The proposed BlockE2T-MORL offers a scalable, privacy-preserving, and computationally lightweight solution for next-generation intelligent path planning in cloud-based autonomous systems.
Abstract
Path planning for cloud based autonomous systems such as smart transportation, Internet of Things (IoT) installations and robot fleets need to be secure, energy efficient, time efficient and fulfil privacy constraints. Current reinforcement learning (RL) techniques mainly consider optimisation of single objective, centralised or loosely secured model updates which are susceptible to data poisoning, privacy breach and adversarial model updates. We present Blockchain enabled Energy and Time efficient Multi Objective Reinforcement Learning (BlockE2T MORL) a new decentralised approach for secure, cloud assisted path planning. BlockE2T MORL has three main components: (i) a dynamic multi objective reward function that reduces energy, travel time and security threat; (ii) a lightweight blockchain inspired trust mechanism that assigns continuous trust values to agents, and is incorporated in the reward function to punish dishonest or malicious agents; and (iii) a hybrid actor critic learning strategy that facilitates exploration and exploitation in dynamic environments. Unlike conventional blockchain systems, our approach incurs low computational overhead ([Formula: see text] instead of [Formula: see text] for validation) and operates without heavy consensus protocols. We evaluate BlockE2T-MORL on a simulated grid-based cloud environment with up to 100 agents and varying adversarial ratios. After 200 training episodes (5 independent runs), the proposed framework achieves: (i) energy consumption = 124.3 ± 8.7 J (23.4% reduction vs. standard RL, p < 0.01), (ii) latency = 45.2 ± 3.8 ms (17.4% improvement, p < 0.05), and (iii) trust score = 0.87 ± 0.04 (67% improvement, p < 0.001). The framework converges faster (210 ± 25 episodes in the final optimized configuration, compared with ≥ 520 episodes for baseline methods). BlockE2T-MORL offers a scalable, privacy-preserving, and computationally lightweight solution for next-generation intelligent path planning in cloud-based autonomous systems.
Mobile Cloud Computing (MCC) enhances resource-constrained mobile devices by enabling task offloading to edge and cloud environments, but faces challenges in security, scalability, and dynamic resource management. This study proposes a secure, blockchain-enabled federated deep learning framework integrated with Hierarchical Adaptive Scalable Spiking Reinforcement Learning (HASSRL) for intelligent and privacy-preserving task offloading. A hybrid optimization approach combining Revolution Optimization Algorithm (ROA) and Proximal Policy Optimization (PPO) is introduced to achieve adaptive scheduling, efficient resource allocation, and real-time decision-making. The framework incorporates continuous monitoring and a feedback-driven self-optimization mechanism to improve latency, energy consumption, and overall system performance. Experimental evaluation using standard datasets demonstrates that the proposed model achieves high accuracy and efficiency, making it suitable for scalable and real-time MCC applications in next-generation networks. The proposed Blockchain-enabled Federated Deep Learning with HASSRL and ROA–PPO achieved a latency of 48.689 ms, energy consumption of 63.222 mJ, and CPU utilization of 95.705%, while maintaining a throughput of 42.067 Mbps and task completion time of 2.781 s. Through experimental evaluation on the UNSW-NB15 and CSE-CIC-IDS2018 datasets, the framework demonstrated efficient task execution, improved resource utilization, and enhanced adaptability in dynamic MCC environments, ensuring secure, scalable, and real-time intelligent task offloading
Cheruku Poorna, Venkata Srinivasa, Rao Research et al.· International Conference Com...· 0 citations
The Energy-Aware Hierarchical Green Fog framework is presented, which introduces a unified reinforcement learning (RL) orchestration layer that explicitly incorporates residual energy, renewable energy availability, spatial proximity, and task deadlines into hierarchical fog-cloud decision-making.
M. Harandi, Afshin Yaghoobi· Scientific Reports· 0 citations
BELS-IoT is proposed, a novel decentralized protection architecture that integrates a cryptocurrency-based blockchain layer with a multi-layer ensemble learning engine that rewards honest behavior and penalizes malicious activities while maintaining privacy through federated learning with blockchain-verified reputation scores.
