Effects of Tram Warning Application on Automated Shuttle Bus and Tram Interaction

181 BASt / A 53 Effects of Tram Warning Application on Automated Shuttle Bus and Tram Interaction by Name: M.Sc. Nino Chkhartishvili Institute: VTT Technical Research Centre of Finland Ltd. Address: Tekniikantie 21, Espoo, Finland Phone: +358 503 378 272 E-mail Effects of Tram Warning Application on Automated Shuttle Bus and Tram Interaction by Name: M.Sc. Nino Chkhartishvili Institute: VTT Technical Research Centre of Finland Ltd. Address: Tekniikantie 21, Espoo, Finland Phone: +358 503 378 272 E-mail 182 BASt / A 53 3 YRS 2025 / Effects of Tram Warning Application on Automated Shuttle Bus and Tram Interaction 1 Introduction Urban transportation systems face difficulties due to the increase in urban population, as urbanisation is accompanied by higher levels of travel times, traffic congestion, delays and accidents (Goetz, 2019). These issues will be more prevalent in the future, since United Nations (2019) estimate that by 2050, 68% of the world’s population will live in urban areas. Additional statistics suggest that urban passenger transport is responsible for 40% of greenhouse gas emissions (GHG) in the total passenger transport sector related GHG emissions (International Transport Forum (ITF), 2021). The alarming issue of urban transportation is congestion; it is directly related to motorisation and the expanded usage of automobiles (Rodrigue, 2020). Congestion is followed by an increased need for transportation infrastructure and road space taken away by vehicle parking (Rodrigue, 2020). Automated shuttle buses could provide a solution to these issues related to individual transportation and improve effectiveness of public transportation. This paper presents a case study on an automated shuttle bus that shares a lane with a tram for part of its route. The research question is whether shuttle bus operation affects the tram’s travel time. Specifically, the influence of a traffic management intervention, such as a tram warning, on the travel time of both the automated shuttle bus and tram is investigated. The tram warning application is a new service that has been mainly tested on conventional vehicles. According to Zimmermann et al. (2021) tram warning application operates for two purposes: i. Application informs car drivers about an elevated risk of collisions with public transportation, ii. Application also notifies drivers about the public transport existence at the stop. The recent study by Tarkiainen et al. (2024) tested the tram warning application in a real urban setting, in an automated shuttle bus testbed in Hervanta suburban area in Tampere, Finland. In the testbed, the automated shuttle bus shared a lane with a tram. Before activating the tram warning app, a slow-moving shuttle impeded the movement of a tram when operating ahead of it on a shared route. The shuttle increased the tram’s average travel time from 24 seconds (when the shuttle was not travelling in front of a tram) to 39 seconds (when it started operating). However, once the tram warning was activated, the instances of shuttle obstructing a tram decreased from every 10 days to every 25 days. Automated shuttle buses are intended to operate at higher automation levels, usually SAE Levels 4 or 5, where user’s assistance is not needed (Chaalal et al., 2023). They have the potential to enhance the attractiveness of public transportation for passengers and could function as connectors for public transport systems (Cao & (Avi) Ceder, 2019; Shen et al., 2018). Thus, evolving automated shuttle buses could improve public transport accessibility (Barillère-Scholz et al., 2020) and ensure uninterrupted last-mile connection (Huber et al., 2022). Additionally, automated shuttles can observe their environment since they are equipped with sensors and advanced communication logic (Chaalal et al., 2023). However, automated shuttle buses are still developing, often require human control and operate at a relatively low speed (Iclodean et al., 2020). 183 BASt / A 53 4 YRS 2025 / Effects of Tram Warning Application on Automated Shuttle Bus and Tram Interaction In general, automated shuttle buses have limited passenger capacity and can transport up to 15 passengers (Fournier et al., 2023). However, a compact size makes them suitable for different services. Shuttles may operate on fixed routes or provide on-demand service, but they are most effective for scheduled services since they often function as feeders for the existing transport network (Soteropoulos et al., 2023). Automated shuttle bus operation has been tested across Europe and North America in mixed-traffic conditions. A pilot project from the Swiss city of Zug revealed that automated shuttle buses experience manoeuvring challenges, and their integration with public transport systems requires further technological advancements (Schweizerische Bundesbahnen (SBB), 2020). Trials in Baltic region cities showed that due to an obstacle detection issue, human drivers often took control of driving, even though shuttles were expected to operate at level 4 of automation (Bellone et al., 2021). Beauchamp et al. (2022) investigated automated