Journal of Ad-hoc Network and Mobile Computing (e-ISSN: 3048-9180)

Mobility- and Energy-Aware Next-Hop Selection for Reliable Routing in Mobile Ad Hoc Networks: A Reproducible Simulation Study

2026-11-05T06:25:11+00:00

Background: Routes in mobile ad hoc networks break because nodes move, and they degrade because relays exhaust their batteries or become congested. Hop-count routing, as used by the Ad hoc On-demand Distance Vector (AODV) protocol, ignores all three effects.

Objectives: This study proposes Mobility-, Energy-, and Queue-aware Routing (MEQ-R) and quantifies its benefits and costs against established alternatives.

Methodology: MEQ-R assigns each link a cost that combines predicted link-expiration time, the residual-energy ratio, queue occupancy of the next hop, and hop count, and it rebuilds a route preemptively when its weakest link is about to expire. It was compared with minimum-hop routing, minimum battery-cost routing, and a stability-based scheme in a packet-level simulator with random-waypoint mobility, a first-order energy model, and a fluid queue model, over 1,100 paired runs that varied speed, node count, offered load, and metric weights.

Results: In the reference scenario of 50 nodes moving at up to 10 m/s, MEQ-R raised the packet delivery ratio from 95.9% to 98.7% (paired Wilcoxon p below 0.0001), reduced route failures by 83% and lowered routing overhead by 20% relative to minimum-hop routing, at a cost of 6% more energy per delivered packet. Route failures fell by 45–95 % in every scenario, and the overhead reduction reached 51% in dense networks. The energy term reduced relaying through depleted nodes and the queue term reduced overflow losses under heavy load. MEQ-R gave no benefit in near-static or partitioned networks.

Conclusions: Link-lifetime prediction with pre-emptive maintenance is the key mechanism for reliable routing under mobility.

Machine Learning-Driven Routing in the Internet of Vehicles: A Critical Survey of Learning Paradigms, Gaps, and Open Problems

2026-10-04T10:07:05+00:00

The Internet of Vehicles (IoV) has evolved rapidly from vehicular ad hoc networking into a large-scale, heterogeneous system integrating vehicles, roadside infrastructure, edge servers, and cloud platforms in support of intelligent transportation. A persistent bottleneck in this evolution is routing: selecting reliable multi-hop paths across a topology that reconfigures on the order of seconds due to high vehicle mobility, variable traffic density, and unpredictable link quality. Traditional topology-based, position-based, and cluster-based routing protocols encode fixed heuristics that react to connectivity loss only after it occurs, a poor structural match for the pace of change in IoV. This has driven more than a decade of research applying machine learning (ML), tabular reinforcement learning, deep reinforcement learning, graph neural networks, federated learning (FL), and metaheuristic-assisted clustering to make routing decisions predictive rather than reactive. This study presents a comprehensive survey of ML-based routing protocols for IoV, drawing on more than seventy studies published largely between 2020 and 2025. The study first characterizes the IoV architecture and the specific challenges that high mobility imposes on routing. The literature is organized into ten methodological categories: reinforcement learning, deep reinforcement learning, graph neural networks, hybrid DRL + GNN architectures, FL, software-defined networking integration, blockchain-assisted trust management, mobility and traffic prediction, digital-twin-assisted edge computing, bio-inspired clustering, and multi-agent reinforcement learning, offering a critical analysis of the strengths, weaknesses, and evaluation practices of representative protocols in each category rather than a simple enumeration. A comparative analysis distills the trade-offs across paradigms along latency, packet delivery ratio, overhead, scalability, and deployability. It identifies open research issues, including inconsistent convergence-time reporting, limited cross-scenario generalization, training–inference asymmetry on resource-constrained onboard units, and the near absence of standardized benchmarks. Future research directions are outlined toward hybrid prediction–decision architectures, privacy-preserving FL at scale, and trustworthy, explainable routing decisions. The survey is intended as both a reference map of the field and a foundation for designing the next generation of learning-based IoV routing protocols.

Mesh-Routed Deferred Settlement: A Proof-of-Concept Architecture for Offline Digital Payments

2026-10-27T11:10:54+00:00

Digital payment systems such as India’s Unified Payments Interface (UPI) require continuous connectivity at the moment of payment, an assumption that fails in basements, remote regions, and during network outages. This study describes a proof-of-concept implementation of Mesh-Routed Deferred Settlement, focusing on offline payment creation, hop-by-hop message relay, and deferred backend settlement after connectivity is restored. Three threats in this setting are addressed: loss of confidentiality and integrity through intermediate carriers, mitigated by hybrid RSA-OAEP and AES-256-GCM encryption; duplicate settlement caused by concurrent uploads, prevented through an atomic idempotency claim based on the SHA-256 hash of the ciphertext; and replay of captured packets, mitigated by a bounded freshness window and per-payment nonces. Settlement runs in one atomic transaction with optimistic concurrency control and a database uniqueness constraint as a backstop. A Spring Boot backend with a software-simulated mesh reproduces the full packet lifecycle, and automated tests cover encryption correctness, tamper rejection, and exactly-once settlement under concurrent delivery. The central limitation of deferred settlement is that, because the backend cannot verify the payer’s balance during offline transit, multiple offline payments may collectively exceed the available funds.