**Autonomous Decision Making for Driverless Cars**
Nicholas Gamez, Nicolas Gallardo
*Autonomous Controls Lab, University of Texas at San Antonio, San Antonio, Texas, USA*
{jyi358, hbq774}@my.utsa.edu
**Project Definition:** Autonomous Driving has been a hot topic with companies like Google, Uber, and Tesla and they have had some success in applying algorithms to commercial cars. Using Deep Learning techniques and the Tensorflow framework, the goal is to navigate a driverless car through an urban environment. The novely in this system is the use of Deep Learning vs. traditional methods of real-time autonomous operation.
The dataset chosen is comprised of a simulation enviroment emulating real life driving situations for training[1]. Feature points such as direction of the steering wheel, location of nearby cars, and specific road markings will be used to determine the best course of action to safely navigate to a desired endpoint. After training, the goal is to apply this alrogithm to an autonomous vehicle instead of more simulation.
The rest of the data will be used to learn different components of driving i.e, car, road sign, pedestrian, cyclist identification, etc. Also as a general supplement to the primary dataset mentioned above.
[1]: C. Chen, A. Seff, A. Kornhauser and J. Xiao, DeepDriving: Learning Affordance for Direct Perception in Autonomous Driving, Proceedings of 15th IEEE International Conference on Computer Vision (ICCV2015) .
**Outcome:** Apply deep neural network to determine the best course of action for a car given an endpoint.
**Dataset:** The simulation training data can be found in [http://deepdriving.cs.princeton.edu/TORCS_trainset.zip][1]. This directory contains over 50 GB of simluation data of a car with the direction of the steering wheel and surrounding caras and road markings.