shuttle operation in the Canadian urban traffic context. Recorded data revealed that shuttles travelled at relatively low speeds, 7-12 km/h, compared to motorised vehicles using the same path. Overall, case studies show the need for more research to integrate automated shuttles into public transport before increasing their use and automation. Like the above-described study by Tarkiainen et al. (2024), this paper presents a case study on the effectiveness of tram warning applications for an automated shuttle bus and tram interaction in the Hervanta testbed, but under different traffic conditions and in a simulated environment. The research findings have been published as part of a Master of Science thesis at the University of Helsinki, titled as Effects of an Automated Shuttle Bus on Urban Traffic in Tampere, Finland (Chkhartishvili, 2024). 184 BASt / A 53 5 YRS 2025 / Effects of Tram Warning Application on Automated Shuttle Bus and Tram Interaction 2 Methodology Given the safety, technical, and infrastructural difficulties of real-world testing of automated shuttle buses, conducting experiments in a simulated environment emerged as a practical alternative. For this reason, the study applied traffic microsimulation to investigate the effects of the shuttle bus operation itself, and its operation combined with a tram warning application on the travel time of the tram. Multimodal traffic data was integrated into the simulation model created with the traffic microsimulation tool PTV Vissim. This methodology is based on our own study conducted in the Hervanta testbed (Chkhartishvili, 2024). 2.1 Study area Hervanta is a testbed for innovative solutions, and testing automated shuttle buses is among them (Business Tampere, n.d.). It is a suburban district of Tampere. A tram line provides a fast connection between Hervanta and Tampere city centre. In the area of interest for this study, the automated shuttle shares a lane with a tram in trial runs. Figure 1 shows the characteristics of the respective area in Hervanta that serve as the basis for the simulation model. The simulation network spans a total length of 500m. It includes one intersection and a road section where an automated shuttle shares the road with a tram and other road users, including pedestrians, cyclists, and private vehicles. Figure 1-2 Map illustrates the intersection of interest in the research, where the automated shuttle bus and tram travel in an overlapping lane. It shows the routes and stops of both the automated shuttle bus and the tram in Hervanta, Tampere. The basemap source is Esri Community Maps Contributors. 6 YRS 2025 / Effects of Tram Warning Application on Automated Shuttle Bus and Tram Interaction 2.2 Simulation model The traffic microsimulation allows a systematic assessment of testing an automated shuttle bus operation under higher traffic volumes and shuttle departure frequencies. The microsimulation traffic model replicates the characteristics of the study area. Creating the simulation model in the Vissim software encompassed building a multi-modal road network, placing speed restrictions, incorporating road elements (pedestrian crossings, road markings, tram stops), assigning right of way and creating vehicle input. The described steps resulted in a realistic road system where pedestrians, bicyclists, core public transport, private vehicles and automated shuttle buses could operate. Additionally, we modified the driving behaviour traits of the shuttle using the parameters of the Wiedemann 99 car-following model (Sukennik, 2020). Figure 2 shows the finalised simulation model. Figure 2-2 Simulation model of an automated shuttle bus testbed in PTV Vissim software. A bright green vehicle depicts the automated shuttle bus. 2.3 Traffic data Simulations used data from the testbed, information collected during a site visit, and relevant open traffic data. 2.3.1 Automated shuttle bus and tram An automated shuttle bus operated in a living lab in 2023 and was characterized by low speed (limited up to 20 km/h) and automation level 4. Its manufacturer was Auve Tech, and it was operated by Remoted Oy. During test runs, the shuttle bus had a fixed route in mixed traffic conditions and shared a part of the route with an existing tram network. It was travelling in a single direction and stopped only at the designated two stopping areas. We were particularly interested in a shuttle-tram interaction, therefore, only the intersection where they affected each other’s travel was investigated. We integrated an existing tram schedule and automated shuttle bus operation in the simulation network. Information about peak hour tram traffic and departure times was obtained from the Tampere Regional Transport Authority—Nysse website. 