Content: Tenth Workshop

Tenth Workshop


The tenth Hybris Workshop will take place on Nov 6th-7th 2017 at RWTH Aachen University, Germany.

Venue

The workshop will be held at RWTH Aachen University, Germany in room 5053.2 of the BIT Research School on the ground floor of the Computer Science building (the same location where we held the project review talks in 2014). When you enter the building from Mies-van-der-Rohe-Straße, you will find a ramp leading to the left, but right next to the ramp is a glass door, which you have to pass. The room will be right in front of you.

Program

TimeTopicSpeaker
Monday, November 6th
12:45-13:00Registration and Welcome
13:00-14:00Agent Planning Programs: Between Planning and ProgrammingSebastian Sardina
RMIT University, Melbourne, Australia
14:00-14:30Planning Approaches for Logistics Robots in SimulationTim Niemueller
RWTH Aachen University
14:30-15:00Planning for RoboCup Logistics League: Lessons Learned and Future ImprovementsAndreas Hertle
University of Freiburg
15:00-15:30Coffee Break
15:30-16:00Generating Macro Actions from a Plan Database for Task Planning on Mobile RobotsTill Hofmann
RWTH Aachen University
16:00-16:30Complexity of Projection with Stochastic Actions in a Probabilistic Description LogicBenjamin Zarrieß
TU Dresden
16:45-18:15Hybris PI Meeting
Tuesday, November 7th
9:00-10:00Decision-theoretic Planning via Probabilistic ProgrammingVaishak Belle
University of Edinburgh, UK
10:00-10:30Coffee Break
10:30-11:00Answer Set Programming for a Knowledge-intensive Planning Problem: Warehouse PlanningSteffen Schieweck
TU Dortmund
11:00-11:30Making Data Science more Accessible by Bridging Data Mining and Artificial IntelligenceFrancois Laferriere
University of Leipzig
11:30-12:00Extending the Description Logic ALC with More Expressive Cardinality Constraints on ConceptsAndreas Ecke
TU Dresden
11:30-12:00Strong Syntax Splitting for Iterated Belief RevisionGabriele Kern-Isberner
TU Dortmund
12:30-13:30Lunch
13:30-15:30Tutorial: Unifying Logic, Dynamics and Probability: Foundations, Algorithms and ChallengesVaishak Belle
University of Edinburgh, UK
15:30Coffee and Farewell

Abstracts

Agent Planning Programs: Between Planning and Programming

Invited talk by Sebastian Sardina, RMIT University, Melbourne, Australia

Abstract: In this talk I will present the Agent Planning Programs high-level paradigm for modelling and controlling agent behaviour, which mixes automated model-based planning with a form of agent-oriented programming. Agent planning programs are finite-state programs, possibly containing loops, whose atomic instructions are not actions but declarative goals. Such programs requires generating plans that meet the goals specified in the atomic instructions (in possibly non-deterministic domains), while respecting the program control flow. After motivating and describing the problem, I will present techniques to automatically synthesize the required plans by exploiting recent significant advances in the fields of reactive synthesis and of model-based planning.

This is joint work with Giuseppe De Giacomo and Fabio Patrizi (Sapienza Università di Roma), Alfonso Gerevini and Alessandro Saetti (Università degli Studi di Brescia): Agent planning programs (in Artificial Intelligence, 231:64--106, 2016).

Slides

Decision-theoretic Planning via Probabilistic Programming

Invited talk by Vaishak Belle, Univerity of Edinburgh, UK

Abstract: We study planning in Markov decision processes involving discrete and continuous states and actions, and an unknown number of objects. Planning in such domains is notoriously challenging and often requires restrictive assumptions. We introduce HYPE: a sample-based planner for hybrid domains that is very general, which combines model-based approaches with state abstraction. Most significantly, the domains where such planners are deployed are usually very complex with deep structural and geometric constraints. HYPE is instantiated in a probabilistic programming language that allows compact codification of such constraints.

In our empirical evaluations, we show that HYPE is a general and widely applicable planner in domains ranging from strictly discrete to strictly continuous to hybrid ones. Moreover, empirical results showed that abstraction provides significant improvements.

In the final part of the talk, we turn to the question of whether there is any hope of developing computational methodologies that are not based on sampling. In particular, it is tricky in hybrid domains to deal with low-probability observations, and most sampling-based schemes only provide asymptotic guarantees.

This talk is based on a Machine Learning Journal article (2017), and is joint work with Davide Nitti, Tinne De Laet and Luc De Raedt.

Slides

Unifying Logic, Dynamics and Probability: Foundations, Algorithms and Challenges

Tutorial by Vaishak Belle, Univerity of Edinburgh, UK

Abstract: Probabilistic reasoning and Bayesian learning are widely used in data management. However, these models are limited in representational power as dependencies are only expressed at the level of propositions. An emerging research initiative attempts to bridge two historically distinct approaches to AI: probabilistic models, which are well-equipped to handle the inherent uncertainty seen in many applications, are merged with first-order logic (FOL), which offers a mathematical framework to formally represent and reason about objects, properties, relations and dependencies.

Nonetheless, such initiatives, mostly for computational reasons, have restricted the kind of syntactical and semantical devices they borrow from FOL. For example, disjunctions in FOL can be used to reason about qualitative uncertainty, and existential quantifiers can be used to reason about identity uncertainty. To accommodate the needs of contemporary and near-future applications in logistics, robotics, planning and data-intensive machine learning, we motivate and review representation languages and automated reasoning techniques that combine the power of logical and probabilistic reasoning in (stochastic) dynamical systems. The material is approached in a general way, and concludes with some open questions in the area.

Slides