AI support for dispatchers

The KARL project brings together research on AI in various application areas in the Karlsruhe region and unites 17 project partners under the leadership of the Karlsruhe University of Applied Sciences. They all gather insights into the design of tomorrow’s working world and test models for explicit AI use cases in education, manufacturing and knowledge-intensive services and ICT systems. The project partners include renowned scientific institutions such as the Research Centre for Information Technology (FZI), the Karlsruhe Institute of Technology (KIT), the Fraunhofer Institute of Optronics, System Technologies and Image Exploitation (Fraunhofer IOSB) and the Fraunhofer Institute for Systems and Innovation Research (Fraunhofer ISI) as well as partners from various industries.

In one of the two mobility use cases, INIT is investigating how AI can facilitate the work of dispatchers in an operations control centre. Work in the control centre is often characterised by stress and time pressures. As soon as a disruption occurs, several decisions with sometimes far-reaching consequences have to be made as quickly as possible. Various processes, from the police accident report, to driver communication, to the establishment of a functional replacement operation, to passenger information, all have to be initiated. Many of these measures are carried out on the basis of many years’ experience.

Together with the Karlsruhe University of Applied Sciences, the FZI and lavrio.solutions, an AI applications and data science company, INIT is investigating ways to support new employees in the operations control centre during these stressful situations using AI. Therefore, in the research project, data of previous dispatching measures is being evaluated with the help of AI methods and, based on this, a dispatching assistant is being developed that supports dispatchers when making difficult decisions.

The AI is trained in such a way that a large number of relevant factors are taken into account and measures are then suggested that are precisely adapted to the particular situation. The dispatching assistant will be linked to the ITCS so that dispatching measures are automatically triggered when dispatchers accept the AI suggestion. This is how dispatchers’work is made easier, but at the same time they are still in full control of the operations. The dispatching assistant is particularly useful to quickly deploy new control centre staff or to support dispatchers who have limited experience (e.g. replacement dispatchers). In this way, the dispatching assistant helps to optimise operational processes.

AI uses historical data from the ITCS which contains vehicle positions, timetable data and the respective dispatching measures that have been triggered. In order to ensure that the dispatchers’ requirements are fully considered, surveys were conducted at the beginning of the project among the control centre staff of Verkehrsbetriebe Karlsruhe GmbH and Albtal-Verkehrs-Gesellschaft GmbH.

The biggest challenge is to build up a consistent and comprehensive database that can be used to train the AI in a meaningful way. This is because information is often available in different systems and reaches the control centre via different channels, such as radio or telephone, and partly also from external parties, e.g. the police or fire services. This is how dispatchers receive information about accidents and, for example, closed sections of roads. Of course, this information is documented, but due to extreme time constraints, it may be poorly structured and possibly mixed with personal data, which makes automated evaluation difficult or in some cases impossible. In addition, non-digitised channels such as radio or telephone cannot be used for AI. These are some of the issues that pose challenges to training AI from historical data and are currently being addressed with the help of extensive data preparation and other data sources.

The project partners have already presented the results achieved so far to participants of the INIT ITCS Working Group to obtain further input from experienced transport companies. This feedback will help to further specify the requirements for the dispatching assistant and identify further useful sources of information. The aim is to provide dispatchers with the best possible support without distracting them from other important activities, such as communicating with a driver who has been involved in an accident. The focus is on practical suitability in the stressful daily routine of an operations control centre.

The digital dispatching assistant for operations control centres supports dispatchers in their work and relieves them, in particular, in the event of any faults. This enables them to make sound decisions, which helps to avoid errors and ultimately leads to an improvement in operational processes.