Real-Time Road Surface Unevenness Recognition for ITS Using BWO-BiGRU-Transformer With Adaptive CKF

The recognition of the road surface unevenness can provide an important support for optimizing the maintenance and management strategies of local intelligent transportation networks. This study proposes a BiGRU-Transformer model optimized with Beluga Whale Optimization (BWO) and combined with the Cubature Kalman Filter (CKF) for road surface unevenness recognition in intelligent transportation systems. The model addresses the limitations of existing methods in time series data processing and nonlinear state estimation. It uses BiGRU to model short-term vehicle vibration dependencies and incorporates the Transformer’s self-attention mechanism to capture long-term and global dynamics. BWO optimizes the global hyperparameters of the model, avoiding local optima and improving accuracy. CKF updates the noise matrix in real-time, enhancing prediction and generalization through recursive nonlinear state estimation. Experimental results demonstrate significant improvements in performance, including recognition accuracy, noise immunity, and convergence speed, when applied to real road datasets. The proposed model outperforms existing methods, offering enhanced road condition identification and providing effective real-time decision support for intelligent transportation systems, thereby improving management efficiency and responsiveness.