C2205: Exploration of Machine Learning and Neural Networks for ADAS and L4 Vehicle Perception

Abstract

Content: Convolutional neural networks are the de facto method of processing camera, radar, and lidar data for use in perception in ADAS and L4 vehicles, yet their operation is a black box to many engineers. Unlike traditional rules-based approaches to coding intelligent systems, networks are trained and the internal structure created during the training process is too complex to be understood by humans, yet in operation networks are able to classify objects of interest at error rates better than rates achieved by humans viewing the same input data.

In this course participants examine an existing convolutional network to understand how it is constructed, trained and then using open-source python tools you’ll create your own convolutional neural network. You’ll train your network using a supplied training data set and then test your networks performance.

Learning Objectives

Content

By attending this course, you will be able to:

Describe the differences between a convolutional and other types of neural networks and where each should be used in the context of ADAS and L4 vehicle development
Explain how a convolutional network performs the perception and classification task on raw camera radar or lidar images to identify objects of interest impacting the ADAS or L4 vehicle operation
Select test cases for evaluating the performance of neural network-based perception systems
Identify potential problems in the network performance caused by training problems due to bias or other problems with the training data sets

Who Should Attend

Content: This course is designed for ADAS and L4 development engineers working on perception using camera, lidar and radar sensors, and validation and test engineers responsible for functional safety (ISO 26262) and SOTIF (ISO 21448) where the perception systems are using AI, neural networks, or other machine learning techniques.

Policy makers responsible for regulations regarding public highway testing of autonomous vehicles where perception is based on AI, neural networks or other machine learning techniques will also benefit from this course.

Prerequisites

Content: A bachelor's degree in a technical discipline and familiarity with basic ADAS functions is recommended.

Familiarity with the python programming language (or other scripting language such as visual basic or Matlab) is recommended.

Meta Tags

Topics: Advanced driver assistance systems (ADAS)
Neural networks
Flight recorders
Machine learning
Event data recorders
Education and training
Railway vehicles and equipment
Data exchange
Lidar
Driver assistance systems

Details

Additional Details