Neubility's Case of Expanding Remote Monitoring Based on Amazon Kinesis Video Streams

Neubility has operated delivery and patrol robot services since its founding in 2017, and in 2024 commercialized a robot delivery service with Yogiyo. In the patrol sector, it has partnered with SK Shieldus and Seongnam City, and as of 2026, it has operated 300+ robots across approximately 140 sites domestically and internationally.

As the scale of operations grew, NCC (Neubility Control Center) needed to stably check robot video and status, and this needed to be expanded beyond the internal operating environment to customer and external user environments. The existing RTSP streaming and port forwarding methods depended on specific networks, and as the number of robots increased, the burden of individual configuration and operation grew larger.

Neubility used the WebRTC feature and Signaling Channel of Amazon Kinesis Video Streams to establish P2P connections between robots and clients. Connections could be made without directly building a separate signaling server, and the examples and APIs of the Amazon Kinesis Video Streams WebRTC SDK were helpful during implementation. Additionally, metrics such as Peer Connection status, roundTripTime, and fractionLost were collected to check quality and network anomalies.

In variable LTE environments, a method was applied that used TWCC to adjust bitrate, and also changed the GStreamer encoder bitrate according to network conditions.

• When adding new clients, connections could be made without port forwarding, reducing the initial setup burden that used to take a minimum of 20 minutes to several hours.
• Repetitive tasks such as network checks, port mapping, and connection testing were reduced.
• A structure was secured that could expand from internally-focused monitoring to customers and external users.

This transition was not so much about the video latency itself, but about establishing a foundation that allows connection and operation without separate configuration even in diverse network environments. AWS credits also helped with the adoption decision during the initial testing and validation phase.