| In the automotive industry, the accuracy and processing of sensor data are crucial for ensuring vehicle safety. Among various vehicle dynamics parameters, longitudinal acceleration stands out as a crucial parameter, as its accurate prediction is essential for safety mechanisms such as the anti-lock braking system (ABS) and the activation of airbags. This article presents a Long Short-Term Memory (LSTM) neural network model designed to estimate the future longitudinal acceleration of a vehicle, with a focus on minimizing the hardware resources needed for computational operations. The development of the model was done using a large quantity of experimental data. The model's input consists of the past and current values of the speeds of the four wheels, lateral and longitudinal accelerations, the steering angle of the front axle, and the yaw rate. Our test results demonstrate that the developed model can forecast longitudinal acceleration 0.1 seconds ahead by utilizing the preceding 0.1 seconds of sensor data with high accuracy. Comparing our results with those published in the literature, we can conclude that the model we developed provides significantly more accurate predictions for longitudinal acceleration, achieved through the application of a less complex model. |
*** Title, author list and abstract as seen in the Camera-Ready version of the paper that was provided to Conference Committee. Small changes that may have occurred during processing by Springer may not appear in this window.