Promoting excellence in mobility engineering

  1. FISITA Store
  2. Technical Papers

Vehicle Motion Prediction-based Adaptive Cruise Control
FISITA2016/F2016-VDCD-005

Authors

Lee, Dong-Han; Lee, Chan-Kyu; Hedrick, J.Karl - University of California at Berkeley, the United States of America

Abstract

KEYWORDS Adaptive cruise control, Model predictive control, Motion prediction, Interacting multiple model, Unscented kalman filter

ABSTRACT

Current adaptive cruise control (ACC) systems that have been commercialized in markets focus entirely on maintaining the space between the ego car and the preceding car or tracking a desired speed. However, there are still many limitations with these systems. One such limitation is when a vehicle moves into the ego car's lane from an adjacent lane and cannot be recognised by the system until it is fully inside of the ego car's lane. This condition makes driver feel uncomfortable and unsafe. This paper proposes a new vehicle motion prediction based adaptive cruise control which combines a vehicle motions predication based on Constant Speed and Constant Turn rate (CSCT) model and a vehicle motion prediction based on the Lane Change (LC) model. The combination of these two methods increase the prediction accuracy in both the short-term and the long-term, respectively, that deals with this specific issue. We use the Interacting Multiple Model with Unscented Kalman Filter (IMM-UKF) approach for the vehicle motion prediction based of CSCT model. Also, the LC model based on logistic function which mimics lane change motions is introduced. Using the combined prediction, The Model Predictive Control (MPC)-based upper-level controller gives optimal control inputs the horizon. The overall approach was evaluated by simulations with pre-recorded data for real vehicle lane change motions. The results demonstrated the superior performance in the combined prediction-based ACC.

Add to basket