AIForLEssAuto

Artificial Intelligence for Urban Low-Emission Autonomous Traffic (AIForLEssAuto) investigates how autonomous electrified traffic should be organized and managed in order to maximise the reduction of CO2 emissions in cities.
Highlights
Dynamic on-road emission estimation

Collaboration: WP3 + WP5 

  • We developed a geospatial framework to estimate dynamic traffic emission using floating car data (FCD) '
  • The framework was built using the macroscopic fundamental diagram (MFD) and generalized linear model (GLM), which consider the variation in congestion level
  • Congestion-induced emissions were calculated to account for up to 10% of the total vehicular emissions in Helsinki (Fig1, the figure shows CO2 emission only) 
  • Using the framework, we calculated that the introduction of CAVs could result in emission reductions of 3–14% owing to congestion improvements 

Fung, P. L., Al-Jaghbeer, O., Chen, J., Paunu, V. V., Vosough, S., Roncoli, C., & Järvi, L. (2024). A geospatial approach for dynamic on-road emission through open-access floating car data. Environmental Research Letters. Doi: https://doi.org/10.1088/1748-9326/ad984d 
City-wide emission upscaling for deep RL methods

Collaboration: WP4 + WP5

  • We developed a novel emission upscaling method to estimate city-wide sustainability effects 
  • The method supports areas of restricted sizes for RL experiments (Fig. 3) and, thus, reduces time to obtain the results
  • Applications: ride-pooling, electric charging stations placement. 

Deep reinforcement learning for emission reduction at scale: a case study for Helsinki city area // Transportation Research Part D: Transport and Environment. To be submitted. 
Design of dynamic routing strategies

Collaboration: WP3 + WP4 + WP5

  • Using SUMO simulator and Helsinki data to assign incentives with a MARL algorithm to improve social benefit.
  • Incentives reach the lowest possible total travel time under an unlimited budget and decrease congestion in the network.
  • From SUMO outputs, WP5 calculates CO2 emissions allowing us to measure the effectiveness of incentive schemes on the environment.
Design of reward functions for EV drivers

Collaboration: WP2 + WP4

  • Using SUMO simulator and Helen charging data to model the reward functions of the drivers
  • The reward functions enable the benchmarking of different RL algorithms and more extensive simulation of the drivers
  • Applications: driver modelling, RL benchmarking, cognitive rationality

Pyykölä S., Baimukashev D., Bochenina K., Ruotsalainen L. Learning reward functions for EV drivers using unsupervised feature selection. To be submitted.
Assessing sustainability impact of RL

WP4

  • Utilizing the electricity information from LUMI, calculating the energy efficiency and carbon footprint of CPU and GPU computations in reinforcement learning
  • The model enables an empirical evaluation of RL models’ electricity efficiency, including the analysis from the effects of computing time and hardware
  • Applications: green computing

Kuczkowski D., Pyykölä S., Heljanko, K., Bochenina K., Ruotsalainen L. A Multi-Objective Approach to Reinforcement Learning for Electric Vehicle Charging: Simulating Demand 
Main info

Project duration 1.1.2022 – 31.12.2024 

Funding: This project has received funding from the European Union – NextGenerationEU instrument and is funded by the Research Council of Finland for 2022-2024.

PIs:

  • Prof. Laura Ruotsalainen, University of Helsinki, Helsus
  • Assoc. Prof. Ville Kyrki, Aalto University
  • Assoc. Prof. Claudio Roncoli, Aalto University
  • Prof. Leena Järvi, INAR, Helsus

