-
StatusOngoing
-
Status date2009-02-16
ATSIG aims to develop an automated diagnosis system on top of typical network management / OSS (operations support systems) for satcom network operators. The idea is that the OSS typically provides a lot of valuable information to the user (operator staff), but this still needs to be processed by a skilled technician to make any value. Today a significant amount of troubleshooting work is manual, and the ATSIG project aimed to automate this.
By capturing expert knowledge in a model known as a "Bayesian network", we gather all available information and feed it into a Bayesian network. Then we produce a much more "final" output to the user: A list of possible faults ranked with an accurate probability score – a diagnosis.
The ATSIG system will save operators a lot of time and resources, and also free the "hardcore guys" in the organization for more pro-active tasks like building new network infrastructure.
True innovation is when you bring technologies together to form new products. The ATSIG project combines two competencies from computer science and engineering (Bayesian reasoning and SatCom). It is always difficult to get enough domain knowledge to the software specialist for building a new product, but the effort from the project team and Avanti Communications Group PLC made this happen, leading to the very promising results now documented in the final report.
- Saving time and resources
The automation provided by ATSIG has been estimated to EUR 100,000 – 150,000 per year for a satellite-based broadband provider having 5000 customer terminals. These savings are due to lower demand for human troubleshooting engineers and also lower reuirements on their skills. - Decreased time to solve problems
The automated diagnosis process is much faster than human reasoning, where the troubleshooting engineer will need to check many performance indicator readings to solve issues. - Better customer satisfaction
Decreased time to solve problems leads to better network quality in general, and from this the operator will get better customer satisfaction. - Less sensitivity to human knowledge
Since ATSIG stores a model about associations between faults and their “symptoms” (performance indicators), this knowledge is not lost when staff leave the organization or go on vacations. - Knowledge is captured and stored
The ATSIG system makes knowledge much more explicit by capturing it in a model associating faults and performance indicators. This is a great advantage because the knowledge about the network is often just tacit (implicit) which makes it difficult to communicate and use for training new staff.
The ATSIG toolset provides a graphical user interface (GUI) enabling the user (troubleshooting engineer) to get a quick diagnosis for a problem that a customer calls to complain about. The developed product can be integrated with existing OSS products, such that the user can identify a customer in the OSS, and immediately get both raw data about the customer link as well as a diagnosis.
The resulting diagnosis is a ranked list of faults according to probability. This means that you get information about certainty of the diagnosis as well as alternatives in case the first option turns out not to be the true fault. Many alternative techniques for automated diagnostics (e.g. rule-based systems or decision trees) will only be able to provide a single "most likely fault" in a deterministic manner. By using Bayesian reasoning, ATSIG provides a much more accurate output in the form of the ranked fault/root cause list. This is much closer to the information given by human troubleshooters when they hand over a case to a colleague: "I strongly believe that the problem is X because of ..., but due to ... it could also be Y or even Z".
Since the project is aiming to bring expertise from two different domains together (Bayesian reasoning and SatCom), the ATSIG project was planned in two phases (1 and 2) where the first phase aimed to identify if the proposed technology seemed feasible or not by implementing a prototype version of the ATSIG functionality. The second phase was then implementation and validation. Phase 1 demonstrated the potentially nicely, and Phase 2 produced some great results and the project outcome is now to be commercialized in a new company, 2operate (www.2operate.com).
The ATSIG toolset has now been developed and validated. The results were so impressive that a new company, 2operate ApS, has been established to commercialize the product. More information can be found at www.2operate.com.