Dynamic mass estimation framework for autonomous vehicle system via bidirectional gated recurrent unit

The precise estimation of vehicle mass is crucial for the optimal performance of electronic control systems in autonomous vehicles. However, the nonlinear nature of vehicle dynamics makes it a challenging task to estimate the mass accurately. In response to this concern, this paper proposes a dynamic vehicle mass estimation framework underpinned by a bidirectional gated recurrent unit, developed using deep neural networks. The bidirectional mechanism and gated recurrent unit network are adopted to elevate the precision of the neural network estimator. The dataset used for training and validation is collected from heavy‐duty vehicle simulations and real vehicle road tests. The average root mean square error, mean absolute percentage error, and mean absolute error evaluated over simulation tests are 92.66 kg, 0.93%$\%$, and 79.67 kg, respectively, and those in real vehicle data tests are 16.61 kg, 0.13%$\%$, and 16.61 kg, respectively. The outcomes manifest that the method put forth surpasses the contrasted approaches in relation to accuracy in the conducted experiments. This paper proposes a dynamic vehicle mass estimation framework based on a bidirectional gated recurrent unit, developed using deep neural networks. The bidirectional mechanism and gated recurrent unit network are adopted to increase the accuracy of the neural network estimator.