Topology Construction Algorithms for DECT NR+ in Normal and Ultra-Dense Deployments

I. INTRODUCTION

• Ultra-dense deployments crowd the interference domain and offer abundant association choices, challenging distributed association and relay selection.The study targets topology construction in both normal and ultra-dense conditions with four enhancements that do not modify DECT NR+ MAC procedures.Source: §I · p. 1
• The proposed suite combines load-dependent relay allocation, hysteresis- and blacklist-based association control, and smart beacon placement.The mechanisms relax relay capacity constraints, slow association decisions, and reduce beaconing interference.Source: §I · p. 1
• Proposed mechanisms keep reassociations below 3 events/s in systems up to 700 nodes and reduce active relays from 34% to 15%, versus more than 10 events/s without optimization.The relay reduction is described as more than half.Source: §I · p. 2

A. DECT NR+ Systems

• DECT NR+ uses a cluster-tree topology in which radio devices act as fixed or portable terminations, with fixed terminations relaying traffic toward the sink.Fixed terminations also coordinate local radio resources and anchor portable-termination connections.Source: A. DECT NR+ Systems · p. 2

B. Related Work

• Prior work examined capacity and near-URLLC performance, but topology studies were scarce; addressed instability with dynamic relay selection and hysteresis.This study focuses on topology construction for ultra-dense deployment conditions.Source: B. Related Work · pp. 2–3

A. Association and Relay Selection in DECT NR+

• Topology construction proceeds through UE association, selection of PT or FT mode, and beacon scheduling by FT-mode devices.FT-mode devices act as relays, coordinating local resources and allowing other RDs to connect.Source: A. Association and Relay Selection in… · p. 3

1) UE Association Procedure:

• An RD scans for FT beacons, filters candidates by link quality and capacity, compares route criteria, then requests association over RACH.If a request fails, it may try another candidate or wait for a later opportunity.Source: 1) UE Association Procedure · p. 3

2) Relay Selection:

• An RD may switch from PT to FT after connecting upstream; mode switching considers connectivity, power, and traffic needs.Before switching, it checks its FT uplink and monitors neighboring FT devices against a configurable threshold.Source: 2) Relay Selection · p. 3

3) Limitations of Existing Approaches:

• Local fluctuations in node density can require dynamic FT selection, which the conventional relay procedure does not account for.The conventional approach does not account for the number of nearby nodes when choosing a relay.Source: 3) Limitations of Existing Approaches · p. 4
• Dense deployments can cause short-term reassociations from interference and long-term next-hop switching as slightly better routes become available.The paper identifies a need to suppress short-term churn while retaining the ability to choose better sink routes.Source: 3) Limitations of Existing Approaches · p. 4
• With hundreds of neighbors, random FT beacon times can overlap and suppress beacons; association and data transmission may also coexist.These conditions can increase reassociation likelihood and negatively affect both topology construction and data transmission.Source: 3) Limitations of Existing Approaches · p. 4

B. Load-Dependent Relay Selection

• The method supplements the fixed candidate-count threshold with an estimate of local association capacity.An associated RD scans for at least one beacon interval and stores viable FT candidates; it also gathers occupied and total capacity from neighboring FTs.Source: B. Load-Dependent Relay Selection · p. 4
• An RD selects FT mode when candidate count is below Tth or occupied-load fraction ρ exceeds γ; otherwise it remains PT.The decision uses the candidate-count threshold and the preset free-slot fraction threshold.Source: B. Load-Dependent Relay Selection · p. 4
• The scheme dynamically selects relays according to current load conditions near a node.Its stated purpose includes enabling already-associated RDs to propagate coverage further where UE geometry permits.Source: B. Load-Dependent Relay Selection · p. 4

C. Hysteresis and Bias-Based Association Control

• Under ultra-dense conditions, small RSSI variations can repeatedly change the selected upstream FT when candidates have comparable hop counts and beacon powers.Such changes increase signaling overhead and may disturb topology construction.Source: C. Hysteresis and Bias-Based… · p. 5
• Received-power bias filters weak candidates, while hysteresis requires a stronger RSSI before switching to an equal-hop candidate.A candidate with a shorter path to the sink may still be preferred, preserving the ability to switch to better targets.Source: C. Hysteresis and Bias-Based… · p. 5
• A candidate replaces the current FT if it has fewer hops, or if it has equal hops and RSSI greater than the current FT by margin h.The candidate must also satisfy the minimum RSSI threshold incorporating a bias above the sensitivity limit.Source: C. Hysteresis and Bias-Based… · p. 5

