Fabio Mercorio's Curriculum Vitae - of Computer Science

Curriculum Vitae a - Dr Fabio Mercorio 

a Last Update: November 2012 

Contact 

Information 

C.R.I.S.P. Research Centre 

University of Milano - Bicocca Voice: (+39) 320 8581712 

Vialedell’Innovazione10,Room1022-Milano 

- Italy 

Skype: dannyboodmann 

Date of Birth: November 6th , 1983 E-mail: fabio.mercorio@unimib.it 

Nationality: Italian web: www.di.univaq.it/fabio.mercorio 

Current 

Position At CRISP Research Centre, Department of Statistics, University of Milano 

- Bicocca 

Research Fellow in Computer Science November 2011 to present 

• Supervisor: Prof. Mario Mezzanzanica 

• ResearchArea: DataCleansing, DataAnalysisandData QualitythroughFormal 

Methods. 

Employment 

History At CRISP Research Centre, Department of Statistics, University of Milano 

- Bicocca 

Education 

Visiting Student September 2010 to September 2011 

• Supervisor: Prof. Mario Mezzanzanica 

• ResearchArea: DataCleansing, DataAnalysisandData QualitythroughFormal 

Methods. 

At Department of Computer Science , University of L’Aquila 

Ph.D. in Computer Science March 2012 

• Supervisor: Dott. Giuseppe Della Penna 

• Co-Supervisor: Dott. Daniele Magazzeni 

• ResearchArea: FormalMethods andArtificialIntelligence, focusing ontheapplication 

of model checking to deterministic/non-deterministic systems in planning, 

control and data quality problems. 

• Thesis: Model Checking for the Analysis and Control of Complex and Non- 

Deterministic Systems 


Master degree in Computer Science October 2008 

• Supervisor: Dott. Giuseppe Della Penna 

• Co-Supervisor: Dott. Daniele Magazzeni 

• Thesis title: Un Planner per domini PDDL+ basato su tecniche di Model Checking 

( A PDDL+ planner based on Model Checking Techniques) 

• Score: 110/110 cum laude 

1 of 7


Bachelor Degree in Computer Science February 2007 

• Supervisor: Prof. Guido Proietti 

• Thesis title: Algoritmi Distribuiti per il Mantenimento di una Minima Foresta 

Ricoprenteinun Grafo( Distributed Algorithms to maitain a Minimum Spanning 

Forest in a Graph) 

• Score: 107/110 

Research 

Activity MyresearchinterestsaremainlyinArtificial Intelligence PlanningandData Analysis, 

both addressedin my PhD Thesis [15], with particular attention to the application 

of Formal Methods. My contributions as well as a description of my current research 

and plans for the next few years are reported in the following: 

Artificial Intelligence Planning. 

My research activity started by addressing planning and universal planning for 

systems with continuous change. Indeed, many realistic planning problems in which 

planningcanbeappliedinvolveamixtureofcontinuous(possiblynonlinear)anddiscrete 

behaviours, also known as hybrid systems. To deal with such kind of systems, recently 

the AI Planning Community has defined the PDDL+ language which is also able to 

capture continuous processes and events. 

As a first step, I was co-developer of the UPMurphi planner, as presented in [14], 

whichperformsplanningbyexploitingtheplanning-as-model-checkingparadigm. Then, 

I defined and implemented a formal mapping between the PDDL+ and UPMurphi 

semantics, then applying UPMurphi to a number of real-world planning applications 

(see [2],[11, 13, 12] for details). 

The main issue in planning through model checking is the state explosion phenomenon 

as the region of reachable states is often huge. To mitigate this problem, I worked to 

enrich the UPMurphi planner with a disk-based algorithm, namely the V-UPMurphi 

tool (see [15],[5] for details) which allows it to use the disk for storing both the system 

dynamics and the final solution, using OBDDs to compress the resulting plans through 

the C++ CUDD library. 

My research interest also investigates the planning in non-deterministic domains. 

Indeed, while in classical planning a plan always reaches a goal, in non-deterministic 

domains an action may have several outcomes, which could make the plan unable to 

reach a goal. The concept of strong plan fills this gap by ensuring that a plan always 

achieves a goal regardless of the non-determinism. In this field, I have been working 

on the development and analysis (proving correctness, completeness and complexity) 

of an algorithm which, as a novel contribution, (1) considers a R + cost function looking 

for cost-optimal strong plans; (2) extends the synthesis of strong planning also to 

hybrid and/or nonlinear domains (see [9] for details). I implemented the cost-optimal 

strong planning algorithm on the top of V-UPMurphi tool, namely the SUPMurphi 

tool, applying it to some motivating non-deterministic domains [5]. 

