Communicating Safety of Planned Paths via Optimally-Simple Explanations

Noel Brindise, Cedric Langbort

Research output: Chapter in Book/Report/Conference proceedingConference contribution


Artificial intelligence is often used in path-planning contexts. Towards improved methods of explainable AI for planned paths, we seek optimally simple explanations to guarantee path safety for a planned route over roads. We present a two-dimensional discrete domain, analogous to a road map, which contains a set of obstacles to be avoided. Given a safe path and constraints on the obstacle locations, we propose a family of specially-defined constraint sets, named explanatory hulls, into which all obstacles may be grouped. We then show that an optimal grouping of the obstacles into such hulls will achieve the absolute minimum number of constraints necessary to guarantee no obstacle-path intersection. From an approximation of this minimal set, we generate a natural-language explanation which communicates path safety in a minimum number of explanatory statements.

Original languageEnglish (US)
Title of host publicationKI 2022
Subtitle of host publicationAdvances in Artificial Intelligence - 45th German Conference on AI, Proceedings
EditorsRalph Bergmann, Lukas Malburg, Stephanie C. Rodermund, Ingo J. Timm
Number of pages14
ISBN (Print)9783031157905
StatePublished - 2022
Event45th German Conference on Artificial Intelligence, KI 2022 - Trier, Germany
Duration: Sep 19 2022Sep 23 2022

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume13404 LNAI
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349


Conference45th German Conference on Artificial Intelligence, KI 2022


  • Constraint optimization
  • Explainable AI
  • Human-robot interaction
  • Mental model reconciliation
  • Path planning

ASJC Scopus subject areas

  • Theoretical Computer Science
  • General Computer Science


Dive into the research topics of 'Communicating Safety of Planned Paths via Optimally-Simple Explanations'. Together they form a unique fingerprint.

Cite this