Scout
This company provides a foundation model, FURY, specifically designed for multi-domain defense robotics operations. The model enables robot-to-robot communication and command execution via natural language across ground, maritime, drone, aircraft, and space platforms. It operates in real-time, even in GPS and communications-denied environments, using COTS hardware for deployment flexibility.
Funding
Funding not disclosed
Founders
Product
Problem
Defense robotics often operate in environments with intermittent or absent communication and GPS signals, hindering coordinated operations and real-time command execution. This reliance on external connectivity limits the autonomy and effectiveness of robotic systems in contested operational theaters.
Solution
Scout provides a foundation model for defense robotics, enabling robust robot-to-robot communication and intuitive command inputs through natural language processing. The AI operates in real-time within comms and GPS-denied environments, facilitating generalized decision-making across diverse robotic platforms and mission profiles. This model is deployable across various robotic form factors and operational domains, adapting to different compute requirements and learning from real-world data. It supports swarm coordination and allows commanders to issue commands via voice, text, or map-based inputs without requiring specialized operator training.
Target Audience
The primary customers are defense organizations and prime contractors developing and deploying autonomous robotic systems for tactical and strategic operations.
Features
- Foundation model for multi-domain defense robotics, enabling robot-to-robot communication via a natural language protocol for swarm coordination.
- Natural language command interface supporting voice, text, or map-based inputs, with robot responses and intent understanding.
- Real-time operation in comms and GPS-denied environments, eliminating reliance on external connectivity or cloud infrastructure.
- Pre-trained, generalized decision-making capabilities applicable across diverse platforms and missions.
- Reinforced safety and alignment protocols for complex, dynamic operations.
- Deployable across all robotic form factors and domains, supporting varying compute and power needs.
- Learning capabilities derived from real-world imitation data or simulated environments.
- Optimized for low-power inference and lightweight sensor configurations, utilizing COTS hardware.
- Operates with minimal sensor input, such as a single camera stream (RGB or thermal), reducing dependence on active sensors like LIDAR or radar.
- Supports adaptive learning and secure over-the-air (OTA) model deployments for edge retraining.