An Intelligent Lane-Changing Behavior Prediction and Decision-Making Strategy for an Autonomous Vehicle

In the future complex intelligent transportation environments, lane-changing behavior of surrounding vehicles is a significant factor affecting the driving safety. It is necessary to predict the lane-changing behaviors accurately. The driving environments and drivers are the main factors of lane-changing. To comprehensively consider their relationship, this article proposes a prediction method based on a fuzzy inference system (FIS) and a long short-term memory (LSTM) neural network. First, to highly integrate driving environments with drivers, drivers' cognitive processes of driving environments are simulated using FIS. Fuzzy rules are formulated based on drivers' cognition, and then driving environments information can be transformed into lane-changing feasibility. Second, the obtained lane-changing feasibility and corresponding vehicle trajectory are designed as input variables of LSTM neural network to predict the lane-changing behavior. Third, based on the above prediction results, an intelligent decision-making strategy is designed for path planning of autonomous vehicle to ensure driving safety. The prediction method is trained and tested by the next generation simulation (NGSIM) dataset, which is made up of real vehicle trajectories. The accurate rate of the method is 92.40%. Moreover, the decision strategy is simulated and verified in hardware-in-the-loop system. Results show that the strategy can significantly improve the performance of driving in dealing with lane-changing behaviors.

Automated summary

This article proposes a prediction method based on a fuzzy inference system and a long short-term memory neural network to accurately predict the lane-changing behavior of surrounding vehicles, as well as an intelligent decision-making strategy for path planning of autonomous vehicles to enhance driving safety.