Improvement of Lane Keeping Assistance ADAS Function utilizing a Kalman Filter Prediction of Delayed Position States Selim Solmaz * VIRTUAL VEHICLE Research Center Inffeldgasse 21a, 8010, Graz, Austria selim.solmaz@v2c2.at Georg Nestlinger VIRTUAL VEHICLE Research Center Inffeldgasse 21a, 8010, Graz, Austria georg.nestlinger@v2c2.at Georg Stettinger VIRTUAL VEHICLE Research Center Inffeldgasse 21a, 8010, Graz, Austria georg.stettinger@v2c2.at Abstract—In designing and implementing control systems, converting simulation based results to real life systems is often not straightforward and may need adaptation of the control approach to achieve similar performance levels to the simulation results. Such adaptations are usually required due to the fact that sensors and actuators have a number of imperfections such as delays, offsets and inherent noise processes. Here, such a problem in relation to the development of a lane keeping control algorithm is presented. An in-house developed lane keeping controller based on a high-fidelity simulation environment was planned to be transferred to a real demonstrator test vehicle. First tests showed significantly deteriorated and unstable performance results of the corresponding controller, which was due to sensor delays and actuator imperfections. After the diagnosis of the problem, an approach to mitigate these issues was undertaken by predicting the delayed sensor data utilizing a linear Kalman filter and an a-priori predictor. The Kalman filter and a-priori predictor design approach is based on a discrete-time version of the lane tracking model. The approach and the corresponding results were demonstrated using simulation and real vehicle implementation results that were evaluated in real driving conditions. Index Terms—lane keeping control, lane tracking model, Kalman filter, sensor delay I. I NTRODUCTION AND BACKGROUND Motorway Chauffeur (MWC) developed in the scope of a former project is an autonomous driver assistance technology that is designed to be SAE Level 3/3+ automated vehicle controller [1]. The MWC combines of a number separate control sub-systems or advanced driver assistance technologies (ADAS functions) that include a longitudinal guidance system, lateral guidance system and a trajectory planning system. Speed and target distance regulation in the form of adaptive cruise control (ACC), lane following maneuvers in the form of lane keeping assistant (LKA), and lane change planning in the from of trajectory planner (TP) is covered by the MWC. The general structure and architecture of the MWC is shown in Figure 1. The LKA controller in the MWC is a flatness-based bump- less controller [2], [3]. The implementation of the MWC controller is based on Matlab/Simulink, where the ADAS controllers run on Matlab/Simulink and the simulated vehicle *corresponding author. Tel: +43 316 873 9730, Fax: +43 316 873 9602. Fig. 1. Motorway Chauffeur (MWC) Architecture developed during a former project. dynamics signals are obtained from a commercial vehicle dy- namics simulator known as CarMaker [4]. While the described simulation approach is quite useful for developing and testing various ADAS functions before implementing or deploying in a real vehicle, there are also many limitations with this approach on modeling realistic sensor signals with regards to uncertainties, biases and limited or varying data rates. While it is also possible to include these imperfections in the vehicle and sensor models, the obtained model will only be valid to a specific vehicle type with a defined architecture, where this model needs to be revised and re-parameterized for every different vehicle. A logical approach for developing ADAS systems for such an uncertain system is to choose a control method that can cope with such uncertainty by its design, either in the form of a robust or an adaptive feedback control strategy. However, there are no general control approaches that can provide a solution for any control problem, especially for non-linear systems. So it is a common practice in control theory to choose a linear control design technique, where general stability analysis and control design techniques are well known. However, in this case one often needs to cope with effects of non-linearity in the real-life system as the linear control design techniques for such