185 BASt / A 53 6 YRS 2025 / Effects of Tram Warning Application on Automated Shuttle Bus and Tram Interaction 2.2 Simulation model The traffic microsimulation allows a systematic assessment of testing an automated shuttle bus operation under higher traffic volumes and shuttle departure frequencies. The microsimulation traffic model replicates the characteristics of the study area. Creating the simulation model in the Vissim software encompassed building a multi-modal road network, placing speed restrictions, incorporating road elements (pedestrian crossings, road markings, tram stops), assigning right of way and creating vehicle input. The described steps resulted in a realistic road system where pedestrians, bicyclists, core public transport, private vehicles and automated shuttle buses could operate. Additionally, we modified the driving behaviour traits of the shuttle using the parameters of the Wiedemann 99 car-following model (Sukennik, 2020). Figure 2 shows the finalised simulation model. Figure 2-2 Simulation model of an automated shuttle bus testbed in PTV Vissim software. A bright green vehicle depicts the automated shuttle bus. 2.3 Traffic data Simulations used data from the testbed, information collected during a site visit, and relevant open traffic data. 2.3.1 Automated shuttle bus and tram An automated shuttle bus operated in a living lab in 2023 and was characterized by low speed (limited up to 20 km/h) and automation level 4. Its manufacturer was Auve Tech, and it was operated by Remoted Oy. During test runs, the shuttle bus had a fixed route in mixed traffic conditions and shared a part of the route with an existing tram network. It was travelling in a single direction and stopped only at the designated two stopping areas. We were particularly interested in a shuttle-tram interaction, therefore, only the intersection where they affected each other’s travel was investigated. We integrated an existing tram schedule and automated shuttle bus operation in the simulation network. Information about peak hour tram traffic and departure times was obtained from the Tampere Regional Transport Authority—Nysse website. 186 BASt / A 53 7 YRS 2025 / Effects of Tram Warning Application on Automated Shuttle Bus and Tram Interaction 2.3.2 Cars, pedestrians and cyclists The City of Tampere's open database provided information about the existing traffic situation. The simulation used pedestrian, cyclist, car, and truck traffic counts for the maximum weekday hourly traffic in 2023. Low vehicular traffic in the study area challenged the examination of automated shuttle interactions with other road users. To address this, we increased the actual traffic by factors of 3 and 6 for lowand high-demand conditions, respectively. We gathered open data exclusively from two traffic collection points, and it lacked detailed information about the movement of vehicles in every direction. 2.4 Scenarios The simulation model incorporates traffic data and a tram warning application. The tram warning application instructs the automated shuttle bus to remain at the bus stop until the tram has passed through the intersection being analysed. This guarantees that the slow-moving automated shuttle bus does not prevent the tram from passing the intersection. The tram warning application is not a built-in attribute in simulation software; a separate script is needed to integrate it with public transport. We modelled six scenarios in PTV Vissim. For lowand high-demand conditions, scenarios separately investigate the existing traffic situation, automated shuttle bus operation and tram warning application. The low-demand scenarios represent the current traffic conditions, whilst the high-demand scenarios take future conditions into account and include doubled traffic data for all travel modes apart from the tram and shuttle. Table 1 shows a summary of the simulated scenarios. Scenarios Characteristics 1 Low demand The baseline conditions. Pedestrians, bicyclists, trams and cars travel in the study area. 2 Low demand The automated shuttle bus starts to travel on a fixed route, sharing an overlapping lane with a tram. 3 Low demand The tram warning script is activated, instructing the shuttle to wait at a bus stop while the tram is travelling. 4 High demand Scenario 1 but with doubled traffic (shuttle and tram volumes remain the same). 5 High demand Scenario 2 but with doubled traffic (shuttle and tram volumes remain the same). 6 High demand Scenario 3 but with doubled traffic (shuttle and tram volumes remain the same). Table 1-2 Scenario description. 187 BASt / A 53 8 YRS 2025 / Effects of Tram Warning Application on Automated Shuttle Bus and Tram Interaction We run each scenario 50 times, each lasting 800 seconds, with a random seed increment of 1. During each run, we collected the travel times between two stops for both the tram and the automated shuttle bus. Figure 3 shows the automated shuttle bus and tram stops and routes in a simulated model. Figure 3-2 Travel time data collection points in a simulation model for the automated shuttle bus and tram. The basemap depicting the simulation road network is from PTV Vissim software. 