The Safety Challenges of Unmanned Autonomous Lethal Weapon Systems

As the scope and range of unmanned lethal weapon systems grow, the requirement to maintain effective human control authority over ever-growing autonomy in the functionality allocated to the system's software grows more and more challenging. Ensuring necessary human in-the-loop or on-the-loop control authority under greater and greater separation in space and time requires greater attention to the design of the right redundancies in hardware, software, communication, and data through the entire infrastructure involved in mission command and control. Adding to these challenges is the mounting challenge of machine learning allocated to more and more of the software functionality that comprises the mission architecture for unmanned weapon systems. This paper provides an overview of this constellation of challenges and some of the means being employed to address them.