M2M Model based on Sampling Method

Research Objectives

• uReplacing rule-based collision judgement with a ML-based M2M model

• uEnhancing trajectory feasibility and realism through Trajectory Vocabulary

• uImproving model interpretability by applying multi-cost evaluation

  
  

Research Contents

• uGenerate Trajectory Vocabulary based on open datasets or custom datasets

• uSelect optimal trajectories based on multi-cost evaluation (safety, progress, comfort)

• uEvaluating model performance through open-loop and closed-loop testing based on the nuPlan simulator