In disaster-prone environments, communication networks must sustain operation under severe constraints such as limited energy, damaged infrastructure, and uncertain topology. This study compares Deep Learning (DL) and Swarm Intelligence Optimisation (SIO) as energy-aware strategies for disaster communication. While DL excels in data-rich prediction and situational analysis, its reliance on high-performance hardware and stable connectivity restricts its feasibility during real-time emergencies. In contrast, SIO provides decentralised coordination, lightweight computation, and adaptive routing, making it better suited to infrastructure-independent device-to-device (D2D) networks when central control collapses. A comparative conceptual framework was developed to evaluate both paradigms across five criteria: energy efficiency, adaptability, computational demand, response time, and scalability, based on recent literature between 2023 and 2025. Findings show that SIO demonstrates superior suitability for energy-limited and time-critical operations, while DL remains valuable for pre-disaster prediction and post-event analysis. Hybrid DL–SIO frameworks bridge both paradigms, enabling predictive–adaptive synergy across the disaster lifecycle. The study contributes a context-aware guideline for algorithm selection, shifting the focus from technology-centric performance toward environment-centric deployment in future energy-efficient, resilient, and adaptive disaster communication systems.

Clustering algorithms in device-to-device (D2D) communication require reliable benchmarking, especially during disaster scenarios where network behavior heavily depends on modeling assumptions and simulation parameters. However, many common benchmark algorithms are often reused without proper validation, despite incomplete parameter disclosures and limited implementation details in original studies. This raises concerns about fairness, reproducibility, and the credibility of performance comparisons. This study systematically reconstructs and validates popular clustering and routing algorithms for disaster-focused D2D communication, integrating them into a unified simulation framework. Six representative schemes, i.e., LEACH, EAUCA, EEFUC, FQUCR, SCRSO, and PFCRE, are reimplemented based on their original design principles, energy models, and workflows. Each is tested under simulated conditions that closely match those reported in the original references. Key metrics, such as network lifetime, data throughput, and communication latency, are measured and compared with published results to ensure behavioral accuracy. Validation shows that reconstructed algorithms closely mirror their reference behaviors, with deviations mostly between 5% and 13%, within expected stochastic simulation variation. After validation, all benchmarks are run in a standardized environment to facilitate a fair, unbiased comparison while preserving each algorithm's internal behavior. The results demonstrate that systematic benchmark validation enhances reproducibility and reduces evaluation bias in D2D clustering research. The validated implementations and the assessment framework provide trustworthy baseline references for transparent performance evaluation and support the development of robust, energy-efficient D2D communication solutions for disaster response.

Nowadays, a variety of technologies on autonomous vehicles have been extensively developed, including in-vehicle infotainment (IVI). It have been noted as one of the key services in the automobile industry. In the near future, people will be able to watch some virtual reality (VR) movies through the streaming service provided in the vehicle. However, a person sometime not tend to be joy while watching espcially when the remote controller or audio sensory controller lack of battery or too far from IVI panel. Thus, the purpose of this research is to design a scheme of real time hand gesture detection for in-vehicle infotainment system, in order to create human computer experience. In this research, the image of human palm hand will be taken by using camera for recognize the hand gesture action. This proposed scheme will recognize human gesture and convert to be computer intruction, that can be understood by IVI device. As a result, it show our proposed scheme can be the most consistent in term of accuracy and loss compared to others method. Overall, this research represents a significant step toward improving better user experience. Furthermore, the proposed scheme is anticipated to contribute significantly to the IVI field, benefiting both academia and societal outcomes.