Artificial Intelligence and Machine Learning in Field Service

Using Machine Learning Methods to Optimise Field Service Visits and Avoid Unnecessary Driving

If you ever lose your internet connection and restarting the router does not help, it might be necessary for a technician from TDC NET to pay you a visit. Field service technicians at TDC NET drive approximately 80,000 kilometres every day, the equivalent of circling the Earth twice, whereof some visits turn out being unnecessary and could be avoided. This PhD project is part of a large grant solution initiative called Greenforce funded by Innovation foundation Denmark. The main goal of Greenforce is to decrease kilometres driven by technicians, thus decreasing cost and CO2 emissions. The project will improve the efficiency of field service by developing a software platform in order to optimize demand management, increase the number of problems that can be solved remotely, and improve daily routing of the TDN NET technician fleet. The research topics will lie in the intersection between Artificial Intelligence (AI), Machine Learning (ML) and Operations Research in order to achieve the ambitious goal.

One of the central challenges of this project is to deal with the complex infrastructure of TDC telecommunication network. Being the biggest supplier of digital infrastructure in Denmark, TDC NET oversees a big and complicated network with many connections and cables running both above and below the ground. Being able to identify the root cause among many potential types of errors and having service technicians with adequate skill sets to address these, are central to maintaining a stable connection for all users on the network. Different problems do not require the same skills and will not take the same amount of time, complicating the scheduling procedure.

Multiple approaches will be assessed towards their ability to solve the different problems that arise. These include using ML to accurately predict the time it takes for a given technician to perform a given task; using AI to analyse data in real time in order to find anomalies, thus predicting problems before they happen; and trying to find the root-cause for multiple problems arising simultaneously across the network. Finding good and reliable solutions to these objectives will help simplify the scheduling procedure and bring down the number of unnecessary service visits, hence decreasing cost and CO2 emissions. In the end, you might end up with a more stable connection, which means that maybe you will not need to call TDC in the first place.