Research on a Lane-Changing Model With Advance Decision-Making Function Based on the Type-2 TSK Fuzzy Inference System

With the increase in vehicle population, the issue of traffic congestion is becoming increasingly severe. In addition to meeting conventional driving requirements, the proactive enhancement of traffic flow by intelligent vehicles holds significant research importance. Therefore, we propose a lane-changing decision-making model with an advance lane-changing function based on Bi-directional Long Short-Term Memory (BiLSTM) and Type-2 Takagi-Sugeno-Kang Fuzzy Inference System (T2 TSK FIS). Our model features a two-level structure. The first level is a fuzzy inference system responsible for primary decision-making. The second level comprises a BiLSTM advance lane-changing model and another fuzzy inference system, facilitating advance lane-changing decisions. In order to enhance model reliability, the most critical features are selected for model construction by analyzing real vehicle driving data of the Next Generation Simulation (NGSIM) dataset, ultimately achieving a decision accuracy rate of 90%. The validation of the advance lane-changing decision also utilizes the NGSIM dataset. Results indicate that our model facilitates better lane-changing conditions for vehicles, mitigates traffic oscillations, and reduces traffic congestion.