Probabilistic Trust-Based Enhancement for simultaneous transmission in AOMDV Routing Protocol

Nikhil Mishra

REVIEW 4 major objections 6 minor 2 references

Probabilistic Trust-Based Enhancement for simultaneous transmission in AOMDV Routing Protocol

T0 review · 4 major / 6 minor · reviewed 2026-08-12 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read Splitting AOMDV traffic in proportion to path trust minimizes expected delay.

desk verdict A readable course-project-style design sketch whose split rule doesn't survive the paper's own load-dependent trust model; no simulation, so the delay-reduction claim is unsupported. read the letter →

arxiv 2411.13227 v1 pith:EVZPLJTH submitted 2024-11-20 cs.NI

Nikhil Mishra This is my paper · ORCID

classification cs.NI

keywords AOMDVtrust-basedroutingBetadistributionoptimaldatadelayminimizationmobileadhocnetworksmultipath

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper tries to show that a mobile ad hoc network using the AOMDV multipath routing protocol can reduce the expected delay of a data transfer by splitting traffic across paths in proportion to each path's computed trust value. The proposal adds a probabilistic trust model, built from a Beta-distributed estimate of acknowledgment success, battery level, node availability, and hop count, to a protocol that currently chooses routes without regard to node reliability. If the derivation is right, the optimal split for two paths is $f_1 = p_1/(p_1+p_2)$, and for $n$ paths it is $f_i = p_i / \sum_j p_j$, so a node can choose fractions purely from local reliability estimates. A sympathetic reader would care because this is a parameter-free allocation rule that promises lower delay in multi-path ad hoc routing without changing the underlying route-discovery machinery.

What carries the argument

The load-bearing object is the minimax allocation rule. With the stop-and-wait expected delay $E[\text{delay}] = 2t/p$ for each path, and the assumption that a path's delay scales linearly with its assigned data fraction, the delay of path $i$ is proportional to $f_i/p_i$. Minimizing the largest of these, $\max_i f_i/p_i$, gives $f_i = p_i / \sum_j p_j$. This split is driven by the trust model: $p$ is a Beta-distributed link reliability updated by ACK outcomes, then adjusted by battery level $f(B_t)$, an availability index $A_i(t)$ that accounts for recent transmissions, and the number of hops to the destination.

What would settle it

A testbed or simulation with two AOMDV paths that share a common intermediate node, where path 1 has much higher trust than path 2, would refute the rule if the trust-proportional split produces higher end-to-end delay than a deliberately equal split under the same traffic load.

Watch

Extended reading notes

Core claim

The paper claims that when a node transmits simultaneously over multiple paths, the expected delay of the whole transfer is minimized by splitting data in proportion to each path's trust value. For two paths the optimal fraction is $f_1 = p_1/(p_1+p_2)$; for $n$ paths it generalizes to $f_i = p_i / \sum_j p_j$. This rule follows from modeling each path's delay as inversely proportional to its trust value and proportional to the amount of data sent on it, then minimizing the maximum path delay. The trust values themselves are computed from a Beta-distribution model of ACK success, battery level, node availability, and hop count, with a windowed update rule that forgets very old observations.

Load-bearing premise

The entire allocation rule rests on treating each path as an independent channel whose delay depends only on its own assigned load, so transmissions on different paths do not interfere and paths do not share bottleneck nodes.

Editorial extensions

If this is right

If correct, AOMDV nodes can choose split ratios locally from trust estimates without extra control messages, reducing expected transfer delay.
The Beta-prior update rule gives a memory-limited trust estimate that can react to node misbehavior or battery drain over time.
The minimax derivation yields a closed-form $n$-path split, so the rule can be applied at any node with multiple neighbors to the destination.
Because the derivation works for any path reliability $p$, the same proportion rule could incorporate loss rate, delay, or energy metrics into one allocation.
The scheme preserves the underlying AOMDV route discovery, so it can be layered onto existing multipath routing without redesigning the network layer.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

Editorial extension: The independence assumption is most plausible when paths are node-disjoint and spatially separated; if AOMDV paths share intermediate nodes or spectrum, the optimal split under contention would likely differ from $p_i / \sum_j p_j$, sending less traffic to a path that shares a bottleneck.
Editorial extension: A natural simulation test would compare trust-proportional splitting against equal splitting and against a predictive equal-delay operating point, measuring both mean and tail end-to-end delay under node mobility and packet loss.
Editorial extension: The proposed local rule could be ported to multipath transport protocols like MPTCP by mapping each subflow's loss or throughput to a trust value $p$, giving a congestion-independent splitting rule.
Editorial extension: The paper's own admission that local optimization can be globally suboptimal suggests that aggregating trust over full paths, rather than first-hop trust only, would likely improve the split at the cost of route-discovery overhead.

Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

Colleague,

Short version: this is a cleanly written design sketch for a trust-aware load split in AOMDV. The Beta reputation update and the stop-and-wait delay algebra are correct, and the paper is honest that it contains no simulation. But the main optimization result does not follow from the paper's own model, because the per-path success probability p_i is treated as a constant when it actually depends on the data fraction f_i through the availability index in Eq. (4). That is a load-bearing gap, not a nit.

What's new: relative to the two cited references (AODV RFC 3561 and Marina-Das AOMDV), the composite trust score in Eq. (6) and the split rule f_i = p_i / sum p_j are not present. The derivation from the delay model to that rule is algebraically correct under a fixed-p independence assumption. The paper also states its limitations clearly in the future-work list, including local-vs-global optimality and queue length.

Soft spots, in proportion: (1) The endogeneity problem is real and central. Eq. (4) degrades availability whenever a node transmits, so increasing f_i lowers A_i(t) and hence p_i. The minimax objective max(f1/p1, (1-f1)/p2) ignores this feedback, so the closed-form split is not generally the optimum. (2) Eq. (6) is an ad hoc product of four factors with unspecified parameters and no calibration or sensitivity analysis. (3) No simulation or baseline comparison appears anywhere; the conclusion explicitly says simulation is the next step, which contradicts the abstract's claim that the scheme reduces delay. (4) Related work is thin: only two references, with no engagement of the substantial trust-based MANET routing literature.

If you read the paper as a class-project report, it is internally readable and the math that is done is done correctly. As a research contribution, the central claim is unproven, and the one clear derivation rests on an assumption the paper's own model contradicts. I would not cite this in my own work, and I would not send it to a full journal as is. The fix is straightforward: redo the optimization with load-dependent p, or justify the constant-p approximation, and add at least a small simulation. Until then, it is a work-in-progress, not a published result.

Referee Report

4 major / 6 minor

Summary. The manuscript proposes a trust-based enhancement to the AOMDV routing protocol. It models link trust with a Beta distribution (Eqs. (1)-(2) and Algorithm 1), combines that with a predicted battery level, a per-node Availability Index Ai(t) that degrades during transmission, and a hop-count penalty into a composite link probability (Eq. (6)). It then derives the expected delay of a stop-and-wait transmission as 2t/p (Eqs. (7)-(8)) and argues that, to minimize the worst-case delay when splitting traffic across paths, the split should be proportional to the path trust values (Eqs. (10)-(11)). The paper is purely analytical; the conclusion explicitly defers simulation and experimental validation to future work.

Significance. If the central derivation were valid, the paper would provide a simple closed-form, decentralized load-splitting rule for multipath routing: each path receives a fraction equal to its trust value divided by the sum of trust values. The Beta expectation proof in Eq. (2) is correct, and Eqs. (7)-(11) are internally coherent under the stated stop-and-wait and delay-proportional-to-load assumptions; Eq. (10) is a genuine algebraic consequence of the minimax objective. The main value of the manuscript is therefore as a clearly stated design proposal with a falsifiable prediction. However, the connection between the heuristic trust score and a true per-transmission success probability, and the treatment of path quality as exogenous to the allocation, are not established; these gaps currently prevent the optimality claim from being accepted.

major comments (4)

[Section III-A5 and Section III-B, Eqs. (6)-(8)] The quantity p_{A→B,t} in Eq. (6) is a product of heuristic submetrics, not a well-defined per-transmission success probability. In particular, α/N is an empirical success fraction over the last N observations and differs from the Beta expected value α/(α+β) used in Eq. (2) and from the continuation of the Beta model; multiplying by f(Bt), Ai(t), and 1/num hops produces an unnormalized composite score. The recurrence in Eq. (7) requires p to be the constant probability that a single transmission attempt succeeds, so E[delay]=2t/p and the allocation result are not justified for the score defined in Eq. (6).
[Section III-C, Eqs. (10)-(11), and Section III-A3, Eq. (4)] The optimization treats p1 and p2 as fixed, but Eq. (4) makes Ai(t) decrease by d when a node transmits, and Eq. (6) includes Ai(t) multiplicatively. Increasing the fraction f_i of data assigned to path i increases transmission activity on that path, reducing Ai(t) and hence reducing p_i. Thus p_i is endogenous to the allocation f_i, and the true minimization is a fixed-point problem; the closed-form f_i = p_i / Σ_j p_j is generally not the minimizer of the actual expected delay. The paper's own Fig. 3 and the future-work item on queue lengths acknowledge that load affects path quality, which contradicts the constant-p premise.
[Section III-C, assumption before Eq. (10)] The assertion that 'Delay ∝ amount of data being transmitted' and that each path's delay depends only on its own trust value and load ignores shared-medium interference and possible overlap of AOMDV paths. Even if AOMDV returns node-disjoint or link-disjoint paths (Section II-B), simultaneous transmissions on different paths can still interfere at the physical layer or queue at shared intermediate or neighboring nodes, and these interactions are not captured by the per-path linear objective. Hence the minimax step establishes at best a conditional result for isolated, independent paths, not the 'optimal' distribution claimed in the abstract.
[Section IV, Conclusions] The paper contains no simulation or protocol-level evaluation; the conclusion explicitly states that simulating the method is 'the next step'. For a routing-protocol enhancement with several free parameters (γ, r, d, N, α0, β0) and no calibration procedure, analytical derivations under idealized assumptions are insufficient to demonstrate that the proposed scheme improves AOMDV in delay, throughput, or reliability. At minimum, a simulation study on packet delivery fraction, end-to-end delay, and overhead is needed to make the central claims testable.

