Vehicle Dynamics Modeling Using FAD Learning
Highly precise vehicle dynamics modeling is indispensable for self-driving technology. We propose a model learning framework, which utilizes FAD (The abbreviation of the capital letters of free dynamics, actuator, and disturbance.) learning, motor babblin
- PDF / 3,169,371 Bytes
- 14 Pages / 439.37 x 666.142 pts Page_size
- 67 Downloads / 191 Views
2
Department of Electrical Engineering and Computer Science, Iwate University, 203 Room, 4 Bldg. East, 4-3-5, Ueda, Morioka-shi, Iwate, Japan [email protected] Department of Mechanical Engineering, Shizuoka University, Shizuoka, Japan [email protected]
Abstract. Highly precise vehicle dynamics modeling is indispensable for self-driving technology. We propose a model learning framework, which utilizes FAD (The abbreviation of the capital letters of free dynamics, actuator, and disturbance.) learning, motor babbling, and dynamics learning tree. In the proposed framework, modeling error was decreased compared with conventional neural network approach. Also, this framework is applicable to online learning. In experiments, FAD learning and dynamics learning tree decreased learning error. The dynamics of a simulated car was learned using motor babbling. The proposed framework is applicable to a variety of mechanical systems. Keywords: Learning
1
· Dynamics · Automobile
Introduction
Self-driving technology is getting attention, because it might suppress many types of accidents caused by human error (e.g. inadequate distance between vehicles, traffic jam caused by insufficient driving skill, increase of environmental load caused by unnecessary acceleration and deceleration, loose driving and inattentive driving). In order to realize safe self-driving vehicle, precise modeling is important. In previous research, Lee et al. proposed a path planning algorithm with a novel path representation for self-driving vehicle [1]. The algorithm was successfully implemented to KAIST Self-Driving Car. Li et al. proposed short term trajectory generation algorithm [2]. Kim et al. proposed parallel scheduling method for the physical attributes of a vehicle [3]. Kim et al. proposed linear prediction based uniform state sampling (LPUSS) that effectively find almost the optimal trajectory based on a vehicle model [4–7]. In order to use these trajectory planning algorithms, precise model of a vehicle is required. However, there is much error between the trajectory calculated by the used model and that of real machine in many cases. In recent years, Bang developed a vehicle dynamics model by focusing on spring mass forces and moments [8]. Satar proposed a five degree of freedom (DOF) longitudinal model [9]. Setiawan proposed 14 DOF vehicle model [10]. c Springer International Publishing Switzerland 2016 H. Fujita et al. (Eds.): IEA/AIE 2016, LNAI 9799, pp. 768–781, 2016. DOI: 10.1007/978-3-319-42007-3 65
Vehicle Dynamics Modeling Using FAD Learning
769
However, these mathematical models have also a gap from a real vehicle. Aerodynamic affection, friction around wheels, specification of engine, and so on, are difficult to be determined precisely using mathematical equations. Using real data for the modeling is a key technology to overcome the difficulty. Kabiraj et al. applied nonlinear tire data to their model [11]. Yim et al. developed a learning system on the basis of neural network, fuzzy logic, and evolutionally algorithm [12
Data Loading...
