|
Activity Number:
|
119
- Statistical Learning Applications for Autonomous Systems in Defense and National Security
|
|
Type:
|
Topic Contributed
|
|
Date/Time:
|
Monday, August 3, 2020 : 1:00 PM to 2:50 PM
|
|
Sponsor:
|
Section on Statistics in Defense and National Security
|
|
Abstract #313588
|
|
|
Title:
|
Leveraging Machine Learning for Autonomy Testing and Evaluation
|
|
Author(s):
|
Galen Mullins*
|
|
Companies:
|
Johns Hopkins University Applied Physics Laboratory
|
|
Keywords:
|
Autonomous Vehicles;
Adversarial Testing;
Test Case Generation;
Machine Learning;
Testing and Evaluation;
Performance Benchmarks
|
|
Abstract:
|
Traditional testing methods for unmanned systems involve field testing a handful of vignettes designed by subject matter experts. However, autonomous systems often fail in unexpected ways due to the interactions between multiple sub-systems and these vignettes are rarely sufficient to fully exercise the system. There are also safety considerations, as testing capabilities such as collision avoidance naturally places the vehicle in dangerous situations. New advances in using machine learning allow us to discover adversarial test cases using orders of magnitude fewer simulations than prior Monte-Carlo techniques. This talk will discuss frameworks that apply these advances to solve the problem of testing autonomous vehicles. As well as a potential path forward to make these processes more efficient and easier to deploy
|
Authors who are presenting talks have a * after their name.
