PRESENTATION OF THEOVERLAY
RESEARCH GROUP
Amedeo Cesta, Alessandro Cimatti, Luca Geretti, Alfonso Gerevini, Angelo Montanari, Adriano Peron and Tiziano Villa
IWES 2019, Naples, Sep 30 – Oct 1, 2019
fOrmalVERification,Logic,Automata andsYnthesis
THE NETWORK
35 MEMBERS FROM 13 INSTITUTIONS
• Amedeo Cesta, Riccardo de Benedictis, Andrea Orlandini, Alessandro Umbrico
ISTC-CNR, Rome
• Luca Benvenuti, Enrico Tronci
University of Rome La Sapienza
• Federico Mari
University of Rome Foro Italico
• Marco Bozzano, Alessandro Cimatti, Andrea Micheli, Marco Roveri, Stefano Tonetta, Paolo Traverso
Fondazione Bruno Kessler, Trento
• Guido Sciavicco
University of Ferrara
• Davide Bresolin
University of Padova
• Luca Geretti, Pietro Sala, Tiziano Villa, Matteo Zavatteri
University of Verona
• Alfonso Gerevini, Enrico Scala
University of Brescia
• Massimo Benerecetti, Marco Faella, Fabio Mogavero, Adriano Peron
University of Naples Federico II
• Salvatore La Torre
University of Salerno
• Dario della Monica, Agostino Dovier, Luca Geatti, Nicola Gigante, Gabriele Puppis, Angelo Montanari, Carla Piazza
University of Udine
• Giorgio Delzanno
University of Genoa
• Ivan Lanese
University of Bologna
STRUCTURE
FOUR AREAS
• Alessandro Cimatti
• Tiziano Villa
• Alfonso Gerevini
• Adriano Peron
FORMAL METHODS
AUTOMATED PLANNING
TOOLS AND APPLICATIONS
DISCRETE AND
CONTINUOUS HYBRID
SYSTEMS
General coordination: Amedeo Cesta, Angelo Montanari
MOTIVATION
FORMAL METHODS MEET AI: AN AGENDA
THEO
RETI
CAL
CON
TRIB
UTIO
NS
Establish theoretical limitations of artificial intelligentsystems
Contribute to explainable AI
PRAC
TICA
L CO
NTR
IBUT
ION
S
Empower AI methods with successful methods based on formal/symbolic reasoning (FM for AI)
Enrich formal methods with machine learningapproaches (AI for FM)
Primary objective: improve the synergy between model-free (data driven) and model-based (representation driven)
approaches
DESIGNING RELIABLE HARDWARE & SOFTWARE SYSTEMS
AREA:FORMAL METHODS
Mathematical techniques for the specification,validation, verification, and synthesis of systems
Games
Automata
Logics
Specificationl Temporal Logicsl Multi-Agent Logics
Validation & Verificationl Model Checkingl Automated Test Generation
Synthesisl Infinite Games on Graphsl Strategic Reasoning
MISSION AND METHODOLOGY
AREA:FORMAL METHODS
Modeling and Automatic Analysis of Complex Computational Systems
l Modeling languages for: concurrent distributed systems;real-time systems
l Identification of expressive but “tractable” specification languages
l Efficient decision algorithmsl Runtime verification techniques
Tasks and methodologies
DISCRETE AND CONTINUOUS
HYBRID SYSTEMS
AREA:
DESIGNING LARGE-SCALE HETEROGENEOUS SYSTEMS
Distributed networked sensor and actuator platforms introduce new challenges and opportunities
Increased complexity of engineered systems
Need to co-design both the physical and control aspects
Increased reliance on automated control of processes
Environment
Communication
Computation
DISCRETE AND CONTINUOUS
HYBRID SYSTEMS
AREA:
FORMALIZATION OF THE DESIGN PROCESS
Inspired by the success of digital system design, we followa flow comprising specification, analysis and synthesis
• Identification of specificationlanguages
• Identification of component boundaries
• Reachability analysis• Abstraction-based refinement• Synthesis of the unknown
component
Tasks and methodologies
A CONTRACT-BASED APPROACH
HIERARCHY
HIDINGCOMPOSITION
AREA:AUTOMATED
PLANNING
ENABLE AGENTS TO PLAN AND MEET GOALS IN A TIMELY AND SAFE FASHION
An autonomous systemneeds to reasons over
its actions, perceptions, goals
Environment and agent actions through formal models of how the world change
Application: space exploration, logistics, UAVs, mobile robots,..
AREA:AUTOMATED
PLANNING
AUTOMATED PLANNING CHALLENGES IN OVERLAY
● Environment○ Deterministic○ Non-Deterministic○ Partial Observability○ Sensing
● State Spaces, Goal Types○ Discrete○ Numeric○ Timed vs Untimed
● Action-Based vs Timeline-Based
Models and Formalisms
● Plan Generation● Plan Adaptation● Goal Reasoning● Incremental Planning● Dynamic environments...
Tasks in AutomatedPlanning
● Heuristic Search● Local Search● Compilation into other Formal
Languages (SMT)● Learning the Search Space● Relaxation-based abstractions● Counterexample-guided
abstractions
MethodsEmployed
TOOLS ANDAPPLICATIONS
AREA:
DEVELOP AN INTEGRATED TOOL CHAIN
• SAT, SMT, ASP• Analysis of finite and
infinite transition systems• Temporal networks,
Transition-level modeling• Automated Test Case
generation• Run-time monitoring
Backends to integrate
• Harmonization of heterogeneous tools
• High degree of automation
• Scalability to large models• Embedding within
commonly used IDEs• Explanation, qualification
of artifacts
Challenges
TOOLS ANDAPPLICATIONS
AREA:
DEMONSTRATORS OF CAPABILITIES
Collection of real-world (success) stories, to demonstratetechnological capabilities
• Critical reconstruction• Lessons learnt, reasons
for success/failure• What works where,
and why
Methodology
• Domains: avionics, space, flexible production, railways, scheduling, etc.
• Problems: requirementsanalysis, design verification, safety assessment, fault detection, design spaceexploration
Real-world stories
THE WEBSITEOVERLAY.UNIUD.IT
o A description of the group and the areaso List of members o Events
WORKSHOP
OVERLAY.UNIUD.IT/WORKSHOP/2019
First Workshop on
Artificial Intelligence and fOrmalVERification, Logic, Automata
and sYnthesisOVERLAY @ AIIA 2019
November 19-20 2019, Rende (Italy)