-RAN (model deployment, inference over F1/Xn).
How it works: CU/DU exchange AI model parameters via F1AP/XnAP (tensor shapes, quantization). gNB-DU runs inference locally (beam/CSI prediction); results sent to CU. Model update via delta parameters (not full retrain).
Problem solved: No standardized AI integration; vendors used proprietary silos. Result: Interoperable AI across multi-vendor RAN (Ericsson+Nokia validated).
Mobile IAB (node migration, RACH-less handover, NCGI reconfiguration).
How it works: IAB-MT performs L1/L2 handover to target parent; served UEs follow via NCGI (NR Cell Global ID) reassignment. RACH-less: target gNB allocates UL timing via XnAP before migration. Topology advertised in SIB (mobileIAB-Cell).
Problem solved: Static IAB failed during vehicle movement (event coverage vehicles, trains); 60% throughput drop during topology change. Result: Seamless backhaul migration, maintaining 95% UE throughput during 60km/h movement.
SON/MDT enhancements (RACH optimization, NPN logging).
How it works: MDT logs slice-specific RACH failures, L1/L2 mobility events. SON algorithms auto-adjust RACH occasions per slice load. NPN (Non-Public Network) logging includes enterprise identifiers, coverage maps.
Problem solved: Rel-17 SON blind to slice interactions; enterprise NPNs lacked diagnostic data. Result: 40% RACH optimization gain, automated NPN deployment validation.
QoE framework (AR/MR/cloud gaming, RAN-visible QoE over DC).
How it works: gNB collects XR pose data, rendering delay, packet loss via QoE measurements (MAC CE/RRC). Reports to OAM/NWDAF via XnAP/NGAP. Dynamic QoS adjustment based on video stall events, motion sickness indicators.
Problem solved: No RAN visibility into application QoE; operators blind to XR performance degradation. Result: 30% reduction in video stalls through predictive scheduling.
Network slicing (S-NSSAI replacement, partial allowed NSSAI).
How it works: Partial NSSAI allows subset during congestion; dynamic S-NSSAI replacement via NGAP. Timing Synchronization Status (TSS) reported every 10s during GNSS outage, enabling gNB clock correction.
Problem solved: NSSAI mismatch blocked 20% of slice handovers; GNSS outages caused 15% timing drift in FR2. Result: 99% NSSAI consistency, <1μs timing accuracy during outages.
Timing resiliency (NGAP/XnAP TSS reporting).
How it works: NGAP and XnAP protocols are enhanced to include Timing Synchronization Status (TSS) reporting between network nodes to detect and compensate for timing drifts or GNSS outages. This ensures gNBs maintain synchronization by adjusting clocks dynamically based on TSS messages.
Problem solved: Timing alignment is crucial for NR, especially in higher frequency bands and NTN. GNSS outages or network impairments cause timing drift affecting throughput and mobility. The TSS mechanism improves network resilience by enabling fast correction, reducing link failures and service degradation caused by timing errors.
Applications
Vehicle-mounted relays (VMR for event coverage).
Enterprise NPN Phase 2 (SNPN reselection/handover).
Automation (AI/ML SON auto-tunes coverage).
Implementation Aspects
CU/DU: F1AP extensions for AI model parameters (e.g., input/output tensors); mobile IAB MT migration via Xn handover.
Example: Mobile IAB-DU reselection broadcasts mobileIAB-Cell indicator; UEs prioritize via SIB assistance, reducing topology change latency 40%.
RAN4: RF Requirements and New Spectrum
RAN4 defines RF/performance for 5G-Advanced bands/devices.
Key Features
FR1 <5MHz dedicated spectrum (FRMCS migration from GSM-R).
How it works: Specifies ACS/SEM for n100 (1900MHz, 3-5MHz BW) coexisting with GSM-R. Reduced BW filters, power class adjustments for narrowband operation. RRM requirements ensure <1% interference to legacy rail.
Problem solved: European rail migration from GSM-R lacked NR spectrum; 5MHz minimum prevented coexistence. Result: Live coexistence trials (n28+n100) show zero interference.
RedCap evolution (positioning via freq-hopping PRS/SRS).