Collaborators: Traficom, Finnish Climate Fund, Sensible 4, HSL, HSY 
Project goals
  • developing a plan how autonomous electrified traffic should be organized and managed in order to maximise the reduction of CO2 emissions in cities
  • developing a framework of computational modelling tools to evaluate the CO2 emissions originating from the introduction of electrified automated vehicles in the urban road context; 
  • developing AI-based control strategies for from vehicle-level to city-center wide traffic control in which CO2 emissions are minimized. 
Research questions
  1.  How much do the different driving routines and connected driving of AVs (for example accelerations, change of lanes, traffic light communication) impact the electricity consumption of electrified vehicles and furthermore CO2 emissions?
  2. What kind of traffc organization, i.e. electric vehicles, public transport, proportion of autonomy, positioning of electric charging devices, would create least CO2 emissions?
  3. Are there significant impacts of introducing autonomous electrified vehicles to traffic scenarios to city-level CO2 emissions?
Work packages
Title Principal investigator Description
1. Stakeholder interaction Laura Ruotsalainen This WP aims to define traffic scenarios with key stakeholders during joint workshops and to form recommendations on future traffic implementations based on the results of the project.
2. AI-based minimization of vehicle level emissions Ville Kyrki This WP aims to answer the question, what level of emission reduction can be gained if control of autonomous vehicles is optimized based on minimizing emissions on the level of single vehicles and local fleets.
3. Novel online traffic decisions Claudio Roncoli This WP aims at developing traffic management strategies to reduce the emissions at local levels (i.e., links and junctions) and at network level (group of intersections and routing decisions within a network).
4. RL traffic level modelling Laura Ruotsalainen This WP aims at simulating the traffic with different vehicle types and traffic planning decisions, and optimize the system to reduce the resulting CO2 emissions.
5. Calculation of CO2 emissions Leena Järvi This WP aims at estimating the total CO2 emissions of mixed road traffic in different vehicle, local and city level scenarios at different temporal scales.
Results