A complete summary of this part of my work has been accepted for the publication [3]. 

Data Analysis. 

Business and governmental applications, web applications, ongoing relations between 

citizens and public administrations generate a lot of data, many of which can be considered 

as longitudinal data (i.e., repeated observation of the same object along the time). 

Such data are often used in several decision making activities as evaluation of active 

policies, resource allocation, service design and improvement. However, the lower the 


Teaching 

Students 

Supervision 

quality of the data, the lower the effectiveness of the statistics derived from it. Hence, 

data quality analysis and cleansing activities became required tasks before the decision 

making process. 

Given my background on Model Checking, as a first contribution I addressed the problem 

of the data quality verification of databases as a model checking problem. 

In this settings, a solution for the latter (if any) represents a data quality problem for 

the former. I extended the application of model checking by defining the Robust Data 

Quality Analysis, a technique to support and improve existing cleansing activities, applying 

it to the CO System of the Italian Labour Market Domain studied at the CRISP 

Research Centre (see [10],[6]). 

Since Model Checking is well suited to explore huge state spaces, I have been workingto 

perform a sensitivity analysis on the cleansed data. The impact that the cleansing 

process may have on the aggregate indicators computed on the cleansed data is investigated 

by considering all the alternative cleansed versions of the original dirty data. 

The results can be seen as a confidence interval about the reliability of the computed 

indicator (see, e.g. [7]). 

Recently, I inherited the concept of AI universal planning by defining the universal 

cleanser: a repository of all the feasible actions able to cleanse a dirty data. It is 

automatically synthesised off-line on the basis of the UPMurphi algorithm. As a characteristic, 

it is domain-dependent, i.e. it can deal only with database data conforming 

to the model that generated it, but data-independent since, once computed, it can be 

used to cleanse any dataset conforming to the model [8]. 

At Master in Business Intelligence and Decision Support, University of Milano-Bicocca 

Lecturer about Data Analysis, academic year 2011/2012. 

At Department of Statistics , University of Milano-Bicocca 

Teaching Assitent, academic years 2011/2012 and 2012/2013. 

• Informatica at Course of Statistics - Faculty of Statistics Science - University of 

Milano-Bicocca 

• Laboratorio di Informatica at Course of Statistics - Faculty of Statistics Science 

- University of Milano-Bicocca 


Teaching as lecturer, academic years 2008/2009 and 2009/2010. 

• Advanced invariants modelling using Linear Temporal Logic and Computational 

Tree Logic (LTL, CTL, and CTL*), at Formal Methods for Complex Systems 

Verification course 

• Introduction to The SPIN model Checker, at Formal Methods for Complex Systems 

Verification course 

Co-supervised Bachelor’s theses: 

• FrancescoScoccia. Planning con Algoritmi Genetici (Jointly supervised with Dr. 

Giuseppe Della Penna). December 2010. 


Research 

Software 

Developed 

Publications 

* 

* 

• FrancescoAntonelli. Controllo automatico di unapparato di sollevamento idraulico 

per installazioni antisismiche (JointlysupervisedwithDr. GiuseppeDellaPenna). 

March 2010. 

• Luigi Grassano. Pianificazione di un processo di produzione chimica tramite tecniche 

di planning basate su model checking (JointlysupervisedwithDr. Giuseppe 

Della Penna). October 2009. 

• UPMurphi: The Universal Planner Murphi tool representsa computational engine 

to support universal planning for possibly nonlinear Discrete Time Hybrid Systems 

defined with a PDDL+ model. We provided a formal mapping process which translates 

a PDDL+ model into a semantically equivalent UPMurphi model. The tool 

translates PDDL+ models into UPMurphi models and then outputs a collection of 

optimal plans according to the PDDL+ syntax. Moreover, UPMurphi is able to validate 

(using the standard PDDL+ VAL validator) all the generated plans. Related 

Publications:[15],[14, 11, 12, 13][2, 4, 3]. 

