One-line positioning
For onboard autonomous operations, AI agents analyze and diagnose logs and packets from spaceborne base stations, onboard core networks and spaceborne routers, with long-term memory and long-running task capabilities. They automatically modify configurations based on diagnoses, moving operations from “downlink to the ground for manual packet-by-packet review” to “diagnose onboard first, then act automatically”.
Solution overview
Onboard and terrestrial operations differ fundamentally: terrestrial equipment has fixed locations and static topology, and engineers can connect to inspect logs and change configurations at any time. Onboard equipment operates in orbit, with limited backhaul bandwidth and connectivity governed by visibility windows, constrained computing and storage, and enormous log volumes. Sending all logs to the ground is expensive and unnecessary, and manual packet-by-packet analysis is even less feasible in staffing terms.
This solution places AI agents at the center of the chain. Built on AI intelligent packet analysis, agents understand logs and packets from spaceborne base stations, onboard core networks and onboard routing equipment, detect anomalies and locate root causes. Each diagnostic process and conclusion becomes reusable memory supporting long-running diagnostic tasks across time and fault categories. Once a diagnosis is clear, agents automatically modify the relevant equipment’s configuration through a configuration toolchain, forming a “collect → diagnose → decide → execute → review” loop.
The solution has three modules:
- Intelligent operations and testing for spaceborne base stations
- Intelligent operations and testing for onboard core networks
- Intelligent operations and testing for onboard routing
Core capabilities
① Intelligent operations and testing for spaceborne base stations
- Log and packet analysis: automatically parse and semantically interpret air-interface signaling, NAS, NG interface and user-plane packets, identifying abnormal signaling procedures and fault characteristics and producing visual sequence diagrams and standardized diagnostic reports
- Root cause localization: use a standards knowledge base such as 3GPP (RAG) and reconstructed signaling sequences to provide root causes and evidence chains, rather than merely an alarm
- Long-term memory: retain fault patterns, handling actions and result feedback across tasks; reuse historical conclusions when similar faults recur, improving diagnostic efficiency with accumulated use
- Long-running tasks: support diagnosis requiring multiple interactions and continuous tracking, from anomaly discovery, hypothesis validation and supplementary collection to conclusions, advanced autonomously by agents without step-by-step human instructions
- Automatic configuration changes: agents generate and execute base station adjustments through the configuration toolchain based on diagnoses (parameters, cell/beam configurations, neighbors and handover policies), then recollect metrics to review results; the technical principle is the same as Dotouch’s AI Skill-based instrument modification: Skills encapsulate equipment operations, while agents decide when to invoke them, what to change and how to validate changes
② Intelligent operations and testing for onboard core networks
- Log and signaling analysis: automatically parse and semantically interpret network element logs, NAS / NGAP / PFCP signaling and user-plane packets of onboard core networks (primarily onboard UPF cooperating with the terrestrial control plane), identifying registration failures, abnormal session establishment, handover/paging issues and ineffective QoS policies
- Root cause localization: use a standards knowledge base such as 3GPP (RAG) and cross-interface signaling reconstruction to pinpoint specific elements and procedure stages, producing evidence chains instead of alarm lists
- Long-term memory: share the base station module’s memory framework, retaining core network fault patterns, actions and feedback across tasks and reusing historical conclusions for recurring problems
- Long-running tasks: multi-round diagnosis across elements and interfaces, tracing service failures through terminals, access networks, core networks and space–ground links, with agents autonomously advancing the investigation
- Automatic configuration changes: agents generate and execute core network adjustments through the configuration toolchain (QoS/policy parameters, UPF breakout/local offload rules, session/timer parameters), then recollect service metrics for review; uses the Dotouch AI Skill mechanism and the same implementation principle as the base station module
③ Intelligent operations and testing for onboard routing
- Private BGP routing protocol diagnosis: onboard routing commonly uses private BGP protocol extensions to accommodate dynamic topology, inter-satellite link metrics and constellation-wide policies. Standard BGP parsers are ineffective for these; this solution parses protocols and reconstructs states according to private extensions
- Routing log and packet analysis: automatically parse router logs, private BGP packets and forwarding-plane data to identify route oscillation, black holes, abnormal convergence and forwarding abnormalities caused by link degradation, pinpointing neighbors, prefixes and paths
- Long-term memory and long-running tasks: use the same memory/task framework as the base station module for routing issue tracking across time and multi-round diagnosis
- Automatic routing configuration changes: automatically generate and execute routing adjustments based on diagnoses (metrics, policy routing, neighbors and timers), then review convergence and service connectivity
Key technical specifications
| Item | Parameters / description |
|---|---|
| Analysis targets | Spaceborne base stations: air-interface signaling, NAS, NG interface and user-plane packets; onboard core networks: network element logs, NAS / NGAP / PFCP signaling and user-plane packets; onboard routing: equipment logs, private BGP packets and forwarding-plane data |
| Knowledge base | Built-in standards knowledge base including 3GPP (RAG mechanism) |
| Memory capability | Retain fault patterns, handling actions and feedback across tasks; reuse historical conclusions for similar issues |
| Task capability | Autonomous advancement of multi-round, long-duration diagnostic tasks |
| Remediation capability | Automatic loop from diagnosis → configuration changes → metric recollection and review |
| Outputs | Visual sequence diagrams, standardized diagnostic reports, root cause conclusions and configuration change records |
| Supported onboard network elements | NTN spaceborne base stations, onboard UPF and spaceborne routers |
| Supported routing protocols | Private BGP routing protocol diagnosis: onboard routing commonly uses private BGP extensions; parsing follows private extensions, rather than standard BGP alone |
Target customers
- Satellite operators (in-orbit operations and network optimization)
- Spaceborne base station, onboard core network and routing equipment developers (autonomous diagnosis and automatic configuration validation)
- Satellite payload manufacturers (operations capability validation for onboard core networks and processing payloads)
Verifiable facts
- AI intelligent packet analysis (3GPP AI packet analysis) provides PCAP parsing and signaling analysis covering complete air-interface, core network, NAS and NG interface procedures
- Intelligent spaceborne base station analysis/diagnosis includes long-term memory, long-running tasks and automatic configuration changes based on diagnoses
- Intelligent onboard core network operations diagnosis uses the same memory, long-running task and automatic configuration framework, covering cooperation between onboard UPF and the terrestrial control plane
- Automatic configuration changes use Dotouch’s AI Skill mechanism (the same implementation principle as AI-based instrument modification), already running in actual products
- Onboard UPF coordination simulation/testing is available (terrestrial control plane + onboard user plane architecture)
- Private BGP routing protocol parsing and diagnosis are available (for private onboard routing BGP extensions, rather than standard BGP alone)