How it works: Reduced BW UEs (20MHz) use frequency-hopping PRS across 100MHz aggregate BW. gNB coordinates hopping pattern; UE reports ToA per hop for cm-level accuracy.
Problem solved: Rel-17 RedCap positioning limited to 10m accuracy due to narrow BW. Result: <1m accuracy for wearables/industrial sensors.
NTN above 10GHz, sidelink/ITS RF.
How it works: RF requirements for Ka-band (17-31GHz) NTN: ±50kHz Doppler tolerance, 1000ms propagation delay. UE power class 3 with beam correspondence mandatory. Channel models include atmospheric attenuation, rain fade.
Problem solved: Rel-17 NTN limited to L/S-band; mmWave satellite blocked by propagation. Result: GEO satellite coverage at 30GHz for backhaul/IoT.
RRM for L1/L2 mobility, XR KPIs.
How it works: RRM specs for L1-RSRP measurements (<2ms latency), L2 handover execution (<5ms). Interference requirements during multi-TRP operation (ICIC coordination).
Problem solved: No RRM specs for L1/L2 mobility; 30% measurement failures during high load. Result: Standardized performance targets for equipment certification.
Channel models for smart repeaters.
How it works: New channel models simulate propagation between base stations, smart repeaters, and UEs including realistic reflections, shadowing, and Doppler effects. These models capture the enhanced beamforming and relay functions of smart repeaters in complex urban and indoor scenarios.
Problem solved: Existing channel models did not accurately capture the behavior of smart repeaters, leading to suboptimal designs and testing results. The new models enable precise performance predictions and validation of repeater-assisted coverage extension and capacity improvement techniques, helping implementers optimize deployment and operation.
Applications
Rail communications (n100 3-5MHz for FRMCS coexistence).
Wearables (RedCap cm-level positioning).
mmWave FWA (71GHz RF specs).
Implementation Aspects
RF design: UE power classes for <5MHz (reduced BW filters); test models include Doppler for NTN >10GHz.
Example: FRMCS validation in n28; gNB/UE meet ACS/SEM limits while operating 3MHz NR alongside GSM-R.
RAN5: Conformance Testing Framework
RAN5 develops test specs for all Rel-18 features.
Key Features
AI/ML model testing, mobile IAB conformance.
How it works: TTCN-3 test cases verify MT handover (no served UE drop), NCGI reassignment timing (<50ms), topology change signaling. Emulates multi-layer backhaul trees.
Problem solved: No standardized test framework; operators rejected mobile IAB deployments. Result: First Anritsu/Keysight test suites certified (Dec 2024).
XR QoE KPIs, L1/L2 mobility tests.
How it works: Test cases simulate 6DoF motion patterns, measure pose prediction accuracy (<2cm), rendering delay (<20ms), motion sickness probability. Validates multisensory QoS prioritization under congestion.
Problem solved: No standardized XR KPIs; vendors claimed different performance metrics. Result: Common pass/fail criteria for XR certification.
NTN/RedCap positioning, FRMCS RF tests.
How it works: Channel emulators simulate 1500km GEO delay, ±30kHz Doppler. RedCap tests verify 20MHz BW positioning (<1m accuracy). FRMCS validates ACS/SEM coexistence with GSM-R in 3MHz BW.
Problem solved: No test infrastructure for new verticals/devices. Result: First certified NTN smartphones, RedCap wearables (Q1 2025).
MBS Inactive, timing resiliency validation.
How it works: Validates Inactive UE joining/leaving MBS sessions without RRC transition. Tests HARQ combining (multicast+unicast), group key handling, session mobility across cells.
Problem solved: Rel-17 MBS testing only covered RRC_CONNECTED. Result: Battery-optimized multicast for stadiums/IoT validated.
Applications
Vendor certification (Anritsu/TTCN-3 suites for AI CSI).
Operator IOT (mobile IAB topology tests).
Implementation Aspects
Testbeds: Update TS 38.523/38.533 for L1 mobility (e.g., CHO trigger on L1-RSRP).
Example: Mobile IAB test case verifies MT handover without served UE drop, using Spirent emulators
References