D. Smart Beacon Randomization

• When many devices start together, aligned association searches and FT beacons can cause access bursts, beacon overlaps, missed candidates, and reassociations.These conditions may occur after blackouts or during mass deployment.Source: D. Smart Beacon Randomization · p. 5
• Randomized association and beacon start times spread activity in time, while conflict-aware placement avoids observed neighboring FT transmission windows.The described effects include reduced beacon interference, lower association contention, and fewer reassociation waves from transient beacon loss.Source: D. Smart Beacon Randomization · p. 5
• An associating device delays its search by a random jitter offset drawn uniformly over the association contention window.The association start time is the boot or wake-up time plus that offset.Source: D. Smart Beacon Randomization · p. 5

1 Phase 1: Randomized association start

• An RD operating in FT mode enters the randomized-association-start procedure.Source: §1 · p. 6

5 Phase 2: Conflict-aware beacon placement

• Neighbor beacon timestamps and advertised periods are observed for at least one local beacon cycle to predict their recurring transmission windows.Source: §5 · p. 6
• If no suitable candidate beacon position is available, the procedure generates a random beacon jitter.Source: §5 · p. 6
• The FT-mode RD selects a candidate time within its local beacon cycle that maximizes the minimum temporal distance from predicted neighboring beacon windows.Source: §5 · p. 6

• The blacklist temporarily excludes association targets that recently produced unsuccessful or undesirable behavior, complementing hysteresis without preventing later route discovery.Source: E. Association Target Blacklisting · p. 6
• Blacklist triggers include disappearing or stale targets, unanswered association or renewal attempts, explicit rejection, and the previous FT after a successful handoff.Source: E. Association Target Blacklisting · p. 6
• The active blacklist stores FT identifiers with expiration timestamps, allowing excluded targets to become available again after the transient condition passes.Source: E. Association Target Blacklisting · p. 6

1 Step 1: Expire old blacklist entries

• Expired blacklist entries are removed before the algorithm evaluates candidate-target events.Source: §1 · p. 7

5 Step 2: Evaluate trigger events and update blacklist

• The blacklist is updated when a selected FT disappears, an association or renewal attempt times out or is rejected, an association becomes stale, or the RD changes FT.Source: §5 · p. 7, §16 · p. 7

16 Step 3: Filter candidate set

• The viable candidate set is filtered by excluding targets present in the active blacklist.Source: §16 · p. 7

IV. SYSTEM AND EVALUATION

• The evaluation describes its simulation tool and modeling approach, deployments, radio models, performance metrics, and data collection.Source: §IV · p. 7

A. Simulation Tool and Modeling Approach

• WINTERsim models radio devices and their functional stack, using a physical-layer abstraction that tracks bidirectional path loss and processes signals above the noise floor.Source: A. Simulation Tool and Modeling Approach · p. 7
• The simulation models uplink traffic from portable terminations to the sink and explicitly includes listen-before-talk with binary exponential backoff.Source: A. Simulation Tool and Modeling Approach · p. 7

B. Considered Deployment

• The study models ultra-dense conditions with free-space path loss, keeping node density and per-node load the same for comparison with normal conditions.Source: B. Considered Deployment · p. 7
• The single-sink deployment varies network size from 200 to 700 nodes, with the sink at the origin and coverage limited to a 170 m radius.Source: B. Considered Deployment · p. 7

C. Propagation and Antenna Models

• Free-space path loss represents unobstructed propagation, while the devices use single-element omnidirectional antennas in transmission mode 1.Source: C. Propagation and Antenna Models · p. 8
• The normal-condition propagation model uses 3GPP-based UMa, InH, or UMi path loss according to whether a link involves the sink and its node distance.Source: C. Propagation and Antenna Models · p. 8

D. Metrics and Data Collection

• Topology metrics include associated-node totals, association and reassociation rates, mean hops to the sink, and active relays; data metrics include mean and 95th-percentile delay and PDR.Reassociation should allow better-path discovery without overloading the network with signaling and interference.Source: D. Metrics and Data Collection · p. 8
• Ten independent runs per configuration use different random seeds, with steady state declared when EWMA reassociation-rate variation falls below 0.01 events per second.The evaluation uses transient simulations and independent replication.Source: D. Metrics and Data Collection · p. 8