Publications

Work package Authors Title of the paper Venue Year Status Link to the paper
WP2 Herm, F., Mazumdar, A., Chugh, T. Multi-Objective Multi-Agent Reinforcement Learning for Autonomous Driving in Mixed-Traffic Environments Evolutionary Multi-Criterion Optimization (EMO) 2025 Accepted  
NA Mazumdar, A., Jain, B., Mitra, M., & Dhar, P.  Interactive Evolutionary Multiobjective Optimization of Primer Design with Uncertain Objectives Genetic and Evolutionary Computation Conference 2024 2024 Published https://dl.acm.org/doi/abs/10.1145/3638529.3654167
WP2 Azam, S., Kyrki, V. Multi-Task Adaptive Gating Network for Trajectory Distilled Control Prediction IEEE Robotics and Automation Letters 2024 Published  
WP2 Azam, S., Munir, F., Kyrki, V., Kucner, T. P., Jeon, M., Pedrycz, W. Exploring Contextual Representation and Multi-Modality for End-to-End Autonomous Driving Engineering Application of Artificial Intelligence 2024 Published  
WP2 Baimukashev, D., Alcan, G., Luck KS, Kyrki, V. Learning Transparent Reward Models via Unsupervised Feature Selection CoRL 2024 Published https://openreview.net/pdf?id=2sg4PY1W9d
WP2 Kujanpää, K., Baimukashev, D., Zhu, S., Azam, S., Munir, F., Alcan, G., Kyrki, V. Challenges of Data-Driven Simulation of Diverse and Consistent Human Driving Behaviors AAAI Workshop 2024 Published https://arxiv.org/pdf/2401.03236
WP2 Lu J., Alcan G., Kyrki V. Integrating Expert Guidance for Efficient Learning of Safe Overtaking in Autonomous Driving Using Deep Reinforcement Learning   2024 Published  
WP2 Mazumdar, A., Kyrki, V. Hybrid Surrogate Assisted Evolutionary Multiobjective Reinforcement Learning for Continuous Robot Control EvoApplications 2024 (Applications of Evolutionary Computation). 2024 Published https://link.springer.com/chapter/10.1007/978-3-031-56855-8_4
WP3 Hu S. et al. High time-resolution queue profile estimation at signalized intersections based on Extended Kalman Filtering   2023 Published https://doi.org/10.1109/TITS.2022.3173515
WP3 Khavarian K., Vosough S., Roncoli C. How do electric bikes affect the route choice of cyclists? A case study of Greater Helsinki 11th Symposium of the European Association for Research in Transportation 2023 Published https://transp-or.epfl.ch/heart/2023/abstracts/hEART_2023_paper_5563.pdf
WP3 Mohammadi R., Vosough S., Roncoli C. User-based transit signal priority in a connected vehicle environment accounting for schedule delays and environmental impacts Transportation Research Board 102nd Annual Meeting 2023 Published https://trid.trb.org/view/2117822
WP3 Vol O., Vosough S., Roncoli C. Drivers’ compliance with social route recommendations: Stated intentions vs actual behavior 8th International Conference on Models and Technologies for Intelligent Transportation Systems 2023 Published https://nextcloud.eurecom.fr/s/Brio7KWdESd6DX5 
WP3 Vosough S., Roncoli C. Drivers’ social routing behavior: Evidence from stated and revealed preferences experiments in two European cities Transportation Research Board 102nd Annual Meeting 2023 Published https://trid.trb.org/view/2087445
WP3 Mohammadi R., Vosough S., Roncoli C. Analysing the Environmental and Social Impacts of a Novel User-Based Transit Signal Priority Strategy in a Connected Vehicle Environment Journal of Advanced Transportation 2024 Published https://doi.org/10.1155/2024/8712813
WP3 Liaquat M., Vosough S., Roncoli C., Charalambous T. Assessing the Performance of Max-Weight Traffic Signal Control in the Presence of Noisy Queue Information - An Environmental Impacts Evaluation IET Intelligent Transport Systems 2024 Published https://doi.org/10.1049/itr2.12571
WP3 Khavarian K., Vosough S., Roncoli C. Bike users’ route choice behaviour: Expectations from electric bikes versus reality in Greater Helsinki Journal of Cycling and Micromobility Research 2024 Published https://doi.org/10.1016/j.jcmr.2024.100045
WP3 Tarkkala K., Vosough S., West J., Roncoli C. Evaluating the influence of cyclists’ route choices incorporation into travel demand modelling: A case study in greater Helsinki Transportation Research Interdisciplinary Perspectives 2024 Published https://doi.org/10.1016/j.trip.2024.101224
WP3 Vosough S., Roncoli C. Achieving social routing via navigation apps: User acceptance of travel time sacrifice Transport Policy 2024 Published https://doi.org/10.1016/j.tranpol.2024.01.026
WP3 Dibaj S., Vosough S., Kazemzadeh K, O’Hern S., Mladenović M. An exploration of e-scooter injuries and severity: Impact of restriction policies in Helsinki, Finland Journal of Safety Research 2024 Published https://doi.org/10.1016/j.jsr.2024.09.006
WP3 Niroumand R., Vosough S., Rinaldi M., Connors R., Roncoli C. Balancing congestion and emissions in urban networks via path-based incentives 5th Symposium on Management of Future Motorway and Urban Traffic System (MFTS) 2024 Published  
WP3 Niroumand R., Vosough S., Rinaldi M., Connors R., Roncoli C. Beyond links: The power of path incentives in alleviating congestion and emissions in urban networks 12th Symposium of the European Association for Research in Transportation (hEART) 2024 Published https://transp-or.epfl.ch/heart/2024/abstracts/hEART_2024_paper_1587.pdf
WP3 Tarkkala K., Vosough S., West J., Roncoli C. Incorporating cyclists' route choice models into travel demand modelling: A case study in Greater Helsinki Transportation Research Board 103rd Annual Meeting 2024 Published  
WP3 Khavarian K., Vosough S., Roncoli C. Bridging the Expectation vs Reality Gap on Route Choice Behavior of E-bike Users: Evidence from the Greater Helsinki Region in Finland Transportation Research Board 103rd Annual Meeting 2024 Published  
WP3 Muhammad H., Roncoli C., Niroumand R. Optimization-based Urban Network Traffic Management with Mixed Autonomy Incorporating Dynamic Saturation Rates 5th Symposium on Management of Future Motorway and Urban Traffic Systems (MFTS) 2024 Published https://research.aalto.fi/files/159424460/MFTS_extended_abstract.pdf
WP4 Bochenina K., Taleiko A., Ruotsalainen L. Simulation-based origin-destination matrix reduction: a case study of Helsinki city area SUMO User Conference 2023 2023 Published https://researchportal.helsinki.fi/en/publications/simulation-based-origin-destination-matrix-reduction-a-case-study
WP4 Bochenina K., Ruotsalainen L.  A reinforcement learning-based metaheuristic algorithm for on-demand ride-pooling IE 2024 (20th International Conference on Intelligent Environments) 2024 Published https://ieeexplore.ieee.org/abstract/document/10599906/
WP4 Pyykölä S., Bochenina K., Ruotsalainen L. Conciliator steering: Imposing user preference in multi-objective reinforcement learning Transactions on Machine Learning 2024 Published https://openreview.net/pdf?id=XAD2kcBS50
WP5 Fung P. L., Al-Jaghbeer O., Pirjola L., Aaltonen H., Järvi L. Exploring the discrepancy between top-down and bottom-up approaches of fine spatio-temporal vehicular CO2 emission in an urban road network Science of the Total Environment 2023 Published https://doi.org/10.1016/j.scitotenv.2023.165827
WP5 Al-Jaghbeer O., Fung P. L., Paunu V.-V., Järvi L. Mapping CO2 traffic emissions within local climate zones in Helsinki Urban Climate 2024 Published https://doi.org/10.1016/j.uclim.2024.102171
WP5, WP3 Fung P. L., Al-Jaghbeer O., Chen J., Paunu V.-V., Vosough S., Roncoli C., Järvi L. A geospatial approach for dynamic on-road emission through open-access floating car data Environmental Research Letters 2024 In print https://doi.org/10.1088/1748-9326/ad984d

