Affiliations: [a] Electronics and Communication Engineering, Sengunthar Engineering College, Tiruchengode, Tamilnadu, India
| [b] Computer Science and Engineering, Dr. NGP Institute of Technology, Coimbatore, Tamilnadu, India
| [c] Electronics and Communication Engineering, Builders Engineering College, Kangeyam, Tamilnadu, India
Correspondence:
[*]
Corresponding author. Arunkumar Muthumanickam, Electronics and Communication Engineering, Sengunthar Engineering College, Tiruchengode, Tamilnadu, India. E-mail: arunkumarmece01@gmail.com.
Abstract: The field of self-driving cars is one that is rapidly growing in popularity. The goal of autonomous vehicles has always been to avoid accidents. It has long been argued that human errors while driving are the primary cause of traffic accidents, and autonomous cars have the potential to remove this. An intelligent transportation system based on the Internet of Things (IoT) is required at some point for the vehicle to make an instant choice to evade accidents, regardless of the competence of a decent driver Mishaps on the road and in the weather are those that occur due to unfavourable weather circumstances such as fog, gusts, snow, rain, slick pavement, sleet, etc. There are many factors that might cause a vehicle to lose control, including speed, weight, momentum, poor fleet maintenance. It has the potential to lessen the number of collisions caused by poor weather and deteriorating road circumstances. An IoT-based intelligent accident escaping system for poor weather and traffic circumstances is presented here. A variety of sensors are used to check the health of the vehicle. Data from sensors is processed by a microcontroller and displayed on the dashboard of a car after it has been received. The proposed model combines both an IoT system that monitors weather and road conditions and an intelligent system based on deep learning that learns the adverse variables that impact an accident in order to anticipate and prescribe a harmless speed to the driver. The experimental results show that the proposed deep learning technique achieved 94% of accuracy, where the existing LeNet model achieved 80% of accuracy for the prediction process. The proposed ResNet is more effective than LeNet, because identity mapping is used to solve the vanishing gradient problems.
Keywords: Accidents-free driving, autonomous vehicles, deep learning, fleet management, internet of things, microcontroller, sensors
