Contrastive Natural Language Explanations for Multi-Objective Path Planning

Raj Korpan, Susan Epstein

Keywords: contrastive explanations, multi-objective path planning, natural language, robot navigation

TL;DR: This paper introduces a flexible, scalable approach that generates contrastive explanations of navigation plans based on multiple objectives.

Abstract: This paper introduces a flexible, scalable approach that generates contrastive explanations of navigation plans based on multiple objectives. These explanations in natural language describe a robot controller's beliefs, intentions, and confidence to any person who travels with or near the robot. A new multi-objective path planning algorithm generates optimal single-objective plans, evaluates each of them with respect to the other objectives, and selects one. The objectives that favored the selected plan over the others become reasons in the explanation. Extensive evaluation in simulation demonstrates the system's ability to produce diverse, readily understandable explanations that provide counterfactual examples.