یک مدل جدید و خودکار برای تاکسی در شهرهای هوشمند
Abstract
I- Introduction
II- Motivation and Background Theory
III- Experimental Setup
IV- Results and Discussions
V- Conclusions
REFERENCES
Abstract:
Autonomous taxies are in high demand for smart city scenario. Such taxies have a well specified path to travel. Therefore, these vehicles only required two important parameters. One is detection parameter and other is control parameter. Further, detection parameters require turn detection and obstacle detection. The control parameters contain steering control and speed control. In this paper a novel autonomous taxi model has been proposed for smart city scenario. Deep learning has been used to model the human driver capabilities for the autonomous taxi. A hierarchical Deep Neural Network (DNN) architecture has been utilized to train various driving aspects. In first level, the proposed DNN architecture classifies the straight and turning of road. A parallel DNN is used to detect obstacle at level one. In second level, the DNN discriminates the turning i.e. left or right for steering and speed controls. Two multi layered DNNs have been used on Nvidia Tesla K 40 GPU based system with Core i-7 processor. The mean squared error (MSE) for the detection parameters viz. speed and steering angle were 0.018 and 0.0248 percent, respectively, with 15 milli seconds of real-time response delay.
Introduction
In the early 1920s the first attempt was made towards driverless vehicles [1] and got momentum in the 1980s when researchers managed to develop automated highway systems [2]. After this many work towards semi-autonomous and autonomous vehicles were made largely in Germany and U.S. The U.S. defense agency, Defense Advanced Research Projects Agency (DARPA), which is responsible for the development of emerging technologies organized driverless car competition many times which also encourage the research in the direction of autonomous vehicle navigation. Renowned car manufacturers like, Bavarian Motor Works (BMW), Ford, General Motors (GM), Google/ Waymo, and Tesla are also trying to build autonomous cars [3]. The main perceptual cues include road color and texture, lane markings, and road boundaries. So, semi-autonomous and autonomous vehicles are relying on the same perceptual cues as humans do. Till now road and lane perception via the traditional cues considered to be most effective for autonomous driving. We have to include the extent of the road the number and positions of lanes, merging, splitting and ending lanes and roads in urban, rural and highways scenarios for understanding of road and lane. Although many work related to this has been made in recent years still it is beyond the current perceptual systems. There are varieties of sensors used for road/ lane detection such as, camera, stereo, LIDAR, etc [4]. Vehicle mounted cameras based on autonomous driving method is one of the best method as per the current research methodology. Therefore, machine learning algorithms becomes helpful and efficient method to extract the appropriate features from the video frames and detect road/ lane and control as well as steer the vehicle autonomously. Deep Neural Network (DNN) algorithm is well capable of learning these complex features and improved the pattern recognition techniques tremendously. It is also emerged as much powerful artificial intelligence tool in machine learning [5]. DNN architecture is a multilayer architecture of simpler layers stacked upon one another such that it can learn complex nonlinear input – output mappings easily. Therefore, DNN can be utilized for navigating vehicles autonomously. In 2016, Bojarski et al. proposed CNN architecture for learning the frames captured from single front facing camera images such that it can map the steering angles of the steering wheel of the car [6].