When Soldiers enter a new environment, their mission demands they stay one step ahead of the enemy; however, they may find it challenging to maintain a high level of alertness if they’re driving a combat vehicle across unfamiliar or dangerous terrain.
The U.S. Army Combat Capabilities Development Command’s Army Research Laboratory designated several research programs as essential for future Soldier capabilities. Of these major flagship programs, the Artificial Intelligence for Maneuver and Mobility, or AIMM, Essential Research Program, endeavors to reduce Soldier distractions on the battlefield through the integration of autonomous systems in Army vehicles.
Dr. John Fossaceca, AIMM program manager, said he seeks to develop the foundational capabilities that will enable autonomy in the next generation of combat vehicles. This include the construction of a robotic combat vehicle that operates independently of the main combat vehicle.
“The main purpose of this essential research program is to build autonomous systems that help the Army effectively execute Multi-Domain Operations,” Fossaceca said. “We don’t want Soldiers to be operating these remote-controlled vehicles with their heads down, constantly paying attention to the vehicle in order to control it. We want these systems to be fully autonomous so that these Soldiers can do their jobs and these autonomous systems can work as teammates and perform effectively in the battlefield.”
Considerations that go into the development of military autonomous vehicles differ significantly compared to those intended for commercial use.
Manufacturers typically design commercial self-driving cars to operate on pristine roads, where heavy traffic and crowds of pedestrians play a significant factor in its capabilities. In contrast, the Army often faces environments with diverse terrains and areas that may not even have a road to travel on.
“Soldiers may have to operate in forests or deserts, and they may have to operate in a certain manner like moving stealthily in order to achieve some objective,” Fossaceca said. “This is very different than what’s happening in the autonomous vehicle industry, which is the main model of autonomy available today in terms of autonomous vehicle research.”
As a result, this research intends to fill the gaps to cover these unique situations that Soldiers face.
Autonomy in the next generation combat vehicle will not only help Soldiers make decisions but also clearly explain the rationale behind its suggestions so that a Soldier can either approve the choice or intervene if necessary, Fossaceca said.
“Future military missions are going to require autonomous vehicles that can determine what the passable routes might be, calculate the best route and make assessment about what’s happening in the environment,” Fossaceca said. “We want to integrate this autonomous behavior in modern military vehicles so that it feels less alien to the Soldier and can decrease the cognitive burden of the decision-making process.”
In the short term, Army researchers plan to take advantage of recent advances in computing architecture and create a single platform that will first perform narrow AI, or artificial intelligence that can complete very specific tasks consistently, and leverage these capabilities to build a teammate for the Soldier.
“We want to perform very specific tasks at first but make them very robust and make them operate really well,” Fossaceca said. “And then, as time goes on, we want to make things more general so that systems can do a lot more than just very narrow, specific tasks on the battlefield.”
One of the program’s near-term objectives pertains to the development of intelligent mobility with narrow AI in environments with minimum a priori training data. While self-driving cars normally need tremendous amounts of data to train the vehicle’s intelligent system, Army researchers have developed techniques to reduce the training time and data required.
Other examples include Natural Language capabilities to enhance communication between Soldiers and their vehicles as well as what Fossaceca referred to as “tactical teaming with distributed assets,” where the intelligent system changes and adapts as the mission progresses.
In the long term, Army researchers want future combat vehicles to be able to look at the context of the environment and obtain helpful clues. This capability would require the intelligent system to analyze complex, adversarial environments and develop possible courses of action.
“For example, let’s say the vehicle’s intelligent system notices a roadblock,” Fossaceca said. “An obstacle like a fallen tree could just be there by coincidence and require additional navigation in circumventing it, while sandbags may indicate signs of a trap and require the system to warn the driver about the possibilities of an ambush. We want our systems to process these kinds of contextual clues in the future.”
So far, researchers in the program have worked to improve the autonomy software stack, a special collection of software algorithms, libraries and software components for autonomous vehicles.
Originally the product of the laboratory’s 10-year Robotics Collaborative Technology Alliance, the software stack consists of computer programs that allow autonomous vehicles to perform functions such as navigation, planning, perception, control and reasoning. The autonomy software stack also contains a world model that the intelligent system can use as a virtual brain to keep track of information about the world.
“With the autonomy software stack, we can feed capabilities into the [CCDC] Ground Vehicle System Center’s software enterprise, so that they can incorporate those capabilities into their next generation combat vehicle program in the future,” Fossaceca said. “This is foundational research, and it represents the program’s core architecture.”
Recently, Army researchers achieved another important milestone with the Scalable Adaptive and Resilient Autonomy program, which leverages external collaborators outside of the laboratory to accelerate the pace of emerging research in scalable, heterogeneous behaviors.
“We have several different types of robots and systems working together, and we’re asking external performers to come in and help us with increasing the complexity of scenarios that we can handle,” Fossaceca said. “This is really going to help us toward our goal of autonomous maneuver in multi-domain operations, so this is pretty exciting.”