Sheila (self driving car)

Author: Zaal Dzindzibadze
Co-authors: Mate Bersenadze, Mikheil Lomidze, Aleksandre Sarchimelidze
Keywords: Deep Learning, Self driving car, Machine learning, Autonomous system
Annotation:

We’d like to present a microservice, that can make the car move in unknow and new environment without human intervention. It should avoid all blockages and if necessary stop, slow or increase the movement speed, without causing any damage. We solve the problem using couple of models of Neural Networks, Deep Learning, that help us determine which way the car should move in a specific time and location. The Nueral Networks make these decisions from analysing the information provided by sensors and the camera on the car. The calculations are made locally and on server, connecting through internet, but in case by any reason the connection is down, the decisions are made by the minimal Neural Networks on the car, that use Raspberry pi 3 / intel modulus 2. We have to admit, that due to specifications of Deep Learning, the system needs huge amount of data for starting to function, the data is accumulated through with human help, by controlling the car for couple of hour in the beginning.