Sentinel

Sentinel is a teleoperation software for robots designed to help robotics companies scale their Physical AI data and deployments. It offers a unified platform for reliable control, data capture, and deployment, enabling operators to control robots from anywhere in the world. The core value lies in its ultra-low latency teleoperation, which ensures smooth and precise motion even over long distances. Aimed at teams working on AI training, research, and real-world robotic applications, Sentinel reduces the barriers of geographic distance and high-latency connections, making remote robot operation as effective as local control. By providing immersive views and haptic feedback, it bridges the gap between operator and machine, whether for data collection or supervising deployed robots.

The primary pain point Sentinel addresses is the difficulty of achieving reliable, low-latency remote control for robots, particularly when collecting high-quality data for Physical AI training. Traditional teleoperation suffers from lag and instability, leading to jerky movements and poor data quality. With latency as low as 10 milliseconds, Sentinel enables smooth teleoperation that feels responsive and natural. This matters because high-quality teleoperation data is essential for training robust AI models. Without it, robots cannot learn effective manipulation or navigation skills. Sentinel's low-latency infrastructure ensures that operators can perform complex tasks accurately, such as grasping objects or navigating cluttered environments, even from thousands of miles away.

One of Sentinel's key feature groups is real-time control combined with an immersive view. Real-time control delivers smooth motion with latency as low as 10 milliseconds, making remote operation feel immediate and precise. This is achieved through optimized data transmission protocols and edge processing. The immersive view feature provides true depth perception from a distance, using stereoscopic camera feeds and advanced rendering. Operators can perceive the robot's surroundings in three dimensions, which is critical for tasks requiring spatial awareness like picking and placing objects. Together, these features drastically improve task success rates and operator confidence.

Another major capability is haptic feedback and multi-feed streaming. Haptic feedback allows operators to feel the robot's interactions with the environment, such as contact forces and texture, through real-time tactile signals. This enhances situational awareness and enables delicate manipulation. Multi-feed streaming supports up to six full high-definition video streams simultaneously, providing comprehensive visual coverage from multiple camera angles. The system maintains full operation without performance degradation, even when all streams are active. These features are particularly useful for complex operations like assembly, inspection, or surgery where multiple views and force feedback are essential.