Professor Maria Gini

This research group studies a variety of problems in artificial intelligence and multi-agent systems. The problems can require machine learning methods, optimization algorithms, search algorithms that work in very large search spaces, image processing, disitributed computations, and more.  When the search space is very large or the training datasets are huge, the computing resources have to scale up to the size of the problem, necessitating the use of MSI resources. Projects being undertaken include:

• Large scale simulation of mobile robots in a swarm. The simulator models physics in the interactions of the robots with the environment, so it takes significant computing cycles. The researchers have managed to run more than 10,000 robots using MSI resources and want to continue the study to see how well the algorithms scale. They are extending this work in collaboration with a biologist to get inspiration from insects to design algorithms for the robots to learn from other robots and adapt to the environment, which can be unknown and changing.
• Building conversational agents, with deep learning used to train the agent. The conversational agent currently being built is designed for elderly people, who will benefit from personalized interactions and support. The work has the potential to improve the standard of care for the aging population, but could be used as well for other applications. The project needs very large datasets of natural language dialogues to be used to train the system. The group is also working on other projects related to natural language, such as understanding the various meanings of modal verbs, such as can, could, need, will, would, and others, to get subtle insights and understand nuances in human communications. These tasks are computationally intensive and require storage space to store the datasets that have to be processed and used for training.

Research by this group was featured on the MSI website in January 2022: Robot Swarms.