 

Pending work

Work package Authors Title of the paper Venue Year Status Link to the paper
WP2 Alcan G., Abu-Dakka F. J., Kyrki V. Constrained Trajectory Optimization on Matrix Lie Groups via Lie-Algebraic Differential Dynamic Programming Automatica 2024 Submitted https://arxiv.org/abs/2301.02018
WP2 Lu, J., Azam, S., Alcan, G., Kyrki, V. Data-driven Diffusion Models for Enhancing Safety in Autonomous Vehicle Traffic Simulations ICRA 2025 Submitted  
WP2 Baimukashev, D., Alcan, G., Kyrki, V. Feature selection for inverse reinforcement learning IROS 2024 Submitted  
WP2 Baimukashev, D., Le, T.N., Azam, S., Kyrki, V. Harnessing Suboptimality: Weakly Supervised Imitation Learning with Diverse Demonstrations ICRA 2025 Submitted  
WP2 Kujanpää, K., Baimukashev, D., Munir, F., Azam, S., Kucner, T.P., Pajarinen, J., & Kyrki, V. Discrete Contrastive Learning for Diffusion Policies in Autonomous Driving ICRA 2025 Submitted  
WP3 Niroumand R., Vosough S., Rinaldi M., Connors R., Roncoli C. Mitigating network traffic congestion via link- and path-based incentives Transportation Research Part B: Methodological 2024 Submitted  
WP3, WP2 Pardo-González G., Vosough S., Papadaki K., Roncoli C. Alleviating congestion in the urban transportation network by implementing incentive schemes in SUMO: A case study in Helsinki 13th Symposium of the European Association for Research in Transportation (hEART) 2025 Work in progress  
WP3 Muhammad H., Roncoli C. Optimizing Urban Traffic Networks With Dynamic Saturation Rates in a Mixed Autonomy Environment 23rd European Control Conference (ECC) 2024 Submitted  
WP3 Muhammad H., Roncoli C. Network-wide Urban Traffic Optimization With Dynamic Saturation Rates in a Mixed Autonomy Environment IEEE Transactions on Intelligent Transportation Systems 2024/25 Work in progress  
WP4 Bochenina K., Agriesti S., Roncoli C., Ruotsalainen L. From urban data to city-scale models: A review of traffic simulation case studies  IET Intelligent Transport Systems 2024 Submitted  
WP4 Bochenina K., Beimuk V., Ruotsalainen L.  Parallel reinforcement learning for on-demand ride-pooling: a scalability study and performance model PDP 2025 (33th Euromicro International Conference on Parallel, Distributed and Network-based Processing) 2024 Submitted  
WP4, WP5 Bochenina K., Fung P. L., Al-Jaghbeer O., Pyykölä S., Järvi L., Ruotsalainen L. Deep reinforcement learning for emission reduction at scale: a case study for Helsinki city area Transportation Research Part D: Transport and Environment 2025 Work in progress  
WP4 Kuczkowski D., Pyykölä S., Heljanko, K., Bochenina K., Ruotsalainen L.  A Multi-Objective Approach to Reinforcement Learning for Electric Vehicle Charging: Simulating Demand and Assessing Sustainability Impacts CCAI+JMLR special issue 2025 Work in progress  
WP4 Pyykölä S., Kuczkowski D., Bochenina K., Ruotsalainen L. Offline multi-objective reinforcement learning for electric vehicle charging demand ICML 2025 2025 Work in progress  
WP4, WP2 Pyykölä S., Baimukashev, D., Bochenina K., Ruotsalainen L. Learning reward functions for EV drivers using unsupervised feature detection   2025 Work in progress  
WP5 Fung P. L. Congestion radials: Visualising dynamic traffic patterns across Europe Environment and Planning B: Urban Analytics and City Science 2024 Submitted  
Contact

Coordinator: Laura Ruotsalainen, laura.ruotsalainen@helsinki.fi