Reinforcement Learning for Low Probability High Impact Risks

We demonstrate a method of reinforcement learning that uses training in simulation. Our system generates an estimate of the potential reward and danger of each action as well as a measure of the uncertainty present in both. The system generates this by seeking out not only rewarding actions but also...

Full description

Bibliographic Details
Main Author: Hunt, Gareth David
Format: Thesis
Published: Curtin University 2019
Online Access:http://hdl.handle.net/20.500.11937/77106
Description
Summary:We demonstrate a method of reinforcement learning that uses training in simulation. Our system generates an estimate of the potential reward and danger of each action as well as a measure of the uncertainty present in both. The system generates this by seeking out not only rewarding actions but also dangerous ones in the simulated training. During runtime our system is able to use this knowledge to avoid risks while accomplishing its tasks.