Researchers of the US Army along with the University of Southern California’s Institute for Creative Technologies are developing a conversational Artificial Intelligence interface that allows two-way conversation between soldiers and autonomous robotic systems. The new interface, which is designed to reduce training costs and improve soldier-robot teamwork is named JUDI (Joint Understanding and Dialogue Interface).
The next generation of military robots will be much more intelligent with a high degree of autonomy, playing an ever-larger role in the military. Also, the problem of control becomes increasingly urgent. These military robots will become less simple remote sensors and tools and more teammates with which soldiers to interact at a much higher level. This level requires more advanced training for the soldiers to work with these machines. JUDI facilitates two-way verbal conversations between robots and soldiers to perform tasks. Conversational AI is the use of speech-based assistants or chatbots to automate communication which creates a personalized customer experience at scale.
Dr Matthew Marge, a research scientist at the laboratory said that the dialogue will be a critical capability for these autonomous systems operating across multiple echelons of multi-domain operations so that soldiers across land, air, sea and information spaces can keep a situational awareness on the battlefield. JUDI allow soldiers to more easily team up with autonomous robots for the safe and effective completion of missions in scenarios like reconnaissance and search-and-rescue.
Conversations enable these systems to supplement their understanding of the world by conversing with human teammates instead of relying on pre-specified and out-dated information about a mission. JUDI allows the soldiers to interact with autonomous systems through bidirectional speech in tactical operations where verbal instructions can be used for commanding and controlling a mobile robot. The technology also provides the robot, the ability to ask for clarification or to provide with status updates as tasks are completed.
A military system cannot rely on a cloud connection or distant labelled databases, which were not secure and not available on the battlefield. JUDI uses a bespoke database to understand a soldier’s intention based on their spoken words. The dialogue processing uses a statistical classification technique that was trained on a small range of human-robot dialogue where humans initially stood in for the robot’s autonomy.
Systems like JUDI not only need to be able to precisely control robots but also need to use situational awareness to evaluate their surroundings and request more information from a human teammate because of high-level needs and incoming data may be incomplete and ambiguous.