The snipers were perched around the hockey rink in Florida. As its human handlers exited through a side of the rink, a counter-sniper team robot into position, before unleashing foam orbs on its turreted foes. A succession of teams passed through, each trying to best both the defenders and their competition, armed with an arsenal you could find in a toy store. Even with comical armaments, there are still lessons for the future of war from foam orb live fire exercises.

Held May 22, the Panama City Division of the Naval Surface Warfare Center’s 2019 Director’s Cup was designed to see if autonomous robots could perform a simulated anti-sniper mission. It’s a task that is, in theory, well suited for robots, whose sensors can read the environment differently than human eyes and whose artificial bodies lack the soft fleshy bits that snipers like to target. For the robots to be effective at the mission, they need to complete a set of fixed tasks: navigate the space, find the snipers, place shots on target, and return to where it had launched. In addition, teams were detracted points from their score for hitting objects other than the target and for each shot after the first 10 on a target, with score bonuses available for more quickly completing the challenge.

Launched by the Center in 2015, the competition takes place every other year and requires teams to create and field “a fully autonomous, artificially intelligent, ground-based vehicle to neutralize a dangerous battlespace.” (A secondary goal of the competition is fostering teamwork among the teams tasked with building the robots.)

While the ammunition was nerf and the battlefield abstracted, the exercise was designed to showcase full autonomy, leaving humans to set the machines in motion and then get out of the way as they performed target identification and shooting. Each team used a Clearpath Robotics Jackal uncrewed ground vehicle as the basis for their robot. Prior competitions required the teams to build both hardware and software from scratch for the robot; with the same standard robot body, the 2019 teams were able to focus on the autonomous features of robot fighting.

Besides the basic robot body, the teams outfitted their jackals with a range of sensors, including LMS-111 laser sensors, inertial measurement units, ZED stereo cameras for image identification, NVIDIA Xavier processors, and “an Arbotix-M microcontroller for classification, control and neutralization of a target,” according to Clearpath.

For armament, the robots sported Nerf Rival Nemesis guns, with a 100-round capacity. Or at least, that was the basis for what they were supposed to use. The ammunition remained the same across teams but the actual gun was subject to interpretation.

How those parts were all integrated together varied team to team. One team mounted the stereo camera directly above the barrel of their Nerf gun. Another mounted two Nerf guns on individual turrets in special 3D-printed casings. The winning team chose to largely eschew the Nerf design, building a custom hopper and using brushless motors with velocity control for greater accuracy. Like beetles iterating around a similar body, the range of options for robots on the same chassis was surprisingly varied.

All told, this was a small-scale competition with a goofy weapon in a low-stakes format. It is iteration in this space, as the tools of autonomy become cheap and the means of implementing autonomy become more available, that will make autonomous machines a more durable presence on the battlefields of the future. Targeting systems, navigation algorithms, target distinction and learning how to place the precise number of shots on target? These are the precursors to lethal autonomy, the laboratory experiments and field tests.

The Nerf guns add a goofy, almost cartoony aesthetic to the whole affair, but the gun is the easiest part of a robot to change.