Model-Based Hybrid Control of Pure Pursuit and Stanley Methods for Vehicle Path Tracking

Abstract

In this study, a new method was applied to systematically combine the two controllers, which can help overcome the limitations of non-systematic combinations such as rule-based methods. For the model-based process, the bicycle model was used. Then, the model probability was calculated through the interactive multiple model filtering algorithm, which stochastically determines the most appropriate model that fits the current dynamic situation of the vehicle well. Based on this result, a hybrid path tracking controller was developed using the model probability of each method. The superiority of the proposed method was validated using the MORAI Drive simulator, which reflects the real road environment well enough. The results showed that the RMS tracking performance error was reduced by 6.0–8.8% in quarter-circle path and 3.3% in general path compared to single methods.