مقاله انگلیسی رایگان در مورد حرکت وسایل نقلیه الکتریکی با یادگیری عمیق – IEEE 2018

AFTER almost 100 years of domination by internal combustion engines, there has been a return of interest in electric vehicles (EVs). Many factors stimulate this resurgence. On the one hand, it is widely believed that EVs can solve energy and environmental problems to some extent. On the other hand, EVs with individually controlled motors, namely four-wheel independently actuated (FWIA) EVs, allow a significant improvement in motion performance due to their remarkable actuation flexibility [1], [2]. So far, various studies have been conducted on how to fully utilize the potential of the FWIA EVs to enhance the maneuverability and stability of vehicles [3]–[7]. It is also believed that the combination of electrically driven and autonomous driving is the future of vehicle technology [8], [9]. Sophisticated control systems in FWIA EVs or autonomous vehicles are based on the advancement of electrical and electronics technology. In the past, electronic devices in vehicles are connected by point-to-point wiring systems. As the growth of electronic control units (ECUs), continuing to adopt this conventional Electrical/Electronic architecture became impossible because it resulted in an expensive and messy wire system. Nowadays, the data signals between ECUs are transmitted through networks, i.e., controller area network (CAN) or FlexRay, forming a networked control system (NCS). However, as the number of ECUs rising further, which is typical in over-actuated FWIA EVs, the unknown and timevarying communication delays of networks may be large enough to degrade the performances of feedback control systems [10]. Some work has been carried out to deal with the delays induced by in-vehicle networks [11]–[14]. In these studies, two approaches are widely employed to describe the timevarying delays as uncertainties. One is a deterministic method, and the other is a stochastic method. In the deterministic method, delays are assumed as being taken from a fixed probabilistic distribution [11], [12]. Such description is quite simple and keeps the synthesis of the NCS easy. However, it neglects the correlation between current and previous delays, which may increase the conservativeness of the control system. In the stochastic method, the NCS is modeled as a Markov chain [13], [14]. Such random process model takes the dependence between adjacent delays into account and is more precise allowing for the real network phenomena. Once the models of NCS has been established, appropriate robust control methods can be designed to make the performance of the system as high as possible under the uncertainties induced by network delays