A. Topology Construction Phase

• Load-dependent relay selection addresses varying network sizes, while hysteresis, blacklisting, and smart beaconing target ultra-dense conditions; subsequent results keep relay allocation enabled.The results section considers topology convergence, association and reassociation, active relays and hops, then data-phase performance and sleeping patterns.Source: §V · pp. 8–9, A. Topology Construction Phase · p. 9

1) Load-Dependent Relay Selection:

• All nodes connect in the 400- and 600-node mixed-propagation systems with load-dependent relay allocation enabled, while geometric placement leaves 5–8% disconnected at 200 nodes.The mechanism elects relays dynamically based on local load.Source: 1) Load-Dependent Relay Selection · p. 9, Figure 7 · p. 9

2) Normal Conditions: Mixed Indoor-Hall/Outdoor Model:

• In the mixed-propagation 700-node case, peak association rate reaches 25 per second without the mechanisms versus approximately 8 per second with them; enabling them completes association approximately 200 seconds later.The passage attributes the delay to blacklisting, which freezes nodes for a period to penalize frequent reassociations.Source: 2) Normal Conditions: Mixed… · p. 10
• With mechanisms enabled in the mixed-propagation model, large reassociation peaks disappear, but occasional reconfiguration still allows discovery of better paths to the sink.In both enabled and disabled cases, reassociation peaks coincide with association-rate peaks.Source: 2) Normal Conditions: Mixed… · p. 10

3) Ultra-Dense Conditions: FSPL Model:

• Under FSPL, reassociation can exceed 10 per second at 500 and 600 nodes without the mechanisms, while enabled mechanisms keep the 700-node rate around 2–3 per second.For smaller systems, the enabled-mechanism rate stays below 1 reassociation per second.Source: 3) Ultra-Dense Conditions: FSPL Model · p. 11

4) Number of Hops:

• With mechanisms enabled, mean hops to the sink remain approximately 9 for mixed propagation and decrease to four under FSPL; the system continues converging toward shorter routes after initial topology construction.Worse propagation in the mixed model forces longer multi-hop routes.Source: Figure 12 · p. 10, 4) Number of Hops · p. 11

5) Number of Relays:

• At 700 nodes under FSPL, active relays fall from around 240 (approximately 34% of nodes) to 100 (approximately 15%) when the other mechanisms are enabled.At 500 nodes, the relay fraction is 10%.Source: 5) Number of Relays · p. 11, Figure 13 · p. 11

B. Data Transmission Phase

• With all mechanisms disabled under FSPL, PDR drops sharply at startup and takes more than 10 minutes to plateau, alongside spikes in mean and 95th-percentile packet delay.The passage attributes the PDR drop to traffic colliding with beacons and association requests after nodes establish sink routes.Source: B. Data Transmission Phase · p. 11, Figure 14 · p. 12
• With all mechanisms enabled under FSPL, PDR and latency change smoothly at startup, with only a slight absolute advantage over the disabled case at 10 packets per second.The passage attributes the smoother behavior mainly to delayed reassociations from blacklisting and reduced association interference from smart beacon positioning.Source: B. Data Transmission Phase · p. 11, Figure 15 · p. 12
• Under FSPL, mean and 95th-percentile latency are similar to mixed propagation, but arise mainly from interference-driven retransmissions rather than the mixed model's longer routes.Mean hops are four under FSPL, while the mixed-propagation model has approximately nine.Source: B. Data Transmission Phase · p. 12

C. System with Periodically Sleeping Nodes

• With asynchronous sleeping patterns under FSPL and 700 nodes, all nodes connect by approximately 200 s; PDR reaches steady state in approximately 200 s, versus around 400 s with all mechanisms activated.Steady-state PDR and delay values are similar to those without periodic sleeping.Source: C. System with Periodically Sleeping… · p. 12

VI. CONCLUSIONS

• Proposed mechanisms maintain reassociation around 2–3 per second at 700 nodes and reduce active relays from 34% to 15%; without them, PDR recovery takes over 10 minutes.The conclusions state that the joint mechanisms avert PDR drops and substantial latency spikes associated with unstable association dynamics.Source: §VI · p. 13