3 October 2019 - Esra Erdem(Sabanci University)

Abstract

Hybrid reasoning for robotics necessitates combining discrete high-level reasoning with continuous feasibility checks and perception. We propose to address such hybrid reasoning problems with a causality-based approach using Answer Set Programming (ASP). The idea is (i) to represent the robotic actions and change in a high-level language of ASP, which allows embedding continuous feasibility checks into logical formulas by means of formal interfaces, and (ii) to compute feasible solutions to hybrid reasoning problems using state-of-the-art ASP solvers and relevant feasibility checkers within a modular architecture. We discuss this hybrid approach in the context of two main reasoning tasks: planning of robotic actions for a feasible execution, and explanation generation of failures via diagnostic reasoning. We illustrate its applications in different robotic domains.