Multi-Performance Coupling Optimization for Edge Computing in Light Industry Production Lines via an Optimized Memory-Hard Proof-of-Work Consensus Mechanism

Source abstract

Edge computing for light-industry production lines must reconcile inherently conflicting requirements: real-time control, bandwidth efficiency, security, energy consumption, and collaborative throughput. To address this multi-objective trilemma, we propose a three-tier perception-edge-cloud architecture grounded in an adapted memory-hard proof-of-work (ItsukuPoW) consensus mechanism. We formalize the joint optimization as a non-deterministic polynomial-time hard (NP-hard) mixed-integer problem with five coupled performance indicators. To solve it in a decentralized manner, we reformulate the problem as a potential game and design a distributed multi-performance optimization (DMPO) algorithm that provably converges to a pure Nash equilibrium. The DMPO algorithm represents the primary artificial intelligence (AI) contribution, enabling autonomous, game-theoretic resource allocation without central coordination. Evaluated on a real automated packaging line—the core engineering application—across five device configurations and nine load levels, the proposed scheme achieves, under full load (50 devices, 350 products/hour), an end-to-end response time of 181 ms (75% improvement over conventional edge computing), a data compression rate of 62% (saving 42% bandwidth), an attack cost of 64×105, an average power consumption of 3.92 W, and a production cycle of 180.2 s (21% shorter than the baseline). The comprehensive objective value F reaches 0.32, substantially outperforming baseline schemes (0.56-0.87). The compact 16 KB proof, enabled by Merkle-tree compression, directly underpins the bandwidth gains. These results demonstrate that the proposed architecture effectively harmonizes security, efficiency, and energy conservation, offering a practical and deployable solution for trusted real-time edge computing in Industry 4.0 settings.