VELO-RL - Improved traffic light control and optimization of bicycle traffic through reinforcement learning
Project leaders: Dipl-Ing. Sebastian Pape, Dr. rer nat. Tobias Nousch
Project duration: 10/2025 - 12/2027
Short description: With the help of a traffic control system based on reinforcement learning (RL), bicycle traffic is to be accelerated while at the same time taking into account the needs of other modes of transport.
In many German cities, various political and transport planning measures are being implemented, but cycling is often insufficiently considered, as the focus is mainly on the operational requirements of motorized private transport and local public transport. A central problem is that cycling is not taken into account in most traffic-dependent traffic light (LSA) control systems and the coordination is primarily geared towards motorized private transport. This leads to delays, frequent stops and the feeling of being at a disadvantage when it comes to traffic control. Although there are already green waves for cycling, these are often unable to adapt dynamically to the current volume of cycling traffic or its varying progressive speeds. The volume of cycling traffic can change greatly depending on the time of day, time of year and weather conditions, which means that adjustments to the control process are always necessary. In addition, there is often a lack of suitable systems for recording the traffic volume of cyclists as well as control algorithms that can appropriately weigh up the interests of cycling and reconcile them with those of motorized and public transport.
The aim of the project proposed here is therefore to develop a control system based on reinforcement learning (RL) that accelerates the cycling traffic and at the same time takes into account the needs of other modes of transport. To this end, a fictitious traffic network with several signalized junctions will be created in the open-source simulation environment SUMO. In addition to motorized private transport, local public transport and vulnerable road users (VRU) are also modeled. Three interconnected junctions are optimized using RL-based control, with a focus on the coordination of bicycle traffic. At the same time, the effects of this control on the surrounding junctions are systematically investigated. With the help of RL-based control, the travel times of bicycle traffic within the traffic network are to be optimized in comparison to the existing standard controls (fixed time control, partial traffic-dependent control) as well as on route sections with and without coordination. The aim is to minimize the negative impact on other road users. In addition to travel time optimization, a situation-dependent activation of a green wave for bicycle traffic (depending on weather, bicycle traffic volume, environmental factors) is also to be tested and its effects evaluated. For the detection of road users, detection systems that are already available on the market are to be implemented in the simulation in order to ensure compatibility with existing traffic management systems and thus make implementation in the field economically viable.
Funding: The research project FE 88.0240/2025/AE03 is funded by the Federal Highway Research Institute on behalf of the Federal Ministry of Transport as part of the 3rd funding call of the "Road Innovation Program" funding guideline with the focus "Green light for cycling - Acceleration of cycling on routes or at junctions". Responsibility for the content lies solely with the author.