Building Environments for Autonomous Vehicle Training

Explore the world of autonomous vehicle training in this 46-minute conference talk from Unite Berlin 2018. Discover how CVEDIA leverages Unity to create dynamic environments for simulation requirements, pushing the boundaries of modern simulation and modeling. Delve into the challenges faced, innovative techniques developed, and cutting-edge sensor simulations. Learn about automotive areas, autonomous systems, and the data problem in autonomous vehicles. Understand SAE levels, simulation systems, and workflows. Gain insights into Unity's role in AV development, including Asset Store resources, strengthening AV support, and sample scenes with sensors and shaders. Explore the Unity ML Agents workflow and imitation learning. Examine the Vision Zero concept, autonomy, and ADAS applications in highway scenarios and edge cases. Investigate pedestrian interactions, object entropy, environment entropy, and thermal entropy. Uncover the complexities of dataset generation and realize that this is just the tip of the iceberg in autonomous vehicle training.

Intro
Automotive Areas
Autonomous Systems
The autonomous vehicles data problem
SAE levels
Simulation systems
Simulation workflow
How can Unity help AV development
Asset Store
Strengthening AV support
Sample scenes, sensors, shaders
Unity ML Agents Workflow
Unity ML Agents: Imitation Learning
Strong ecosystem
Vision Zero
Autonomy & ADAS
Highway
Edge cases
Case study
Pedestrians
Object Entropy
Environment entropy
Thermal entropy
Dataset generation
Tip of the iceberg