Tutorial on AI Planning for Robotics

Abstract

The main goal of this tutorial is to show how state-of-the-art formalisms in planning and scheduling (P&S) are used to model robotics domains; some of the challenges and solutions involved in dispatching plans on-board robots and autonomous systems; and to provide an overview of the ROSPlan framework. AI planning for long-term autonomy means an autonomous agent planning for unsupervised periods of days or months. Many interesting robotics problems are problems for long-term autonomy. For example, in service environments, or seabed inspection and maintenance. Planning for these scenarios requires rich models to capture the uncertain and evolving environment, and robust methods of execution. The first part of the tutorial will focus on modelling and solving P&S problems for robotics systems, including a description of how PDDL+ can be used to handle complex dynamics, and some examples of P&S solutions taken from various domains. The second part of the tutorial will cover the challenges that arise in plan execution, such as: incomplete knowledge, handling temporal constraints, state estimation, error detection and recovery, and exploiting opportunities. We will describe the ROSPlan framework, and show some solutions to these challenges. The final part aims to summarize the main challenges, open issues, and new opportunities related to planning for autonomous robots. In particular, we will report on the recent news from the Dagstuhl Workshop on Planning and Robotics.

URL: http://kcl-planning.github.io/ROSPlan/tutorials/tutorialICRA2017