Architectures for Embedded Artificial Intelligence

Starting date
January 1, 2020
Duration (months)
12
Departments
Computer Science
Managers or local contacts
Fummi Franco

The objective of the research contract is to identify methodologies for the realization of embedded architectures able to run AI edge applications.
The contract will be based on the following activities:
1. Definition of a Matlab-based environment for the modeling of the AI problem to solve and identification and optimization of the AI technique for its solution.
2. Definition of a Simulink-based environment for simulating the AI technique. This technique is identified in an environmental context managed by BOX-IO for the acquisition of real data and the execution of any activation commands.
3. Configuration of the automatic generation environment of C code optimized for execution on the reference BOX-IO platform.
4. Creation of an interface library between the embedded BOX-IO software levels and the AI application for the effective inclusion of these applications.
5. Creation of an interface library between the AI application and the BOX-IO cloud environment in order to reuse the same application in the two contexts.
These developments will be applied to different contexts including the ICE laboratory of the department where BOX-IO architectures are present.

Sponsors:

EDALab s.r.l.
Funds: assigned and managed by the department

Project participants

Alessia Bozzini
Research Scholarship Holders
Franco Fummi
Full Professor
Research areas involved in the project
Sistemi ciberfisici
Embedded and cyber-physical systems

Activities

Research facilities