Age-of-Information Dependent Random Access for Massive IoT Networks

I. INTRODUCTION

• Freshness strongly influences the accuracy of decisions made from IoT measurements at information fusion nodes.The network sends sensed data to fusion nodes for processing, such as control decisions or remote source reconstruction.Source: §I · p. 1
• Throughput and delay do not adequately measure information freshness, and optimizing either does not necessarily deliver the freshest data to receivers.AoI measures destination-side freshness and depends on both packet transmission frequency and delay.Source: §I · p. 1
• Prior approaches include age-independent randomized policies –, round-robin scheduling, and carrier-sensing schemes, each with stated limitations for this problem.The age-independent policies do not use instantaneous AoI; round-robin does not accommodate changing node counts, and leaves parameter design unaddressed.Source: §I · p. 1
• The paper introduces threshold-based ADRA, where devices transmit with a fixed probability only when instantaneous age reaches a threshold, and derives a closed-form average AoI analysis.Simulations validate the analysis and report better performance than conventional age-independent access and the scheme in [21].Source: §I · p. 2

II. SYSTEM MODEL

• N devices report status to an access point over a common wireless channel using equal-duration slots, with each update taking one slot.The model uses a slotted ALOHA-like random-access protocol.Source: §II · p. 2
• An active device samples fresh information and generates its update at the beginning of the slot, following the generate-at-will model.An inactive device remains idle for that slot.Source: §II · p. 2
• A transmission succeeds when exactly one device is active; simultaneous activity by multiple devices causes a collision and failure.The model is interference-limited, so collisions are the only stated cause of update failure.Source: §II · p. 2

A. Average AoI

• Fig. 1 depicts one possible trajectory of device Di's instantaneous AoI across time slots.The illustrated trajectory starts at age one and spans 14 consecutive slots.Source: Figure 1 · p. 2, A. Average AoI · p. 2
• A successful update resets device i's instantaneous AoI to one; otherwise, its age increases by one each slot.A reset occurs when device i is active and every other device is inactive.Source: A. Average AoI · p. 2
• The section defines average AoI for each device based on its instantaneous AoI evolution.The supplied passage introduces the definition but does not state its full expression.Source: A. Average AoI · p. 2

B. Age-Dependent Random Access

• Devices with smaller AoI access less often under the age-dependent policy, giving devices with larger AoI more opportunity to update with fewer collisions.This is the motivation for the policy's expected network-wide AoI reduction.Source: B. Age-Dependent Random Access · p. 2
• In the threshold-based ADRA variant, each device uses access probability p when its AoI is at least threshold δ and otherwise stays inactive.The CAP vector assigns an active probability for each instantaneous AoI value.Source: B. Age-Dependent Random Access · pp. 2–3
• The paper restricts the CAP vector to entries of either 0 or p and leaves the general CAP case for future work.By symmetry, all devices have the same statistical average AoI in the considered model.Source: B. Age-Dependent Random Access · pp. 2–3

• In threshold-based ADRA, each device’s success probability depends on all devices’ instantaneous AoI, coupling their evolution and making multidimensional Markov-chain analysis exponentially costly as N grows.The existing AIRA analysis instead assumes a success probability independent of the other devices’ instantaneous AoI.Source: §III · p. 3
• The analysis decouples devices by approximating their success probability q as constant at transmission, while allowing q to depend on threshold δ and access probability p.Under this approximation, each device follows an identical DTMC characterized by δ, p, and q.Source: §III · p. 3
• Simulations confirm that the adopted approximation is accurate, especially when the number of nodes N is large.The paper notes that the slotted system’s AoI follows a staircase rather than the sawtooth evolution described in prior work.Source: §III · p. 3

• For AoI below δ, the device’s age advances by one; at or above δ, it accesses with probability p and otherwise advances.A successful update requires the other devices to remain inactive, causing the device’s AoI to reset to one.Source: A. DTMC of Each IoT Device · p. 3
• The stationary distribution assigns equal probabilities to AoI states 1 through δ, followed by a tail for states above δ.The stationary probabilities are used to relate the transmission probability η to the successful-update probability q.Source: A. DTMC of Each IoT Device · p. 3
• For δ > 1, the analysis explores p > 1/N by assuming p ≤ 2/N, since fewer than N devices are likely to compete in a slot.The paper states that this setting can yield lower average AoI than δ = 1 and p = 1/N.Source: A. DTMC of Each IoT Device · p. 4
• The analysis finds q numerically from an equation whose solution is unique, then uses q to obtain the stationary AoI distribution.The cited lemmas establish a feasible range, monotonicity, and uniqueness for solving the equation.Source: A. DTMC of Each IoT Device · pp. 4–5
• Fig. 3 presents stationary distributions of instantaneous AoI for different values of N, p, and δ.The paper reports that simulated distributions are close to their analytical counterparts for all simulated cases.Source: Figure 3 · p. 5, A. DTMC of Each IoT Device · p. 5

• Increasing p or decreasing δ can reduce average AoI directly, but can also increase collisions and lower the successful-update probability q.The resulting optimal p and δ are found by a two-dimensional search over the derived expression.Source: B. Analysis of the Average AoI · p. 5
• When δ = 1, the average-AoI expression reduces to the conventional AIRA expression given earlier.Source: B. Analysis of the Average AoI · p. 5

IV. SIMULATION RESULTS AND DISCUSSIONS

• Simulation and analytical stationary AoI distributions are close across all simulated cases, especially when N is relatively large and p is relatively small.That regime is where the fixed-success-probability approximation is reported to be accurate.Source: §IV · p. 5
• An optimal threshold δ minimizes average AoI in all simulated cases, and the optimal δ decreases as N and p decrease.Increasing δ reduces devices’ active probability but also reduces collision probability; similar trends are reported for p.Source: §IV · p. 5, Figure 4 · p. 6
• ADRA achieves significantly better average AoI than AIRA and Algorithm 2 in [21], with performance gaps widening as the number of devices increases.The reported comparison optimizes δ and p for ADRA; the existing AIRA active probability is set to 1/N. The search is limited to p < 2/N, yet performs better than p=1/N, δ=1.Source: §IV · pp. 5–6, Figure 5 · p. 6

V. CONCLUSIONS

• The analysis approximates coupled AoI evolutions by modeling each device’s AoI as a discrete-time Markov chain with a fixed successful probability.Source: §V · p. 6
• Simulations verify the approximation’s tightness and the correctness of the theoretical analysis.Source: §V · p. 6
• The proposed threshold-based ADRA outperforms the state-of-the-art age-oriented random access scheme in all simulated cases.Source: §V · p. 6