- Understanding intent is a critical aspect of communication among people and for many biological systems. This is particularly important in situations that involve collaboration among multiple agents or assessment of potential threats. During the recent years, there has been an increased interest in using autonomous technologies for security and defense applications, in order to reduce the danger for the people involved. However, the current systems deployed by the US Army and the US Air Force rely heavily on input from a human operator who assesses the situation and takes a decision. Thus, in the context of these applications, being able to automatically detect any threatening situations is of critical importance. This reduces to the problem of understanding the intent of the other agents in the environment, from their current actions, before any attack strategies are finalized. Toward this end, the main research problem we address in this project are to develop a tool for modeling the behavior of relevant agents (boats), which will incorporate the predictive capabilities necessary to infer those agents’ intentions.
- The goal of this project is to design an integrated system for automatic behavior modeling, which provides effective detection of intentions, before any actions have been finalized. Our approach relies on a novel formulation of Hidden Markov Models (HMMs), which allows a robot to understand the intent of others by virtually assuming their place and projecting their potential intentions based on the current situation. In contrast with the standard use of HMMs, which solely model transitions between discrete states, in this work the HMMs also model the rate and direction of change for a set of relevant parameters. This approach allows the system to recognize the intent of observed actions before the action has been completed, thus enabling the system to take preemptive actions for defense.
- Logan Carlson, Dalton Navalta, Monica Nicolescu, Mircea Nicolescu, Gail Woodward, “Early Classification of Intent for Maritime Domains using Multinomial Hidden Markov Models”, Frontiers in Artificial Intelligence – Special Issue on Advances in Goal, Plan and Activity Recognition, vol. 4, pages 1-11, October 2021.
- Logan Carlson, Dalton Navalta, Monica Nicolescu, Gail Woodward, Mircea Nicolescu, "Multinomial HMMs for Intent Recognition in Maritime Domains", Proceedings of the International Conference on Autonomous Agents and Multiagent Systems, Montreal, Canada, May 2019.
- Logan Carlson, Dalton Navalta, Monica Nicolescu, Mircea Nicolescu, Gail Woodward, "Intent Classification in Maritime Domains with Multinomial HMMs", Proceedings of the International Workshop on EXplainable TRansparent Autonomous Agents and Multi-Agent Systems, pages 1-15, Montreal, Canada, May 2019.
- Mohammad Saffar, Mircea Nicolescu, Monica Nicolescu, Daniel Bigelow, Christopher Ballinger, Sushil Louis, "Intent Recognition in a Simulated Maritime Multi-Agent Domain", Proceedings of the International Workshop on Machine Learning, Optimization and Big Data, pages 1-12, Taormina, Italy, July 2015.
Detecting a blockade by a group of small boats
Detecting a hammer and anvil attack by a group of small boats
Detecting of deceptive behavior (hiding behind other boats)
Detecting of deceptive behavior (hiding behind land)
Detecting threatening intentions in a complex scenario in San Diego Harbor (boats going in formation, potential attacks)
Detecting threatening intentions in a complex scenario in open water (boats forming groups, splitting, hiding and attacks)
Detecting threatening intentions in a complex scenario in the Straits of Hormuz (small boats harassing and attacking destroyer in channel)
- Intent Recognition for On-Water Adversarial Agents, Office of Naval Research, PI (Co-PI: Mircea Nicolescu), Amount: $900,000, March 2022 - February 2025.
- Intent Recognition for On-Water Dynamic Maritime Domains, Office of Naval Research, PI (Co-PI: Mircea Nicolescu, Terry Huntsberger - JPL), Amount: $685,355, March 2016 - February 2018.
- Understanding Intent Using an Activation Spreading Architecture, Office of Naval Research, PI (Co-PI: Mircea Nicolescu, Sushil Louis), Amount: $590,992, July 1, 2012 - June 30, 2015.
- Context-Based Intent Understanding for Autonomous Systems in Naval and Collaborative Robotics Applications, Office of Naval Research, PI (Co-PI: Mircea Nicolescu, Sushil Louis), Amount $543,280, August 1, 2009 - July 31, 2012.
- Understanding Intent Using a Novel Hidden-Markov Representation, Office of Naval Research (ONR), PI, (Co-PI: Mircea Nicolescu), Amount: $619,584, June 1, 2006 - May 31, 2009.