Direct Drone-to-Satellite Connection Enables Real-Time Video Transmission Through Token Encoding

China Telecom’s Artificial Intelligence Research Institute, TeleAI, has developed the “AI Flow” technology framework, introducing a new method of communication between edge devices such as drones and robots and satellites. According to reports from Quantum Times, this technology enables lightweight drones to directly connect to satellites and stably transmit high-definition video in real-time. Furthermore, this framework has won the Empowerment Award under the Outstanding Artificial Intelligence Leader Award (SAIL) at the World Artificial Intelligence Conference (WAIC) 2026.

Mechanism of Token-based Video Encoding

Traditionally, video transmission from drones required the setup of temporary communication equipment on-site. Drones would send video to ground receivers, which would then relay information to command centers via satellite. This approach faced two physical limitations: instability in connection when drones moved too far from ground communication equipment, and the limited bandwidth of satellite links due to payload and power restrictions in lightweight drones, which made real-time transmission of high-definition video challenging.

TeleAI’s approach fundamentally changes the nature of transmitted information. At the sending terminal, the video is first analyzed to extract meaningful elements such as scenes, people, object structures, and movement states, which are then encoded into token sequences. These tokens are transmitted via satellite links, and at the receiving end, generative models are used to reconstruct the video. By transmitting token streams instead of traditional bitstreams, high-quality video can be sent even with limited bandwidth.

The concept of tokens has primarily been discussed in the context of large language models (LLMs) until now. TeleAI has extended this token concept beyond chatbots to facilitate communication between drones, satellites, robots, and underwater devices.

Practical Applications in Disaster Sites

The generative intelligent transmission capabilities of AI Flow are already being utilized in real-world emergency scenarios. TeleAI has partnered with China Telecom Emergency Services to deploy over 1,000 devices equipped with AI Flow technology across flood response operations in 31 provinces. Transitioning from demonstration stages to interprovincial operations signifies that this technology is now undergoing real-world validation.

In disasters like floods or earthquakes, simultaneous occurrences of roadblocks, power outages, and communication disruptions are common. Previously, temporary reception and transmission equipment had to be set up in affected areas. However, with AI Flow technology, lightweight drones can directly connect to satellites, capture disaster footage, encode it into tokens, and transmit it to command centers. This enables commanders to quickly assess road conditions, structural risks, and optimal entry routes for rescue teams.

Extension to Remote Robot Operations

The drone-to-satellite link technology is now being extended to ground-based robots. For robots equipped with physical intelligence to perform tasks in real-world environments, efficient communication and coordination between humans and machines, as well as among machines themselves, are crucial. Operators need to view video from the robot’s perspective, while robots must receive real-time control commands. In scenarios where multiple robots collaborate, sharing information about the environment and task progress is also essential.

By combining AI Flow with the low-latency capabilities of 5G, end-to-end delays can be reduced to 20–50 milliseconds. This expansion enables remote control of robots not only within local areas but also across cities and nationwide. Even in remote areas without base stations, robots can transmit video footage via satellite links and receive operational instructions remotely.

Moreover, AI Flow employs a unified token stream encoding system, which allows various types and manufacturers of robots to connect seamlessly. For example, if one robot detects a roadblock, it can share the information with others. Similarly, if another robot identifies a traversable route, it can share its findings with the system. This facilitates the formation of a distributed intelligence network, reducing redundant recognition and computation across multiple devices.

Application in Underwater Communication

Underwater communication has long been a challenge in the industry due to the absorption and scattering of signals by water, which causes significant attenuation. Real-time transmission of high-definition video over long distances via wireless communication in natural water bodies has been particularly difficult. TeleAI has utilized AI Flow to achieve real-time transmission of high-definition video over long distances in underwater environments. The core payload equipment that facilitates this process is integrated into a self-developed “Air-Sea Cross Submersible.”

This device can transition from air to underwater environments, performing tasks such as seabed cable inspections, underwater exploration, marine farming, and emergency searches while transmitting live footage from the site. By extending the same communication logic across air, satellite, ground, and underwater environments, TeleAI has created a unified solution.

Deployment of Decoding Models on Orbital

Satellites TeleAI is also working on deploying AI Flow’s decoding models directly on orbital satellites. In traditional space-to-ground links, uplink bandwidth is often limited, while downlink bandwidth is relatively abundant. If the decoding model can perform part of the computation and reconstruction directly on the satellite, it reduces dependency on ground-based computing nodes, allowing the satellite to send results to the receivers via the downlink.

This approach means that satellites themselves would function as edge computing nodes. While challenges remain in efficiently operating models within the constrained resources of satellites, successful implementation could significantly enhance overall communication efficiency.

Editorial Opinion

TeleAI’s achievements demonstrate a new stage in the fusion of AI and communication technologies. In the short term, this advancement is expected to enable faster and more flexible video transmission for emergency communications and remote monitoring applications, especially in disaster response scenarios where only lightweight drones may be deployable instead of large-scale ground-based equipment.

In the long term, the unified communication protocol enabled by token streams could improve interoperability among different devices. The ability for drones, robots, and underwater equipment to connect via a single communication standard may accelerate the transition from isolated operations to distributed, collaborative systems. However, it remains to be seen whether video reconstruction quality and latency from generative models are robust enough for critical applications requiring high levels of safety and reliability.

The editorial team is keenly observing whether this technology can serve as a practical solution in bandwidth-constrained environments, especially under adverse conditions. Large-scale field trials will be key to its validation.

References

• “妙啊！无人机直连卫星传Token”, by 十三 — 量子位, 2026-07-18T08:17:18.000Z (ARR)
• Source URL: https://www.qbitai.com/2026/07/453789.html