My long-term research agenda is to develop tools and methodologies to assist in designing, validating, and implementing complex systems known as Cyber-Physical Systems (CPS).

RESEARCH IN CYBER-PHYSICAL SYSTEMS MODELING

CPS can be found all around us, for example, in Avionics or Autonomous Vehicles, but also in applications such as Building Automation Systems (BAS), Energy Management Systems (EMS), Smart Grids (SG), or Intelligent Retail.

CPSs are often regarded as networks of multi-physical (mechanical, electrical, biochemical, etc.) and computational (control, signal processing, logical inference, planning, etc.) processes, often interacting with a highly uncertain and adverse environment, including other CPS and human actors. CPSs are notoriously complex to design and implement, mainly because of their cross-discipline borders, leading to inter-domain interactions.

My current goal is to model these systems at the appropriate level of abstraction so that implementations (possibly partial) can be automatically derived. Overall, this approach can drastically cut the effort required to model and deploy CPS.

At Sensei Tech I coordinate the develoment of an incredibly sophisticated CPS for Autonomous Retail that combines embedded systems with AI and services integration.

SENSOR DATA PROCESSING

Applications such as BAS, EMS, or Smart Grids (SG), can be understood in real-time data flows from sensors to actuators. Technically, these data flows can be expressed as Real-Time Data Streaming Queries that can be efficiently processed using engines known as Data Stream Management Systems (DSMS). I have been particularly interested in expressing complex CPS logic using these systems. Expectably, from an engineering point of view, systems conceived utilising this approach will be much easier to understand, maintain, and evolve.

My hypothesis is that a useful sub-class of CPSs can be successfully addressed as a Sensor Data Processing (SDP).

Several research questions are not yet answered regarding the use of DSMS to implement CPS, among which:

  • Are DSMSs an adequate architectural building block for CPS?
  • Which classes of CPS control problems can be solved in practice with DSMSs? (not every DSMS can run in embedded hardware)
  • What extensions are required for DSMS and, in particular, to Stream Query Languages to address CPS problems? (provisions for declaratively expressing data-retention, timeliness constraints, data quality, and inference models)

MODEL DRIVEN ENGINEERING

Model-Driven Engineering (MDE) is an advanced Software Engineering technique that simplifies the design of complex systems by capturing high-level domain models. MDE works from the idea of describing domains in a language that closes the semantic gap (i.e. minimises the cognitive gap) between the problem and the solution as much as possible. Models can then be analysed and simulated to validate desired properties and then be subjected to automatic synthesis.

I believe that MDE can be successfully employed to model, verify, and automatically synthesise control logic for the subclass of CPS that corresponds to SDP problems. Moreover, I believe that multi-paradigm modelling can be used in the process.

For this to be possible, the following questions, among others, need an answer:

  • What is the best modelling language to express Sensor Data Processing logic?
  • Can functional and extra-functional aspects be modelled separately and then integrated successfully?
  • Can SysML and related tools be used in this effort?
  • Can we successfully integrate SDP logic with logic created in other formalisms? (One idea is to generate SDP specified-controllers and then test them on models physics simulators created in Modelica; these simulators could be created from models as well)

BUILDING INFORMATION MODELLING

Building Information Modelling is a surprisingly rich and effervescent field in problems and solutions regarding complex models. A BIM is a super meta-model that enables the formalisation of multiple aspects of a building. A recent extension proposal to BIM aims at capturing BA aspects as well. BIM is also exciting due to the availability of tools.

I aim to verify and simulate the behaviour of a building (including its BA system)  entirely from a BIM specification.

The idea is to extract a Physics Model and a Control System model from the BIM model. The impact of this research is far-reaching: it would enable tunning the BA system to maximise energy efficiency and occupant comfort.

In this effort several questions have to be answered:

  • What are the required extensions to BIM so that Building Automation behaviour can also be specified?
  • What is the best way to model BA aspects and then translate them into BIM? (BIM is just a way to encode the model)
  • Is it possible to create BA component libraries whose behaviour could then be used in a simulation?

RESEARCH IN DATA PROCESSING AND MANAGEMENT

Another field of study that I am passionate about is Database Systems. I have created efficient tools for handling complex data transformations and then studied how the engines of RDBMS could be extended to support these transformations. This work front has required significant knowledge in:

  1. Theoretical aspects of query languages.
  2. Semantics of programming languages.
  3. Domain-specific languages.
  4. Schema matching and mapping.
  5. Data migration, integration and cleaning.
  6. Query processing and optimisation.

OTHER TOPICS OF INTEREST

  • Computer Vision
  • Automatic Software Verification and Validation
  • Compiler Construction and Program Analysis

RESEARCH PROJECTS

APPLIED R&D and TT PROJECTS

  • 2014-2018 BIMK Project (Industry funded)
  • 2014-2014 PROVIDERESULTS/BIM (Industry funded)
  • 2011-2011 Lumin SI2 (Industry funded)
  • 2010-2010 DF-PROTECT (Industry funded)