Domain Optimization for Hierarchical Planning based on Set-Theory

The design of planning domains for autonomous systems is a hard task, especially when different parties are involved. We present a domain optimization algorithm for hierarchical planners that uses a set-based formulation. Due to an automatic alignment we can compose models from different sources to a larger domain for efficient planning. The combination of domain optimization and hierarchical planning can handle large scale domains very efficiently. Our algorithm reduces the effects of the non-optimality that comes with the hierarchical approach. We demonstrate the scalability with a task and motion planning problem. In the scenario of a robotic assembly with up to 62 parts and plan lengths of over 1000 steps the planning times are kept within 15 minutes. We show the execution of our plans on a real-world dual-robot setup.

[1]  Jianwei Zhang,et al.  A Hierarchical Planner based on Set-theoretic Models: Towards Automating the Automation for Autonomous Systems , 2019, ICINCO.

[2]  Patrik Haslum,et al.  Domain-Independent Construction of Pattern Database Heuristics for Cost-Optimal Planning , 2007, AAAI.

[3]  Demis Hassabis,et al.  Mastering the game of Go without human knowledge , 2017, Nature.

[4]  Shirin Sohrabi,et al.  Delfi: Online Planner Selection for Cost-Optimal Planning , 2018 .

[5]  Jianwei Zhang,et al.  Bridging the Gap Between Semantics and Control for Industry 4.0 and Autonomous Production , 2019, 2019 IEEE 15th International Conference on Automation Science and Engineering (CASE).

[6]  Sergey Levine,et al.  QT-Opt: Scalable Deep Reinforcement Learning for Vision-Based Robotic Manipulation , 2018, CoRL.

[7]  Alex Graves,et al.  Playing Atari with Deep Reinforcement Learning , 2013, ArXiv.

[8]  Wolfram Burgard,et al.  Planning Reactive Manipulation in Dynamic Environments , 2019, 2019 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS).

[9]  Jörg Hoffmann,et al.  Optimizing Planning Domains by Automatic Action Schema Splitting , 2014, ICAPS.

[10]  Wojciech M. Czarnecki,et al.  Grandmaster level in StarCraft II using multi-agent reinforcement learning , 2019, Nature.

[11]  Jendrik Seipp,et al.  Learning Portfolios of Automatically Tuned Planners , 2012, ICAPS.

[12]  Michael Beetz,et al.  Configuration of Perception Systems via Planning Over Factor Graphs , 2018, 2018 IEEE International Conference on Robotics and Automation (ICRA).

[13]  Sergey Levine,et al.  Improvisation through Physical Understanding: Using Novel Objects as Tools with Visual Foresight , 2019, Robotics: Science and Systems.

[14]  Demis Hassabis,et al.  Mastering the game of Go with deep neural networks and tree search , 2016, Nature.

[15]  Wolfgang Reif,et al.  Automatic Planning of Manufacturing Processes using Spatial Construction Plan Analysis and Extensible Heuristic Search , 2018, ICINCO.

[16]  Lukás Chrpa,et al.  On the effective configuration of planning domain models , 2015 .

[17]  Bartholomew O. Nnaji,et al.  Feature representation and classification for automatic process planning systems , 1993 .