A Smart Deep Learning Based Self Driving Product Delivery Car

Engineering notes

Key topics: self-driving car, self-driving, autonomous vehicle, tesla, control. See the paper for implementation details and experimental results.

Chinese explanation / 中文解读

中文解读待补充：本站会优先为端到端自动驾驶、BEV感知、3D目标检测、轨迹预测、路径规划、LiDAR感知等高价值论文补充中文说明。

Original abstract

Today, Tesla, Google, Uber, and GM are all trying to create their own self-driving cars that can run on real-world roads and Dominos Pizza, Amazon and Walmart decided to create them own proprietary delivery vehicles for commercial use. Intelligent logistics services using self-driving technology, in contrast with the traditional last-mile logistics services, provide a viable alternative for lowering delivery costs and improving quality. Many analysts predict that within the next 5 years, we will start to have fully autonomous cars running in our cities, and within 30 years, nearly ALL cars will be fully autonomous [1]. Wouldn't it be cool to build your very own self-driving car using some of the same techniques the big guys use? I have built my own prototype for a self-driving autonomous car from scratch that uses Deep Neural Network. I built enable the car to detect and follow lanes, recognize, and respond to signs also, it can detect post code box for commercial use. I also used Computer Vision and Deep Learning software needed to recognize the signs, post code box and road lanes. I used python as main tool and second tool is OpenCV powerful computer vision package. To construct a self-driving car, this project proposes an interesting methodology that combines Machine Learning, Image Processing, and IoT concepts. With the help of image processing, the input image is prepared. In order to process the photos, two major models were used. In the first model, I used image processing to recognize road lanes, causing the automobile to move to the right or left to stay within the same lane. The photos of the side of the road right after the lane line, on the other hand, are the region of interest in the second model. The picture of the region of interest is sent into the Convolutional Neural Network in this model, which processes the image. The output of the Neural Network assists in making targeted judgments. After that, the controller just sends the proper signal, and the autonomous vehicle's computer reacts accordingly.

Links and sources

• Official / arXiv page

Comments