• V-UPMurphi: It is a computational engine to enhance the ability of UPMurphi to 

synthesise plans and universal plans for (possibly nonlinear) Discrete Time Hybrid 

Systems defined with a PDDL+ model. V-UPMurphi exploits the use of the disk 

during the exploration of the dynamics. It implements on disk the state space 

reduction techniques inherited from Murphi (as bit-compression, hash-compaction) 

and allows one to pause and resume the system analysis process, using the disk to 

both explore and store the graph of the system dynamics. Related Publications:[3, 

2][15] 

• SUPMurphi: The Strong Universal Planner Murphi implements the Cost Optimal 

Strong Plan Algorithm described in j2. It uses the Murphi description language 

to model non-deterministic behaviour of systems’ actions and, thanks to the use 

of disk-based algorithm, it is able to address strong planning problems. Then, it 

allows one to validate the final strong plan (if any) on the system graph. Related 

Publications:[9][3, 5][15] 

Journal Articles 

[1] GiuseppeDellaPenna,BenedettoIntrigila,DanieleMagazzeni,andFabioMercorio. 

Synthesis of cost-optimal strong plans in non-deterministic domains. Applied 

Intelligence (under review), 2012. 

[2] Giuseppe Della Penna, Daniele Magazzeni, and Fabio Mercorio. A universal 

planning system for hybrid domains. Applied Intelligence, 36(4):932–959, 2012. 

[3] Fabio Mercorio. Model checking of universal planning in deterministic and nondeterministic 

domains. AI Communications (to appear), 2012. 

[4] Giuseppe Della Penna, Benedetto Intrigila, Daniele Magazzeni, and Fabio Mercorio. 

Resource-optimal planning for an autonomous planetary vehicle. International 

Journal of Artificial Intelligence & Applications (IJAIA), 1(3):15–29, 

2010. 

Journal Articles (under review) 

[5] GiuseppeDellaPenna,BenedettoIntrigila,DanieleMagazzeni,andFabioMercorio. 

Synthesis of cost-optimal strong plans in non-deterministic domains. Applied 

Intelligence (under review), 2012. 


* 

* 

International 

Schools and 

Conferences 

[6] Mario Mezzanzanica, Roberto Boselli, Mirko Cesarini, and Fabio Mercorio. Longitudinal 

data consistencyverificationusing formal methods. International Journal 

of Information Quality (under review), 2012. 

Referred Conference Papers 

[7] Mario Mezzanzanica, Roberto Boselli, Mirko Cesarini, and Fabio Mercorio. Data 

quality sensitivity analysis on aggregate indicators. In Markus Helfert, Chiara 

Francalanci, and Joaquim Filipe, editors, DATA 2012 - Proceedings of the International 

Conference on Data Technologies and Applications, pages 97–108. 

SciTePress, 2012. 

[8] Mario Mezzanzanica, Roberto Boselli, Mirko Cesarini, and Fabio Mercorio. Towards 

the use of model checking for performing data consistency evaluation and 

cleansing. In The 17th International Conference on Information Quality (ICIQ 

2012) (to appear), 2012. 

[9] Giuseppe Della Penna, Benedetto Intrigila, Daniele Magazzeni, Fabio Mercorio, 

and Enrico Tronci. Cost-optimal strong planning in non-deterministic domains. 

In Proceedings of the 8th International Conference on Informatics in Control, 

Automation and Robotics (ICINCO), pages 56–66. SciTePress, 2011. 

[10] Mario Mezzanzanica, Roberto Boselli, Mirko Cesarini, and Fabio Mercorio. 

Data quality through model checking techniques. In João Gama, Elizabeth 

Bradley, and Jaakko Hollmén, editors, Intelligent Data Analysis (IDA), Lecture 

Notes in Computer Science vol. 7014, pages 270–281. Springer, 2011. 


A PDDL+ benchmark problem: The batch chemical plant. In Proceedings 

of the The 20th International Conference on Automated Planning and Scheduling 

(ICAPS 2010), pages 222–225, Toronto, Canada, 2010. AAAI Press. 


Planning for autonomous planetary vehicles. In Proceedings of the The 

Sixth International Conference on Autonomic and Autonomous Systems, pages 

131–136, Cancun, Mexico, 2010. IEEE. 

[13] Fabio Mercorio. Planning for continuous domains. In The AI*IA Doctoral 

Consortium, Brescia (Italy) December 1-3, 2010. 


