Abstract

Autonomous systems in remote locations have a high degree of autonomy andthere is a need to explain what they are doing and why in order to increasetransparency and maintain trust. Here, we describe a natural language chatinterface that enables vehicle behaviour to be queried by the user. We obtainan interpretable model of autonomy through having an expert 'speak out-loud'and provide explanations during a mission. This approach is agnostic to thetype of autonomy model and as expert and operator are from the same user-group,we predict that these explanations will align well with the operator's mentalmodel, increase transparency and assist with operator training.

 

Quick Read (beta)