Researchers proposed an aerial edge machine and smart reflecting panel framework to support autonomous mining robots, according to Nature.
Deep-pit mining environments suffer from severe non-line-of-sight blockage in terrestrial networks.
The hardware-constrained SRP model fits its phase-dependent amplitude response to measured varactor-tuned element data and optimizes reflection coefficients directly on a non-convex manifold.
Researchers formulated a worst-case safety-response-time minimization problem by jointly optimizing quasi-static aerial edge machine placement, smart reflecting panel passive beamforming, uplink transmit-power allocation, computation-frequency allocation, and task-partitioning ratios under statistical interference-temperature-limit constraints, as reported by Nature.
An outer bisection searches the target latency, while an inner block coordinate descent loop combines closed-form task and computation updates, primal-dual weighted MMSE power allocation, Riemannian manifold optimization, and a trust-region successive convex approximation for placement, according to Nature.
Surrogate steps are accepted only if they do not increase a common violation function to maintain inner loop monotonicity. Nature said that while global optimality is not guaranteed, the reported latency relies on a strictly checked feasible setup that serves as an upper bound to the optimal solution.
When processing a 3.0 Mbit workload, the system achieves 1.05 s compared to 6.00 s for local processing, representing an (82.5%) decrease. Nature said it preserves Jain’s fairness index higher than 0.9 alongside a fading-averaged cognitive safety headroom ranging from 2.5–12.5 dB below the limit.
The achieved latency lies within 0.05 s of the ideal-reflection bound. Nature reported that this narrow gap stems from the deployment being edge-computation-limited rather than near-optimal beamforming, with uplink transmission accounting for about (5.5%) of offloading time so that the panel acts as insurance against blockage.