UPMurphi: atool foruniversalplanningon PDDL+ problems. In Proceedings 

of the 19th International Conference on Automated Planning and Scheduling 

(ICAPS 2009), pages 106–113, Thessaloniki, Greece, September 2009. AAAI 

Press. 

Thesis 

[15] Fabio Mercorio. Model Checking for the Analysis and Control of Complex 

and Non-deterministic Systems. PhD thesis, Department of Computer 

Science - University of L’Aquila - Italy. (public available at 

http://hdl.handle.net/10281/36432). 

• Speaker at 8th International Conference on Informatics in Control, Automation 

and Robotics. 


Foreign 

Languages • Mother tongue: Italian 

Noordwijkerhout, The Netherlands July 28-31, 2011 

• Speaker at the AI*IA Doctoral Consortium on Artificial Intelligence. 

Brescia, Italy December 1-3, 2010 

• At eBISS 2012 Summer School on Business Intelligence 

Brussels, Belgium July, 16-21, 2012 

• At VTSA 2009 Summer School on Verification Technology, Systems & 

Applications 

Nancy, France October 12-16, 2009 

• AtICAPS 2009 Summer School on Automated Planning and Scheduling 

Thessaloniki, Greece September 16-18, 2009 

Language Understanding Speaking Writing 

Listening Reading Spoken Spoken 

interaction production 

English C2 proficient C2 proficient C1 proficient C1 proficient C2 proficient 

user user user user user 

French A1 basic user A1 basic user A1 basic user A1 basic user A1 basic user 

Work 

Experience At Cyborg S.r.l for Mediterranean Games 2009 

Role: Information Technology Manager May-July 2009 

• Activity description: I managed the Information Technology Sector (ITS) for 

the Boxing competition held in ”Avezzano - Palaghiaccio” during the sixteenth 

edition of the ”Mediterranean Games“. 

• Abilities and Skills achieved: Cisco Systems and RAC configuration, Windows 

and Linux Printer server configuration, Windows Active Directory, PRINCE2 

management system, project and staff management. 

and Software costumer support, Network configuration and security managing. 

Technical 

knowledge • Programminglanguages: AdvancedknowledgeaboutJAVA(J2EE,J2SE,J2ME), 

C/C++, PHP, Javascript, Assembler, COBOL, Coq, ML, Scheme, Prolog, DALI, 

Matlab. 

Advanced knowledge about Object Oriented Design Pattern. 

Advanced knowledge about C/C++ optimization programming. 

Advanced knowledge about client-server programming and mobile application programming 

• Markup languages: Advanced knowledge about HMTL, XML, XSLT, XQUERY, 

XPATH 

• Develop Frameworks: Advanced knowledge about Rational Rose, NetBeans, 

Eclipse, Vistual Studio C++ Builder 

• DBMS: Advanced knowledge about SQL Server, DB2, MySql, Interbase 

• Web Server: Advanced knowledge about Apache 1/2.x, Microsoft IIS. 


• Operating Systems: Advanced knowledgeaboutWindowsVista/XP/NT{Server, 

Professional} editions - Linux (Debian, RedHat, Ubuntu, Slackware) and their system 

calls. 

• Network Protocols: Good knowledge about GSM ,GPRS, UMTS protocols. 

Advanced knowledge about ISO/OSI layers, in particular TCP/IP - UDP- ARP - 

MAC and Wireless communication protocols. 

Advanced knowledge about routing protocols as Open Shortest Path First (OSPF), 

BGP and RIP 1/2.x 

Advanced knowledge about Microsoft Office Suite (I have obtained the European 

Computer Driving Licence certification) 

• Theoretical Qualifications: Advanced knowledge about Formal Methods and 

complex System Software Verification 

Advanced knowledge about Model Checking and AI Planning problems 

Advanced knowledge about Planning and Scheduling problems algorithms and their 

formulations 

Advanced knowledge about Distributed Algorithms and Graph Theory 

Good knowledge about Game Theory 

Advanced knowledge about Operating Research, Combinatorial Optimization, PL 

and PLI problems formulation 

Advanced knowledge about Genetics and Artificial Intelligence 

Good knowledge about Neural Networks.

Fabio Mercorio's Curriculum Vitae - of Computer Science

Create successful ePaper yourself

Delete template?

Save as template?