Future autonomous systems to understand surroundings

By U.S. Army CCDC Army Research Laboratory Public AffairsJune 18, 2020

Army researchers develop a one-of-a-kind computational resource known as REO that will enhance human-agent teaming and Soldier protection.
Army researchers develop a one-of-a-kind computational resource known as REO that will enhance human-agent teaming and Soldier protection. (Photo Credit: Shutterstock) VIEW ORIGINAL

ADELPHI, Md. -- Army researchers developed a one-of-a-kind computational resource to enhance human-agent teaming and Soldier protection by allowing autonomous systems to understand both the world around them and Soldiers’ conveyed intent in a given mission.

Dr. Claire Bonial, a social scientist for the U.S. Army Combat Capabilities Development Command’s Army Research Laboratory, is one of the researchers behind the resource, known as Rich Event Ontology, or REO, which supports the Soldier Lethality Army Modernization Priority.

REO is a one-of-a-kind resource that supports computational recognition, understanding and reasoning about events and the things of the world both in language and conceptually, Bonial said.

“This research supports the state-of-the art in event detection, which is truly pivotal for autonomous systems to be able to understand what is going on, who is doing it, where is it happening, how and why,” Bonial said.

In REO, four of the most widely-used semantic markup schemas, each with valuable manually-labeled data showing who is doing what to whom, are unified into one resource.

“This allows, for the first time, these distinct efforts to be combined into a larger, more diverse training corpus, supporting machine learning and automatic recognition of events and their participants across the widest variety of events possible,” Bonial said.

Additionally, the ontological structure of REO contributes temporal and causal relations between events (for example, taking sides is a precondition for protest) that do not exist in any of the original markups, she said.

REO also adds commonsense information about how things relate to events—specifically, what the purpose of something is (armor is for protecting) and how it generally comes about (tire tracks are created by a wheeled vehicle driving through a soft surface), Bonial said.

“For example, REO would support an autonomous system in reporting back that a space was likely recently occupied, given the presence of a soda can, and that a fire had taken place, given smoke and charring of objects,” Bonial said. “Such reasoning goes far beyond much of the current state-of-the-art, which focuses on mere recognition of objects and events, without any deeper reasoning.”

According to Bonial, REO facilitates transformational overmatch as it allows our autonomous systems to understand both the world around them and the Soldier’s conveyed intent better than our adversaries’ systems can understand those things.

“This allows our systems to be a deterrent and provide protection to our Soldiers as they coordinate and execute missions more efficiently than our adversaries with the help of superior autonomous aids,” Bonial said.

Bonial and team are eager to continue this research and get it into the hands of Soldiers as soon as possible.

“I am optimistic that this research will help future Soldiers, and it is very important to me that we can transition this technology to the Soldier’s advantage, as it is the culmination of research I have done as both a student, postdoctoral fellow and CCDC ARL researcher throughout the past decade,” Bonial said.

The next step for this research will be to implement and evaluate REO into autonomous systems.

“We will be leveraging REO to assist in agents’ understanding mission instructions, recognizing partially obstructed objects, and for reasoning about how entities and events in the surroundings affect, or should be reported for, a particular mission,” Bonial said.

The team reported their research in a book chapter, The Rich Event Ontology: Ontological Hub for Event Representations, by Caselli, T., Palmer, M., Hovy, E., and Vossen, P. (editors) in Computational Analysis of Storylines: Making Sense of Events, Cambridge University Press.

The book, which is set to be available later this year, emerged from ongoing research discussions, presentations and shared tasks of the 2016 and 2018 EventStory Workshops, which Bonial helped to organize.

Bonial’s chapter of the book reviews the current landscape of ontological and lexical resources that motivated the development of REO.

(Photo Credit: U.S. Army) VIEW ORIGINAL

CCDC Army Research Laboratory is an element of the U.S. Army Combat Capabilities Development Command. As the Army's corporate research laboratory, ARL discovers, innovates and transitions science and technology to ensure dominant strategic land power. Through collaboration across the command's core technical competencies, CCDC leads in the discovery, development and delivery of the technology-based capabilities required to make Soldiers more lethal to win the nation’s wars and come home safely. CCDC is a major subordinate command of the U.S. Army Futures Command.