This paper addresses the challenges of deploying reinforcement learning (RL) models for traffic signal control (TSC) in real-world environments. Real-world training can prevent mismatches between simulation environments and the actual traffic conditions, thereby achieving better performance of agent upon deployment. However, free explorations by agents during real-world training can disrupt traffic operations. To mitigate this, this paper proposes a reference mechanism to guide the decision-making process within the RL framework. A reference timing policy, typically a model-based signal strategy, is integrated into the learning process to provide agents with reference actions. Specifically, an additional Q-value function is introduced to evaluate both the agent's actions and those of the reference policy, allowing for adjustments before the actions are executed in real traffic system. Numerical results indicate that the reference mechanism effectively enhances system performance early in the training process, thus accelerating learning. We also combine the reference RL method with a pretraining procedure and a jump-start algorithm, respectively. Experimental results demonstrate their effectiveness in further enhancing system performance and facilitating real-world training.

@en

When making trips in urban environments, cyclists lose time as they stop and idle at signalized intersections. The main objective of this study was to show how augmenting the situational awareness of traffic signal controllers, using observations from moving sensor platforms, can enable prioritization of cyclists and reduce lost time within the control cycle in an effective way. We investigated the potential of using observations from connected autonomous vehicles (CAVs) as a source of new information, using a revised vehicle-actuated controller. This controller exploits CAV-generated observations of cyclists to optimize the control for cyclists. The results from a simulation study indicated that with a low CAV penetration rate, prioritizing cyclists by tracking reduced cyclist delays and stops, even with a small field of view. As the delay of car directions were not taken into account in this study, the average car delay increased considerably with an increasing number of cyclists. Future work is needed to optimize the control that balances the delays and stops of cyclists and cars.

@en

Abstract Traffic signals are crucial for regulating road safety, but public opinions and cyclist behaviour regarding traffic signals and push buttons have sparked discussions on social media platforms. This study aimed to explore public opinions and cyclist behaviour related to traffic signals and push buttons by analysing social media data and empirical observations and provide recommendations to municipalities on both suitable push buttons, and communication with the public regarding traffic control. To gather social media opinions, social media scrapers were used to collect the data, including posts from social media platforms and news articles published by municipalities. The analysis revealed that a significant majority (77%) of the social media posts expressed negative opinions toward traffic signals. Among the negative opinions, 52% of users believed that push buttons on traffic signals were ineffective. The main reasons cited for criticism were concerns about push button hygiene, long waiting times at traffic lights, and safety issues at intersections. Additionally, approximately 23% of negative posts did not provide specific reasons for their criticism but expressed a negative sentiment. Understanding the reasons behind these negative opinions is essential to address them effectively. Several factors were identified as influencing the emergence of these opinions, including influential articles published by news websites and time-dependent events such as the Covid pandemic. In response to these opinions, municipalities acknowledged citizen criticisms and implemented various strategies to address concerns. Personalized approaches, actively listening to individuals, and assuring them that their criticisms would be addressed, proved to be effective in engaging with the public. Explaining the control function for specific locations also resulted in positive reactions from 65% of respondents. However, when municipalities announced improvements to intersection layout or control without a personal touch, the reactions tended to be more negative: only 50% positive reactions for layout-related announcements and 40% positive reactions for control-related announcements. Negative reactions ranged from scepticism about effectiveness to questioning why similar solutions were not implemented elsewhere. Concurrently, an empirical study was conducted to explore the relationship between different types of push buttons and cyclist behaviour at signalized intersections. The study made use of detector data and signal data provided by the municipality of The Hague, collected from eight intersections with 3-5 bicycle movements. Each signal for these movements was equipped with loop detectors and one of three types of push buttons: bright buttons, touch buttons, or touch buttons combined with a waiting time indicator, see Fig. 1. The bright button is a simple button which must be pushed and has small feedback lights (LEDs). The touch button indicates "wait BT11: Designing Infrastructure for Bicycle Traffic 178 for green" after being touched. The traffic controller of the intersection logs the data, so-called V-Log data, with which the following events can be determined per movement: 1. The start of red and green of the signals. 2. The moment the detector is occupied. 3. The moment the button was pushed or touched after the start of the red signal. 4. The number of times the button was pushed during red. 5. The duration of each button press/touch. The buttons give feedback (the LEDs or the message “wait for green”) if pushed or touched, but also if a cyclist is detected by the loop detector, the button gives this feedback. The cyclist can still use the button after being detected by the loop detector, and the majority of cyclists will do so. For all movements at all intersections, and all buttons, on average in 0.7% of the cases only the button is used to request green, so without the loop detection. Averaged over the cycles, although the initial detection is done by the loop detector, the bright button is still pushed for 62% of the cases, The touch button is used in 85% of the cycles. It is unclear why this is higher since the touch button indicates the detection more clearly than the bright button. If the waiting time indicator is added, only in 44% of the cycles the button is still touched. There is only one movement with waiting time indicator, but it shows that when the remaining green time is clearly indicated, cyclists will use the button less. There was no correlation found between the maximum waiting time and the number of button presses, suggesting that cyclists at these intersections do not express their impatience by pushing or touching the button more frequently. Further analysis showed that the touch button was touched for a longer duration compared to the bright button. In 85% of the cases, the bright button was pushed for less than 1 second, while only 45% touched the touch button for less than 1 second. Additionally, for the touch button (both with and without the waiting time indicator) a remarkable peak at 9 seconds was observed, with 8% of cyclists pressing the touch button for this extended duration. An explanation can be that the bright button gives clear tactile feedback when being pushed, which lacks for the touch button. In conclusion, despite the presence of negative opinions on push buttons, cyclists continue to use them. Negative opinions expressed on social media often relate to concerns regarding hygiene, waiting times, and safety issues. The recommended approach for municipalities is to implement personalized strategies and provide information to address these concerns. The empirical study found no significant correlation between using the button and waiting time. However, clear detection feedback, the tactile feedback of the ‘push’, and a waiting time indicator is advised for enhancing user comfort and trust. @en

