Driving intention recognition based on COA-LSTM network

Abstract

Driving intention recognition is one of the key technologies in the development of advanced assisted driving systems (ADAS), which can greatly reduce the occurrence of traffic accidents and thus improve driving safety and comfort. In order to recognize driving intention more effectively and accurately, this paper proposes a method to improve the recognition rate of driving intention based on Long Short-term Memory (LSTM) network optimized by Coati Optimization Algorithm (COA). First, the training dataset and test set based on the road information dataset are established by considering the degree of influence of feature factors on driving intention. Second, the COA-LSTM driving intention recognition model is constructed. Finally, the training and validation evaluation metrics are performed based on the dataset, and the results show that the COA-LSTM-based driving intention recognition model proposed in this paper is better than the LSTM and BP models as a whole, and the accuracy rate and the 𝐹 score are both above 0.95.