Portfolio Archive:

CASHMERE – Context-Aware Search and Discovery in Hypermedia-Driven Multi-Agent Environments

In hypermedia-driven Multi-Agent Systems (HyperAgents) the web and its architecture act like a global environment in which different types of agents (whether human or software) perform interactions and compose the functionality of web-enabled things to achieve their goals.

The main objective of this project is the development of a web-based framework facilitating context-aware search and discovery of web-enabled things in hypermedia driven environments. The requirement is to facilitate the detection and understanding of shared context between a producer and a consumer to enable corresponding discovery of and access to relevant web resources to agents operating in the web environment.

To achieve its objective the project aims to:

1) Develop a web-based platform that can expressively model relevant context information of web things and enable a semantic-query and complex-event processing based mechanism to define and detect situations of shared context between web things and agents;

2) Develop a web-based APIs that take the detection of a shared context as input and are able to facilitate the creation of authorization elements that integrate into existing access control and search technologies for web things.

Duration: 2021-2023

AI Folk – Resource Management in Distributed AI

Implementing a vision in which intelligent agents perceive and act in an environment while searching for, exchanging, annotating, and improving machine learning models, forming a culture based on experience and interaction.

Our research goal is to develop a knowledge model and an interaction protocol which allow a system formed of multiple actors — human, software, or organizational — to find, use and share improvements on machine learning resources. Such resources can be datasets, models, or experiences.

We propose the development of the AI Folk framework and methodology, at the intersection of machine learning, knowledge management, and multi-agent systems. It comprises tools and methods that allow the management and discovery of ML-related resources in a distributed system. Federated learning has yet to achieve maturity as a field of study and this is a novel approach which assumes an open system and a variety of resources. We believe that this approach will lead to an advance in the state of the art and will help in the development of standards for open and distributed artificial intelligence.

The project result will be an ontology for describing machine learning resources, a methodology for creating searches for resources, an interaction protocol through which actors can search, transfer and update machine learning resources, the implementation of two applications using this approach, and a general methodology for applying the proposed approach to other application domains.

Duration: 2022-2024

People Detection and Tracking for Social Robots and Autonomous Cars

Detecting, tracking, and recognizing people is a valuable capability for machines. However, these tasks are quite difficult to be achieved autonomously and, although significant results have been obtained, they are still a major technological challenge. People tracking, unlike other recognition and interpretation tasks, is difficult both from the point of view of the recognition and prediction of trajectory, and from the one of the identifications of the ground truth.

The main objective of the PETRA project is the development of a software platform enabling the development of applications requiring people detection and tracking in real environments. The design and implementation of the platform and the set of supported algorithms for people detection and tracking will be such that they can be easily used and integrated in tasks performed by social robots in closed spaces and in tasks in open spaces, such as the case for pedestrian detection and tracking. The scientific and technological challenge of the project is to start from our current developments on people detection and tracking in the contexts of user-robot interaction and autonomous driving to develop and implement novel solutions based on deep learning approaches. One of the project challenges is to extensively test the implementations towards several difficult benchmarks, but also on our own data sets, and strive to achieve results better than current state-of-the-art.

Duration: 2020-2022