US Army Combat Capabilities Development Command, which has been working on robotics research for the past ten years, has just recently begun integrating this technology into the battlefield. They have been testing their robots, and using them to scan for IEDs and recover Ravens. The US Marine Corps is also testing a recon robot for urban warfare. These robots work together with Soldiers. They can carry out tactical actions as planned but also make suggestions for changes.
The US Army is creating an algorithm-based learning platform that will allow Soldier robots to communicate with one another. This system will enable robots to hear the Soldier's voice and take the appropriate actions. It will allow robots to ask questions in order to overcome uncertainty. This will allow robots to complete joint tasks at an operational speed and will reduce the learning curve of Soldiers.
The most difficult part of creating an algorithm-based learning system was figuring out what questions to ask. The research team developed a natural language understanding pipeline that processes the spoken language of a Soldier, and based the system on statistical classification techniques. The system was trained from a small amount of human-robot interaction. This allowed the researchers to manage the natural language variability incrementally. The research team is currently tying the output semantic representations of Soldier's spoken languages to the grounding system for the robot.
Algorithm-based learning systems are designed to improve Soldier-robot cooperation in tactical environments. This ability will enable robots to communicate naturally with Soldiers. The research team developed a natural language processing pipeline that interprets the intent of a Soldier's spoken language. The system will eventually be able to tame the variation in spoken language. This will give the robots the ability to act on the intent of language, which will enable them to become more collaborative.
The algorithm-based learning will also enable the robots ask Soldiers specific questions to gain an understanding of their environment. The questions will use information from maps. These maps will help drive autonomous robot-based decisions. The questions are also tactically relevant. These questions examine the process by which robots make decisions, and the information the Soldiers need. Soldiers also provided feedback. The Soldiers’ feedback was used by the robots to determine if the maps were relevant to tactical and operational decisions.
The Soldier-bot critical variant has a faster attack rate, and fires critical rockets faster than other Soldier-bot variants. These rockets are six times more destructive than regular rockets. The critical variant has a higher health level and can take out players without any resistance.