SCANNER - Secure CollAborative recogNitioN of complEx tasks with Robots
Abstract
The problem of automatically interpreting the surroundings using robots equipped with different sensors, scene understanding, is a widely studied problem that includes all kinds of recognition tasks. As such, nowadays most organizations and divisions with some technological base have their own robotic platforms and scene understanding algorithms. The heterogeneity of such systems can present a challenge in the context of collaborative missions, making the exchange of useful information difficult. At the same time, individual parties need to trust each other and the information they exchange so that they are secure against external entities not involved in the process. Therefore, there is a strong need for new methods and solutions that allow collaborative scene understanding seamlessly and safely, independently of the individual robotic platforms and perception methods. In this project we plan on making this possible by proposing new perception algorithms and distributed communicationprotocols that provide a team of robots the chance to (i) effectively recognize complex tasks that involve multiple agents (e.g., robots and humans working together) and (ii) preserve desired integrity levels in their own information. Target applications may include area monitoring, scouting and construction of infrastructure.We will leverage state of the art distributed optimization algorithms and recent breakthroughs from Foundation models to produce innovative basic research in the general problem of 4D semantic scene understanding with teams of different robots. More in detail, advances in computer vision and machine learning have brought new architectures and algorithms able to provide efficient computation and rich semantic information about the environment. Similarly, distributed consensus-based protocols have made possible to solve complex optimization problems scattered across communication networks. Lastly, formal languages are a powerful tool capable of describing complex systems that can be exploited to demonstrate structural properties with high accuracy and rigor. We will combine these three tools to fulfill the goal of a successful SCANNER (Secure CollAborative recogNitioN of complEx tasks with Robots).In the long term, it is expected that with our work, the Navy will earn the capability of deploying different teams of intelligent robots that are able to collaborate in complex perception tasks despite their different capabilities and peculiarities of their individual perception systems.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Nov 08, 2024
- Source ID
- N629092412081
Entities
People
- Eduardo Montijano
Organizations
- Office of Naval Research
- United States Navy
- University of Zaragoza