Anwar Kalghoum, Leila Azouz Saidane· SN Computer Science· 0 citations
The rapid expansion of Internet of Things (IoT) ecosystems has intensified the need for distributed intelligence mechanisms that reduce direct raw-data exposure while remaining resilient to adversarial manipulation. Federated learning (FL) addresses part of this challenge by enabling collaborative model training without centralizing raw data, but it remains vulnerable to malicious client behavior, particularly model poisoning attacks that can substantially degrade global model quality. This paper investigates a blockchain-assisted, trust-aware FL framework for privacy-aware distributed intelligence in resource-constrained IoT systems, where the blockchain layer is used to support trust coordination, auditability, traceability, and tamper-resistant metadata recording rather than to directly improve predictive performance. The empirical study compares six aggregation strategies: FedAvg, coordinate-wise Median, Trimmed Mean, FLTrust, trust-aware weighted aggregation, and a hybrid trust-trimmed mean method. The primary evaluation is conducted on the UCI Human Activity Recognition (UCI HAR) dataset under Dirichlet-based non-IID client partitioning α = 0.1 and α = 1.0, partial client participation, and sign-flip model poisoning. Each configuration is evaluated over five independent runs. Under the severe 40% malicious-client stress test, FLTrust achieves the strongest mean robustness among the evaluated methods, reaching 0.4195 ± 0.1295 accuracy and 0.3108 ± 0.1294 macro-F1 for α = 0.1, and 0.6472 ± 0.0725 accuracy and 0.6014 ± 0.0949 macro-F1 for α = 1.0. In lower-intensity attack controls with 10% and 20% malicious clients, the trust-aware and hybrid trust-trimmed strategies are the most competitive, achieving the highest or near-highest mean performance without requiring a clean server-side reference set. Mechanism-level analysis shows that FLTrust is particularly effective in the severe setting because its reference-based scoring assigns near-zero weights to malicious clients, whereas trust-aware and hybrid methods rely on relative update consistency and are more affected by the interaction between poisoning and statistical heterogeneity. Globally, the results indicate that no single aggregation rule dominates all adversarial regimes. Instead, reference-based trust provides strong protection under severe poisoning, while trust-aware and hybrid aggregation offer competitive root-free alternatives under lower attack intensities. The findings also clarify that the privacy-preserving scope of the framework derives from FL-based data locality and does not constitute a formal cryptographic privacy guarantee.
M. Reis, Carlos Serôdio, Frederico Branco· Electronics· 0 citations
In modern power systems, the integration of distributed energy resources creates challenges for coordinated control, operational stability, and secure energy transactions, particularly under extreme climate conditions. This work presents a simulation-based proof-of-concept framework comprising a multi-agent reinforcement learning (MARL) coordinator and a permissioned blockchain layer, designed to enhance transparency, resilience, and security in microgrid operations. The hybrid Convolutional Neural Network–Long Short-Term Memory forecasting model predicts load, solar generation, and electricity prices, guiding a multi-agent Proximal Policy Optimization controller. Autonomous agents manage battery storage, generator operation, and grid power exchanges. Energy transactions are represented and validated within a simulated permissioned blockchain layer. The performance of the framework is evaluated within a validated digital-twin microgrid simulation environment, featuring a 350 kW PV array and a 300-kWh battery system, including a 48-h simulated heatwave to assess climate resilience. Compared with rule-based control and MPC, the proposed system demonstrates reductions in operational cost up to 22.8% and improvements in energy efficiency by 24.7%, while achieving high reliability under simulated conditions (99.92%). Under climate stress, the framework achieved 99.1% load satisfaction with only an 18% cost increase. The blockchain layer introduces negligible overhead (<0.5% energy use). Technoeconomic analysis indicates economic feasibility, yielding a 28.5% annual return and a 30-month payback period. Overall, the proposed framework offers a scalable simulation-based approach for investigating secure and climate-resilient smart energy management strategies under complex operating conditions.
Maan Saker, G. Fandi, Z. Muller et al.· Energy Exploration & Exp...· 0 citations
The increasing usage of distributed renewable energy resources has rapidly increased the transition to decentralised smart grids. This is where secure and efficient peer-to-peer (P2P) energy trading is crucial. However, conventional blockchain-based energy trading schemes suffer from high computational overhead, communication latency, and limited scalability, making them unsuitable for resource-constrained Internet of Things (IoT) networks. This research proposes a lightweight blockchain framework that incorporates Hyperledger Fabric with Practical Byzantine Fault Tolerance (PBFT) consensus, ZigbeePro communication, and Long Short-Term Memory (LSTM)-based energy demand forecasting to facilitate secure and intelligent decentralised energy trading. The framework was evaluated using MATLAB/Simulink simulation, NS-3, Hyperledger Fabric, and a Raspberry Pi/ESP32 prototype. The results of the experiment show that the proposed framework achieved an average latency of 48.9 ms, throughput of 185 transactions per second, packet delivery ratio of 97.8 percent, and support for up to 250 IoT nodes while maintaining low energy overhead. The LSTM forecasting model attained an R² of 0.964 with a MAPE of 4.7 percent, delivering accurate demand prediction for intelligent energy allocation. Compared with centralised and Proof-of-Work blockchain models, the proposed framework enhanced communication efficiency, scalability, and security while reducing computational cost. These results demonstrated that integrating lightweight blockchain, low-power communication, and Artificial Intelligence-based forecasting provides a practical and scalable solution for decentralised energy trading for the next-generation smart grids.
Aliyu Musa Kida, C. Ngene, Jafaru Usman et al.· International Journal of Inn...· 0 citations