SHM-Traffic: DRL and Transfer learning based UAV Control for Structural Health Monitoring of Bridges with Traffic

TL;DR

The paper tackles the problem of performing structural health monitoring on bridges while traffic is present by framing the UAV crack-survey task as a Markov Decision Process and solving it with proximal policy optimization (PPO). It rigorously compares two crack-detection strategies—Canny edge detection and a transfer-learning CNN—finding that CNN achieves higher crack-detection accuracy ($99.6\%$) while Canny offers faster per-step processing. The results show CNN yields up to $12\%$ better damage detection and 1.8x higher rewards, whereas Canny reduces task completion time by up to 40\% in some scenarios, highlighting a trade-off between speed and accuracy. The work demonstrates a practical, traffic-resilient SHM framework and points to future directions including multi-UAV deployments and digital-twin bridge representations.

Abstract

This work focuses on using advanced techniques for structural health monitoring (SHM) for bridges with Traffic. We propose an approach using deep reinforcement learning (DRL)-based control for Unmanned Aerial Vehicle (UAV). Our approach conducts a concrete bridge deck survey while traffic is ongoing and detects cracks. The UAV performs the crack detection, and the location of cracks is initially unknown. We use two edge detection techniques. First, we use canny edge detection for crack detection. We also use a Convolutional Neural Network (CNN) for crack detection and compare it with canny edge detection. Transfer learning is applied using CNN with pre-trained weights obtained from a crack image dataset. This enables the model to adapt and improve its performance in identifying and localizing cracks. Proximal Policy Optimization (PPO) is applied for UAV control and bridge surveys. The experimentation across various scenarios is performed to evaluate the performance of the proposed methodology. Key metrics such as task completion time and reward convergence are observed to gauge the effectiveness of the approach. We observe that the Canny edge detector offers up to 40\% lower task completion time, while the CNN excels in up to 12\% better damage detection and 1.8 times better rewards.

Figures (9)

• Figure 1: System Model illustrating bridge survey and crack detection using UAV.
• Figure 2: Worflow for Canny Edge Detector technique
• Figure 3: Transfer Learning based CNN Workflow
• Figure 4: An illustration of the proposed approach with PPO agent and concrete bridge deck with traffic.
• Figure 5: PPO performance comparison for 5 cracks with canny edge detector versus CNN. CNN outperforms the Canny edge detector.