Praxis

PRAXIS addresses the escalating complexity of deploying unmanned systems (UxS) at scale across defense and commercial sectors. Current robotic systems require multiple operators managing single platforms. PRAXIS changes this paradigm - enabling one operator to manage multiple autonomous systems across air, ground, surface, and sub-surface domains, even in GPS-denied and electromagnetically degraded environments.

The value and impact of autonomous systems are limited by their ability to communicate and collaborate. Communication hurdles in contested environments, high operator-to-platform ratios, and the inability to scale robotic operations across multiple domains create significant operational constraints.

PRAXIS deploys intelligent, autonomous, multi-domain heterogeneous robotic systems that display collective agency and decision-making through an AI agent. This is not simply deterministic commands with an agentic wrapper. PRAXIS is a set of modular behaviors that leverage robust training to navigate unfamiliar and unknown situations, incorporating non-deterministic behaviors into your robotics.

The operator does not define routes or micro-manage platforms. Instead, platforms collect data, and the AI agent coordinates collective action for user-defined outcomes. Using techniques from generative and embodied AI, the user experience is outcome-focused - as intuitive as issuing a command like "find red ball."

• One-to-Many Multi-Domain Control: Manage air, ground, surface, and sub-surface robotic systems from a single operator interface
• Communications Maintenance Architecture: RF MANET or mesh, WiFi, Bluetooth with no reliance on a ground control station and fully decentralized nodes
• RF Mapping: Intelligent frequency hopping facilitated by AI agents to maintain connectivity in degraded environments
• Edge Device Reinforcement Learning: Decentralized neural network enabling self-healing and mission-healing capabilities
• Collaborative Robotic Behaviors: Access a growing library of behaviors with dynamic replanning and adaptive autonomy when platforms are lost or destroyed
• Target Identification: Vision-model agnostic TOI identification from multiple cameras looking for similar or various targets
• Mission Planning: Preplan and cache multiple missions to address potential future courses of action and contingencies
• Triggered Missions: Humans can participate in or trigger missions as part of the collective group of collaborative robotic systems

PRAXIS meets customers where they are today with their current inventory of robotic systems. All components are containerized and built on common languages and standards, supporting the accelerated adoption of attritable platforms across multiple domains. The solution decreases costs to train operators, hire larger workforces, and acquire new platforms.

PRAXIS constantly learns through the accumulation of robot operations data, becoming more useful and valuable to customers over time. Edge-based reinforcement learning enables the system to adapt and improve without centralized dependencies.

Deployed as containerized components on AWS infrastructure, PRAXIS leverages cloud scalability for training, simulation, and mission planning while maintaining edge autonomy for field operations in disconnected environments.