AI-integrated non-destructive sensing technologies for road condition assessment: recent advancements, technical bottlenecks, and future trajectories

Abstract

This review presents a comprehensive analysis of recent trends in AI-driven sensing technologies for road condition assessment. Driven by the need for automated monitoring of aging infrastructure, 92 key studies published between 2020 and 2025 were analysed. The review evaluates five primary sensing modalities: vision-based systems, ground-penetrating radar (GPR), infra-red thermography, LiDAR, and mobile sensing platforms. Following a multi-stage selection workflow and a quality appraisal framework, each study was examined using a structured framework based on research objectives, methodologies, and technical results. The findings highlight a significant transition from traditional image processing to deep learning frameworks, indicating strong potential for automated pavement inspection. However, despite these technical advances, critical systemic bottlenecks remain, including edge-computing latency drops, unmodeled environmental noise, and spatial data leakage. This review identifies future research directions, establishing cross-modality operational trade-offs and a standardised four-dimensional benchmarking protocol to enhance real-world scalability and field reliability.