minor comments (6)

[Section III-C, Eq. (11)] Eq. (11) states that n is the number of neighbors of the transmitting node, but the data-distribution problem is over multiple paths to a given destination; not every neighboring node necessarily has a valid path, so the universal quantification over neighbors is imprecise.
[Section III-A3, Eq. (4)] Eq. (4) mixes the time-scaled quantity rΔt with the per-transmission decrement d in the same updating expression; please clarify the intended time interval for one degradation decrement.
[Section III-C, Eq. (10)] The derivation of Eq. (10) is omitted; adding the one-line equalization argument (set f1/p1 = (1-f1)/p2 and solve) would improve reproducibility.
[Section III-A1, Algorithm 1] Algorithm 1 increments α and β indefinitely, but the text says α and β are determined from only the latest N observations; the algorithm does not describe the sliding-window deletion of older observations.
[Section III-A2, Eq. (3)] Eq. (3) uses two closest battery observations Bt0 and Bt1 but does not specify how these are stored, synchronized, or refreshed in a distributed MANET.
[Abstract and Section III-B, Eq. (7)] The phrase 'simultaneously transmitting through multiple paths' is in tension with the stop-and-wait model in Eq. (7), which sends a new packet only after the previous ACK; please clarify whether the scheme is concurrent-flow or interleaved-serial.

Circularity Check

0 steps flagged · score 0.0 of 10

No material circularity: the optimal data split is derived algebraically from explicitly stated delay assumptions.

full rationale

Section III-C derives the allocation rule f_i = p_i / sum_j p_j by minimizing max(f1/p1, (1-f1)/p2). This is a direct algebraic consequence of the paper's stated assumptions (stop-and-wait per packet, Eqs. (7)-(8), and Delay ∝ amount of data being transmitted), not a result that has been fitted, renamed, or imported from prior work. The trust model in Eq. (6) and the availability index in Eq. (4) do make path reliability endogenous to transmission load, so treating p1 and p2 as fixed could be questioned as a modeling approximation; however, this is a correctness or robustness limitation that the paper itself partially acknowledges in Fig. 3 and in its future-work item on queue lengths, not a circular step in which the conclusion is assumed in the premises. There are no self-citations, no imported uniqueness theorems, and no empirical predictions that reduce to fitted inputs. Accordingly, no circularity is established.

Assumptions & free parameters 5 free parameters · 5 assumptions · 2 invented entities

The central claim depends on a multiplicative trust score whose parameters (γ, r, d, N, initial α/β) are free, on a stop-and-wait delay model, and on an independence assumption between paths. No parameter is fitted to data, which keeps circularity low, but every parameter and the multiplicative structure are unvalidated. The Availability Index is an invented state variable with no external evidence. The allocation formula itself requires no additional free parameters beyond the trust values p_i.

free parameters (5)

gamma (γ) in battery influence function f(x)=1-exp(-γx)
Section III-A2 states γ is 'a chosen parameter'; no calibration method or value is given, and the final link probability Eq. (6) scales directly with f(Bt).
Availability regeneration rate r
Section III-A3, Eq. (4): r controls how fast A_i(t) recovers when idle; no value or estimation procedure is specified.
Availability degradation factor d
Section III-A3, Eq. (4): d controls how much A_i drops during transmission; no value or estimation procedure is specified.
Observation window N
Section III-A1: α and β are determined from the latest N observations; N is a design choice that changes the empirical success rate α/N used in Eq. (6).
Initial Beta shape parameters α0, β0
Algorithm 1 increments α and β but never specifies starting values; without an explicit prior the trust value is undefined at the start.

