Overview

In order to autonomously accomplish complex, long-horizon tasks such as doing the laundry, robots need to compute long-horizon, high-level task plans representing the overall strategy, as well as short-horizon, low-level motion plans representing the physical movements required for executing the strategy. Integrating task planning with motion planning has been well established as a challenging sequential decision-making problem: it constitutes one of the major bottlenecks in enabling a variety of applications, including mobile manipulation, household robotics, healthcare robotics, robots for disaster recovery, etc.

This full-day tutorial will present an introduction to the rich and active field of integrated task and motion planning. Starting with a survey of technical problems and diverse state-of-the-art approaches towards addressing them, it will present recent advances in data-driven learning for integrated task and motion planning as well as major open problems in the field and promising directions of research for addressing them.