#selfdriving, #carla, #qlearning, #pytorch
In this lecture we enter the realm of autonomous driving by making use of the Carla simulator and implementing deep q-learning networks in pytorch. The task is to control the steering of a self driving agent. This lecture is only an introduction to doing self driving in this manner (for end to end autonomous systems), and as we learn more reinforcement learning & deep learning algorithms, we'll improve our agent's performance. Stay tuned.

You can find the code referenced in this lecture in the following repository: [ Ссылка ]

The agent was trained for about 7-8 hours, just to show its capability to learn in this environment.

Reference: This paper inspired the approach taken
[ Ссылка ]

0:00 Intro
1:50 Problem Setup
6:20 Reward Function Design
16:45 Image Processing
20:45 Architecture
29:40 Challenges in RL Design
34:20 Programming Solution

свернуть