Adaptive cognitive control of LDPC-coded UAV radio links under electronic countermeasures

Reliable UAV radio links operating under intentional interference require adaptation mechanisms that account not only for channel quality but also for the internal dynamics of iterative decoding. This paper proposes a convergence-aware cross-layer control method for LDPC-coded UAV radio links in which decoder-convergence information is incorporated into the state used for finite-set robust configuration selection. The proposed contribution is not a new LDPC decoding or early-stopping algorithm. Instead, effective SINR and jammer-state estimates are combined with LLR variation, parity-check information, decoding latency, and energy indicators to select the modulation order, coding rate, maximum LDPC iteration limit, and convergence threshold. The control problem is formulated over a finite configuration catalogue using robust feasibility filtering, worst-case Pareto pruning, and minimax selection subject to BLER, latency, normalized energy, and minimum-goodput constraints. Performance is evaluated for IEEE 802.11n QC-LDPC codes with a codeword length of 1296 bits under AWGN and Rayleigh fading with stochastic on–off Gaussian jamming. Practical syndrome/CRC termination is used in the baseline receivers, and the contribution of decoder-convergence information is isolated through an ablation with an otherwise identical Pareto–minimax controller. Under resource-matched transmit-power and MCS conditions, convergence-aware control reduces the mean decoding iteration count by 7.72%, modeled decoding latency by 6.56%, and normalized energy per successfully delivered information bit by 4.39%. The corresponding BLER change is 0.0257% points and the goodput change is approximately 0.03%. These results indicate that decoder-convergence information primarily improves decoding-resource allocation rather than providing a large additional reliability gain. The conclusions are limited to the evaluated channel, coding, and jammer models; hardware implementation and robustness to unmodelled interference remain subjects for further study.