Abstract—Based on a non-linear single track model which describes the dynamics of vehicle, an optimal path planning strategy is developed. Real time optimization is used to generate reference control values to allow leading the vehicle alongside a calculated lane which is optimal for different objectives such as energy consumption, run time, safety or comfort characteristics. Strict mathematic formulation of the autonomous driving allows taking decision on undefined situation such as lane change or obstacle avoidance. Based on position of the vehicle, lane situation and obstacle position, the optimization problem is reformulated in real-time to avoid the obstacle and any car crash. Keywords—Autonomous driving, Obstacle avoidance, Optimal control, Path planning. I. INTRODUCTION UMAN error is the main reason of accident which cause 93.5% of accident based on Audi accident research. Drowsiness and distraction as the driver states are relevant cause of traffic crashes, 15% and 18% of accidents reason each respectively. NHTSA 2008 shows that driver is involved in at least one non-driving activity in 18% of cases. Pedestrian and cyclist as an important part of traffic system need more focus because one pedestrian life lost every two hours in United States and high number of fatalities and injuries for cyclist [1]. These are the main reasons that modern vehicle are equipped with an increasing number of sensors to sense surroundings such as radar, lidar, computer vision sensors and GPS and also are equipped with driving assistant systems such as navigation systems, intelligent speed adaptation, electronic stability control and etc. A further development of these systems is autonomous driving which means driving without or very limited intervention of driver. Classical approaches for path planning [2] are mainly rules and maneuvers based. However predefining all the critical situation is not feasible, this is whyin this research work the autonomous driving is based on mathematic formulation which allows taking decision on unforeseen situations and restrictions related to the road in the case of obstacles, sudden change on the road or other vehicles, which maybe not all covered by predefined rules. Based on the dynamic vehicle model, the geometric data of the track as the known input is delivered to the path optimization level by e.g. a navigation R. Dariani is research assistant at the University of Magdeburg, Germany. (phone: +493916752084; fax: +493916712656; e-mail: reza.dariani@ovgu.de) S. Schmidt is assistant Professor “Autonomous Vehicles” at the University of Magdeburg, Germany (phone: +493916752084; fax:+493916712656; e- mail: stephan.schmidt@ovgu.de). R. Kasper is professor and head of chair of mechatronics at University of Magdeburg, Germany (phone: +493916752606; fax: +493916712656; e-mail: roland.kasper@ovgu.de). system or digital map. Then, based on delivered data in the path optimization level, steering angle and driving force as the inputs of the systems are calculated. These allow leading the vehicle alongside a calculated lane which is optimal for different objectives as energy consumption or comfort and allows taking into account some constraints as width of the lane or maximal lateral or longitudinal acceleration. There is no guarantee that vehicle behavior matches perfectly with the optimal solution found by optimization algorithm due to the model uncertainty or environment disturbances such as different road friction and side wind. This is why a closed loop path control system is provided to generate additional inputs, force and steering angle, in order to correct the longitudinal and lateral distance error. Fig. 1 shows the general hierarchical concept of autonomous driving used in this research work. Fig. 1 General hierarchical concept of autonomous driving The optimal problem due to the big dimension of the system and length of course is numerically hard to solve and require more calculation time to find optimal solution. In another hand optimal problem must be updated frequently to consider the dynamic behavior of road and environment. To solve the problem in an easier way and make the system real-time capable, a “moving horizon approach” is used, in which global optimization problem is partitioned into a sequence of local optimization problems with an adequate smaller horizon. Updating and solving the problem with small horizon offers the possibility to update lane status which is useful in the case of sudden changes, obstacle or other vehicles in the lane. In the case of another vehicle on the road as a moving obstacle or in the case of fix obstacle, the optimization problem is updated and reformulated by considering the obstacle in the objective function in order to avoid it. Based on the obstacle position and its distance to our vehicle a penalty function is added to optimization problem. Optimization Based Obstacle Avoidance R. Dariani, S. Schmidt, R. Kasper H World Academy of Science, Engineering and Technology International Journal of Computer and Information Engineering Vol:8, No:9, 2014 1567 International Scholarly and Scientific Research & Innovation 8(9) 2014 ISNI:0000000091950263 Open Science Index, Computer and Information Engineering Vol:8, No:9, 2014 publications.waset.org/9999297/pdf