Development of A Concept for A Driverless Vehicle Using an Artificial Neural Network

Recent developments in self-driving cars and smart control would make the idea of an autonomous vehicle a viable one. Nonetheless, using of such cars in traffic and some other hostile surroundings successful requires effective solution of a package of navigation, vision, and control problems. In all this, our ultimate goal is to come up with cost effective methods that are operationally sound which in turn can help the larger research community to take self-driving automobiles offering business potential seriously. However, what we are in need of, is a ploy that could transform the traditional motorists into discouraging using self-driving cars and at the same time works out the sorting of the present passenger cars into the future research work. In this vein, the current discussion develops a modular mechanical structure that could be manufactured in especially fast fashion and adapted on a significant portion of modern automobiles. The design brief is an intermediate stage of development on the road to the production of wholly autonomous cars. With the use of commercially available actuators, we prove that it is possible to transform homegrown automobiles of the sort found in abundance in the parking lot of whatever rest stop or convenience store you happen to drop by, quite reasonably to include most of the privately owned vehicles, into autonomous models. In the context of the motor vehicle's automation, motors are often used as actuators Considering the fact that the electro-mechanical motors did not work, pneumatic system was introduced to make one predetermined step automated. Mechanical setup of autonomous cars is the most important aspect and should provide a possibility to make regular updates as the car is designed to work under extremely changing conditions. In order to measure the feasibility of such strategy, two further convolutional neural networks were re-implemented, allowing to conduct an objective comparison of the performance, technical severity and architecture of the proposed system with regards to the reference networks. The architecture is proposed to 300 AlexNet units and three PilotNet units (Increased feature extraction capacity, slightly reduced control model replication ) and operate at a bigger scale than the prevalent models of neural networks that are available today. On the one hand, it has fewer architectural complexities compared to its peers and hence experiences longer latencies, and slower inference time; on the other hand, the system still ranks equally well on the autonomous-driving metrics reported by two similar benchmarks. By meeting this performance objective and at the cost of no longer requesting high-speed computing hardware, the model promotes not only cost-efficiency, scale, and total affordability.