Geloofwaardigheid is belangrijk bij verkeerslichtenregelingen. Verkeersdeelnemers die te lang moeten wachten terwijl het onduidelijk is waarom, zijn geneigd tot roodlichtnegatie, wat gevaarlijke situaties met zich kan meebrengen. Als er idee¨en verspreid worden waarin de geloofwaardigheid wordt aan getast, is het van belang een methode te vinden om deze idee¨en met informatie weer recht te zetten. Als use case is de nieuwjaarsconference van 2021 van Peter Pannekoek gebruikt, die twijfel deed ontstaan aan detectie van fietsers, met name de werkzaamheid van drukknoppen bij verkeerslichten. Om dit te onderzoeken is de onderzoeksvraag gesteld: “Wat zijn de meningen van social media gebruikers over drukknoppen van verkeerslichten, en op welke manier kan de gemeente als reactie op deze meningen het beste voorlichting geven over de werking van verkeerslichten en drukkknoppen?” Met behulp van “social media scrapers” zijn opinies over verkeersregelingen en drukknoppen verza meld, en de informatie die gegeven wordt door gemeentes om verkeersdeelnemers te informeren over verkeerslichtregelingen en drukknoppen bij verkeersregelinginstallaties. Uit de resultaten blijkt dat een aanzienlijk deel van de opinies over verkeerslichtenregelingen een negatieve uiting is (77%). Deze negatieve uitingen zijn gerelateerd aan veiligheid, hygi¨ene (met name gedurende de Coronacrisis) en lange wachttijden. Wat betreft de meningen over drukknoppen, in 52% van de posts over dit onderwerp wordt gesteld dat drukknoppen bij verkeerslichten niet werkzaam zijn. De mening van “influencers” in de media over het “niet-werken van drukknoppen” blijkt van invloed te zijn op de meningen die te vinden zijn op social media. Het percentage posts over drukknoppen in het totaal aantal posten over verkeerslichten stijgt na media-aandacht, van maximaal 26% in jaren dat er geen aandacht voor was, tot 47%-62% in jaren waarin de media er aandacht aan wordt gegeven. De meest effectieve benadering om deze negatieve uitingen om te buigen, is een persoonlijke benadering, waar de gemeente laat zien de negatieve uitingen serieus te nemen met de belofte de situatie te onderzoeken, en te verbeteren.@en

Similarly to Maslow’s pyramid of human needs, we theorize that cities have a pyramid of bicycle network needs that depends on their level of bicycle culture. As an increasing number of data sources emerge for bicycle data collection, transport authorities face the challenge of understanding how to use the data and which data sources are fit for their network needs. This article defines a framework that relates the bicycle network needs of cities with data collection systems. We showcase the need-driven framework through a case study of Melbourne, Australia, a bicycle ignorant city, and surveying 15 municipalities (and their consultancies) of the Netherlands. By using the proposed need-driven framework cities can understand how to fully exploit bicycle data collection systems and make a systematic plan.@en

Is a roundabout still a good solution in the era where ‘intelligent’ intersection control using vehicle connection (i-TLC), is in upswing? To answer this question, the capacity of ‘i-roundabouts’, where the infrastructure communicates with the vehicles (I2V), is determined analytically. The roundabouts considered are single-lane roundabouts and turbo roundabouts (a spiral multi-lane roundabout with reduced number of conflicts). A macroscopic approach explores the capacity gain that can be achieved by taking into account the necessary safety margins with regard to headways and gaps. Furthermore, it is assumed that by using I2V the speed, headway, and also the driving curve of the vehicles can be controlled. For roundabouts with speeds lower than 36 km/h, the conclusions are that:•On a single lane roundabout, roughly a doubling of the capacity can be achieved.•On a turbo roundabout the capacity gain can be surprisingly much higher (about a factor 2.5). This is due to the possibility of gap synchronization on the double-lane segments.@en

Various studies have demonstrated that the state and dynamics of an urban network can be described by Macroscopic Fundamental Diagrams (MFD) and that MFDs can be used for perimeter flow control. Perimeter flow control aims at a higher throughput in an urban network by controlling the flow at the boundaries of sub-networks.

For perimeter flow control it is desirable that the MFD has a favourable and consistent shape, independent of fluctuations in traffic demand and of intersection signal variations. From literature it is known that a consistent shape is related to the homogeneity of vehicle accumulation in the sub-network. However, also the signal controller type may influence homogeneity and the MFD shape.

In this paper we investigate the relationship between the type of intersection control and the shape and scatter of the MFD, and the homogeneity of the sub-network, for Vehicle-Actuated (VA) and Back-Pressure (BP) control. The comparison of the two control methods is performed by means of microsimulation.

The results show that for both control methods the free-flow branch of the MFD has a low scatter with an average relative deviation around 2%. The congested branch shows a much larger deviation, 15% for the Vehicle-Actuated control, 16% for the Back-Pressure control. Furthermore, there is a distinct difference in the shape of the MFDs: for VA control the production increases faster as function of the accumulation than for BP control, but the network breakdown starts at a lower accumulation. So, based on the simulation results, VA control is better in undersaturated situations, and BP is better at higher accumulation levels.@en