188 BASt / A 53 9 YRS 2025 / Effects of Tram Warning Application on Automated Shuttle Bus and Tram Interaction 3 Results Traffic simulation results on Figure 4 and Figure 5 helped to clarify the research questions of this study regarding the impact of the automated shuttle bus on tram travel time, and the effects of a tram warning application on the travel times of both the automated shuttle bus and the tram. Figure 4 depicts the tram travel time descriptive statistics based on six scenarios: 1. “Tram – low demand” is a baseline scenario of low traffic demand conditions before the automated shuttle bus started operating in the study area, 2. “Tram – high demand” is a baseline scenario of high-demand traffic conditions before the automated shuttle bus started operating in the study area, 3. “Shuttle – low demand” is a scenario where an automated shuttle bus started sharing an overlapping lane with an existing tram in the low-demand traffic conditions, 4. “Shuttle – high demand” is a scenario where an automated shuttle bus starts sharing an overlapping lane with an existing tram in high-demand traffic conditions, 5. “Tram warning – low demand” is a scenario in which the tram warning is activated and instructs the shuttle to wait at a bus stop if the tram travels in a shared lane in low-demand traffic conditions, 6. “Tram warning – high demand” is a scenario when the tram warning was activated and instructed the shuttle to wait at a bus stop if the tram was travelling in a shared lane in the high-demand traffic conditions. Figure 5 illustrates simulation results, but for the automated shuttle bus travel time. Simulation results show that the tram’s average travel time increased by 15% (from 62 seconds to 71 seconds) in the low-demand scenarios when a slow-moving automated shuttle shared a lane with the tram. However, the tram warning application prevented a delay caused by the automated shuttle, as the application informed the shuttle bus about the tram location and suggested waiting at a bus station. Thus, the tram warning application decreased the tram’s average travel time to its original duration but increased the average travel time of the automated shuttle bus by 33% (from 141 seconds to 188 seconds). The tram warning application similarly affected shuttle and tram travel times in high-demand scenarios within the research area. In the high-demand scenario, the automated shuttle increased the tram’s average travel time by 9% (from 68 seconds to 74 seconds), while the tram warning application managed to avert this delay. Due to the tram warning service, the automated shuttle bus was also delayed during higher traffic conditions. The shuttle’s average travel time was increased by 34% (from 146 seconds to 195 seconds). 189 BASt / A 53 10 YRS 2025 / Effects of Tram Warning Application on Automated Shuttle Bus and Tram Interaction Figure 4-3 Descriptive statistics of the tram travel time for low and high demand conditions and different traffic scenarios. The top whiskers denote the highest travel time values (i.e., 80.4 s, 84.0 s, 93.0 s, 98.0 s, 80.4 s, 84.0 s), and the lower ones indicate the minimum travel time values (i.e., 55.3 s, 55.8 s, 55.4 s, 55.3 s, 55.3 s, 55.8 s). The line inside the boxplot represents the median values of the data within each scenario; crosses show the average values, which are also given in numeric values. The dots are outliers in the dataset. Figure 5-3 Descriptive statistics of the automated shuttle bus travel time for low and high demand conditions and different traffic scenarios. The top whiskers (i.e., 150.1 s, 167.2 s, 222.7 s, 247.6 s) denote the highest travel time value, and the lower ones (i.e., 132.3 s, 133.5 s, 149.2 s, 152.7 s) indicate the minimum travel time values. The line inside the boxplot represents the median values of the data within each scenario; crosses show the average values, which are also given in numeric values. The dots are outliers in the dataset. 196 BASt / A 53 17 YRS 2025 / Effects of Tram Warning Application on Automated Shuttle Bus and Tram Interaction Pictures Figure 1-2: Study area map in Hervanta, Tampere Figure 2-2: Simulation model in PTV Vissim software: bright green vehicle depicts an automated shuttle bus Figure 3-2: Travel time data collection points in a simulation model for the automated shuttle bus and tram. The basemap depicting the simulation road network is from PTV Vissim software Figure 4-3: Descriptive statistics of the tram travel time for low and high demand conditions and different traffic scenarios. The top whiskers denote the highest travel time values (i.e., 80.4 s, 84.0 s, 93.0 s, 98.0 s, 80.4 s, 84.0 s), and the lower ones indicate the minimum travel time values (i.e., 55.3 s, 55.8 s, 55.4 s, 55.3 s, 55.3 s, 55.8 s). The line inside the boxplot represents the median values of the data within each scenario; crosses show the average values, which are also given in numeric values. The dots are outliers in the dataset Figure 5-3: Descriptive statistics of the automated shuttle bus travel time for low and high demand conditions and different traffic scenarios. The top whiskers (i.e., 150.1 s, 167.2 s, 222.7 s, 247.6 s) denote the highest travel time value, and the lower ones (i.e., 132.3 s, 133.5 s, 149.2 s, 152.7 s) indicate the minimum travel time values. The line inside the boxplot represents the median values of the data within each scenario; crosses show the average values, which are also given in numeric values. The dots are outliers in the dataset 197 BASt / A 53 18 YRS 2025 / Effects of Tram Warning Application on Automated Shuttle Bus and Tram Interaction Tables Tab. 1-2: Scenario description