Voice-Commanded Autonomy for Robotic Ground Vehicles

As the battlefield becomes increasingly complex – as threats become more difficult to identify and track, as the amount of data at the hands of the warfighter increases, as decision cycles become smaller and quicker – it will be critical for emerging technologies to simplify the operational environment and enable soldiers to act quickly and decisively.   Both ground autonomy and natural language processing are significantly robust technologies that have become more and more commonplace in people’s everyday lives. Ground autonomy in complex environments (such as Overland AI’s work with DARPA RACER) is already functional, and is rapidly improving on an almost day-by-day basis, enabling soldiers to offload basic mobility tasks and focus attention on more complex problems. Natural Language Processing is also commonplace in today’s world. Virtual assistants like Apple’s Siri and LLMs such as ChatGPT show that current technology is readily able to understand complex tasks in conversational language and deliver accurate solutions. Voice commands have already been used in managing and tasking UAS platforms, allowing warfighters to operate these platforms without pulling their attention away from critical tasks.  For this proposal, Overland AI plans to develop true voice-commanded autonomy by first integrating Overland AI’s OverDrive autonomy stack and OverWatch map-based C2 application with Primordial Labs Anur Natural Language Command and Control platform through an open systems API. With that integration established, we will then focus on establishing voice-commanded autonomy by starting with basic autonomous movement of a single platform, before moving to complex autonomous maneuvers with multiple vehicles and finally progressing to turret and payload commands. Throughout, we will focus on end-user feedback and validation through two government facilitated demos during the period of performance.