Hiroyuki Furusho

Automatic path tracking using linear quadratic control theory

The authors conducted a study of automatic path tracking performance in the automated highway system (AHS) public trials that were carried out in Japan in 1996. A comparison was made of proportional-plus-derivative (PD) control and linear quadratic (LQ) control as two potential control systems. Because the PD control system lacked state feedback of yaw motion, the interference of yaw motion on lateral motion could not be eliminated. This made it impossible to achieve both satisfactory response and convergence. The LQ control system, on the other hand, had a larger number of state feedback coefficients, enabling lateral motion to be separated from yaw motion. This made it possible to set the response and convergence of lateral motion at will. Driving on an automated highway system is premised on the receipt of feedforward information from the road-side infrastructure. In this study, it was shown both theoretically and experimentally that road curvature can be estimated logically as a vehicle travels along even in the absence of feedforward information on curvature. Logical estimates are obtained with an algorithm in which curvature is treated as a state variable. It was also shown experimentally that an autonomous lane-tracking vehicle can even navigate road curves of unknown curvature by using a regulator configured around the method of estimating road curvature logically.

Investigation of automatic path tracking using an extended Kalman filter

Abstract This paper proposes a lane marker recognition method that uses the steering angle in addition to image information. A Kalman filter was reconfigured regarding the yaw motion and lateral motion of lane markers, previously treated as a stochastic process, as the states of a vehicle model driven on the basis of the steering angle. Driving tests conducted with an actual vehicle verified that this method provides good tracking at the time of steering input and avoids misrecognition of lane marker candidate points in inclement weather.

Research on automated lane tracking using linear quadratic control : control procedure for a curved path

Research has been done recently on automated steering systems for automobiles which would make hands-free driving possible. In the present research, it was shown both experimentally and by simulations that the use of linear quadratic (LQ) control improves lane tracking control performance. It was observed that LQ control improves tracking performance by working to separate yaw motion from lateral motion. It was found that both satisfactory steady-state characteristics and response could be obtained by suitably weighting lateral motion. The first part of this paper shows how the use of LQ control facilitates lane tracking along a straightway. The second half describes a method for enabling an automated vehicle to navigate curves. It is shown that detection of the vehicles lateral deviation alone, without any feed-forward information about the road curvature or other conditions, is sufficient to secure stable cornering performance. For the covering abstract, see IRRD 490001.