BlockLLM-E2MR: Blockchain-attested LLM-guided energy-efficient multipath routing for mobile ad hoc networks

Abstract

Route decisions in mobile ad hoc networks must consider preserving battery energy, resisting forwarding manipulation as well as being tolerant of mobility. Here we propose BlockLLM-E2MR, a blockchain-verified large-language-model-guided multipath routing framework where the language model cannot directly implant a route. Paths violating residual-energy and trust constraints first turn infeasible due to a deterministic feasibility gate, then the surviving paths are ranked with respect to a frozen LLM preference vector, followed by only storing the route digest and trust update as determined by a permissioned committee. Routing objective includes transmission energy, expected delay, posterior misbehavior risk, mobility exposure, residual-energy imbalance distance between any 2 nodes in same cluster and also non-linear interaction penalties. A 1000 m × 1000 m region with 60 nodes configured for mobility between 1–10 m/s, 4096-bit packets, and will execute over independent runs (N = 24) where a fraction of the nodes have been corrupted (ft= [0%,30%]). In an environment with 30% malicious nodes, BlockLLM-E2MR achieves a packet-delivery ratio of 46.27% and consumes 66.270 mJ per offered packet; compared to the stated model, this is a delivery improvement of 8.41 percentage points combined with an energy reduction of 1.35%. Both delivery (p=0.00108) and energy (p=8.07e-08) are impactful hence, have their pairwise comparisons done against the blockchain-based trust routing still significant in favour of delivery and energy respectively. The study can be independently reproduced by a worked example of route selection, complete parameter table, confidence intervals and ablation logic are explained here along with complexity bounds, ledger break-even inequality.