Alternating Transmission Protocol: Receiver-Dependent Allocation of Rapid and Recurrent Processing

Abstract

# Provider Description A nervous system cannot allocate the same amount of return processing to every event. Some actions can proceed from recent evidence; some require earlier context, correction, or recurrent stabilization; and some uncertainties cannot be repaired by more computation. This paper develops the Alternating Transmission Protocol (ATP) as a source-recovered, receiver-dependent policy over rapid and return influence, controlled-loss tolerance, ordering, correction, persistence, receiver readiness, action coupling, energy, and latency. The proposal is evaluated through four progressively stronger preregistered synthetic development studies. Structured return content, receiver-state features, and conditional allocation affected outcomes, but the tested hysteretic controllers did not establish unique effectiveness. In the strongest nonlinear navigation study, the ATP candidate achieved 0.9078 exact accuracy, 0.9149 task action score, and 0.7815 registered utility. It exceeded a trajectory expected-value policy in mean utility, but failed the complete conjunctive endpoint and was outperformed by a simpler direct final expected-value policy. Every adverse result, null, tie, failed margin, and unrealized workload condition is retained. The paper also supplies a finite Lean 4 policy kernel with 30 machine-checked theorems, including explicit nonidentification boundaries; publication-quality architecture, genealogy, result, countermodel, and evidence-flow figures; and a public robot-data preflight. The public data do not contain route assignment, route-caused action change, executed action, and observed consequence in one governed episode. The decisive physical action-consequence study is therefore registered and unrun. The 52-page preprint includes 47 numbered equations, 59 cited references, five figures, a complete source genealogy, claim and strongest-form ledgers, and a reproducibility companion with code, contracts, results, proof logs, manifests, and hashes. ## Keywords neural communication; recurrent processing; feedforward processing; adaptive computation; predictive coding; action consequence; brain-computer interface; Self Aware Networks; NAPOT; Phase Wave Differential; formal verification; reproducibility