assumptions (5)

domain assumption A node's ACK success for a neighbor follows a Bernoulli process with fixed probability p, and the Beta distribution is the conjugate prior.
Section III-A1. The trust update uses Beta prior plus Bernoulli likelihood, but node behavior in a mobile ad hoc network is time-varying and correlated; this stationarity assumption is unstated.
domain assumption Retransmission follows stop-and-wait ARQ: a packet is sent only after the previous ACK is received, and each attempt takes 2t.
Section III-B, Eq. (7). This yields E[delay]=2t/p. If pipelining or windowed transmission is used, the expected delay does not scale as 1/p.
domain assumption Path delay is proportional to the amount of data sent on the path, and the paths do not interact.
Section III-C: 'assuming Delay ∝ amount of data being transmitted'. The optimization treats each path's delay as an independent linear function; simultaneous transmission over a shared wireless medium introduces contention and coupling, so independence is not guaranteed.
ad hoc to paper The trust factors (ACK history, battery, availability, hop count) combine multiplicatively and independently in Eq. (6).
Section III-A5 defines p_{A→B,t} as the product α/N·f(Bt)·A_i(t)/hops. No derivation or empirical justification is given for multiplicative independence, and the product is then treated as a transmission success probability in Eq. (7).
domain assumption Interference between neighboring nodes is adequately captured by the local Availability Index of the immediate next hop.
Section III-A3 and Fig. 3 acknowledge that local optimization can be globally suboptimal; the model still ignores multi-hop path coupling beyond the first-hop availability index.

invented entities (2)

Availability Index A_i(t)
purpose: A per-node state variable, maintained by each neighbor, meant to capture recent transmission activity and interference so that busy nodes are avoided.
Defined in Eq. (4) as an ad hoc thresholded integrator with manually chosen parameters r and d. No measurement or validation shows it corresponds to actual channel availability; it is introduced solely to make the model include coupling.
Composite link-trust probability p_{A→B,t}
purpose: A single scalar used as the success probability in the delay model and the basis for data splitting in Eq. (11).
Defined in Eq. (6) as a product of four factors. It is not validated against measured ACK success rates or actual delay; treating it as p in E[delay]=2t/p is an unverified modeling step.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Probabilistic Trust-Based Enhancement for simultaneous transmission in AOMDV Routing Protocol." pith.science (2026). https://pith.science/paper/EVZPLJTH

@misc{pith2026241113227,
  author       = {Pith},
  title        = {Pith review of: Probabilistic Trust-Based Enhancement for simultaneous transmission in AOMDV Routing Protocol},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/EVZPLJTH}},
  note         = {Machine review of arXiv:2411.13227}
}

read the original abstract

This work addresses a trust-based enhancement to the Multipath Ad hoc On-Demand Distance Vector (AOMDV) routing protocol. While AODV and its multipath variant AOMDV have been fundamental in mobile ad hoc networks, they lack mechanisms to account for node reliability. A probabilistic link-trust model is proposed that incorporates factors such as past behavior, battery levels, and node coupling to distribute data optimally to reduce delay while simultaneously transmitting through multiple paths.

Figures

Figures reproduced from arXiv: 2411.13227 by the authors.

**Figure 2.** Optimization strategy for data distribution across multiple paths [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗

**Figure 3.** Key considerations for future protocol enhancements [PITH_FULL_IMAGE:figures/full_fig_p003_3.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

2 extracted references · 2 canonical work pages

[1]

Ad hoc on-demand distance vector (AODV) routing,

C. Perkins, E. Belding-Royer, and S. Das, “Ad hoc on-demand distance vector (AODV) routing,” RFC 3561, 2003

work page 2003
[2]

Ad hoc on-demand multipath distance vector routing,

M. K. Marina and S. R. Das, “Ad hoc on-demand multipath distance vector routing,” Wireless Communications and Mobile Computing, vol. 6, no. 7, pp. 969–988, 2006

work page 2006

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.

[1] [1]

Ad hoc on-demand distance vector (AODV) routing,

C. Perkins, E. Belding-Royer, and S. Das, “Ad hoc on-demand distance vector (AODV) routing,” RFC 3561, 2003

work page 2003

[2] [2]

Ad hoc on-demand multipath distance vector routing,

M. K. Marina and S. R. Das, “Ad hoc on-demand multipath distance vector routing,” Wireless Communications and Mobile Computing, vol. 6, no. 7, pp. 969–988, 2006

work page 2006