Repository files navigation

This repository contains the code of the algorithms and tools developed in Work Package 4 of Artificial Intelligence for Urban Low-Emission Autonomous Traffic (AIForLEssAuto) project.

The webpage of the project

Repositories:

SUMO Helsinki traffic model — pipeline of demand estimation and calibration for SUMO micro-simulation traffic model of Helsinki city area.

The input data are: (i) origin-destination (OD) matrices for traffic assignment zones (TAZs) covering whole Finland fromHSL Helmet model, (ii) Digitraffic traffic counts from traffic counting stations located within Helsinki. The pipeline consists of two steps: (i) OD matrix reduction (for smaller area, for example, Helsinki metropolitan area), (ii) route calibration with traffic counts as ground-truth data (manuscript, please cite asBochenina, K., Taleiko, A., & Ruotsalainen, L. (2023, June). Simulation-based origin-destination matrix reduction: a case study of Helsinki city area. In SUMO User Conference (pp. 1-13)). SUMO network and route files for Helsinki and three (nested within each other) smaller areas, resulting from the pipeline, are available atdemo folder.

SUMO-RL ridepooling — SUMO Gym environment and reinforcement learning (RL algorithm) for centralized dispatching of a fleet of shared taxis with multiple competing objectives. Description of the algorithm can be foundin the manuscript (please cite asBochenina, K., & Ruotsalainen, L. (2024, June). A reinforcement learning-based metaheuristic algorithm for on-demand ride-pooling. In 2024 International Conference on Intelligent Environments (IE) (pp. 117-123). IEEE.)

SUMO-ridepooling testbed — contains scripts for calculating the passenger satisfaction, taxi fleet usage and greenhouse gas (GHG) emissions for SUMO ride-pooling algorithms, performing batched experiments for the range of parameters, aggregating and visualizing the results.

Kamppi minidemo is a small demo about SUMO running traffic simulations in Kamppi for 10 minutes under different scenarios regarding the number of fuel vehicles. The demo creates a static visualisation about the simulation outputs of all scenarios for further analysis.

Interactive visualization tool is a demo for interactive visulizations about SUMO simulation outputs. The repository has a Kamppi simulation environment with one scenario for fuel and electric vehicles.

Conciliator steering is another output of AIForLEssAuto project. The repository features an implementation for the RL algorithm described in a paper published in TMLR in 2024. The algorhtm is tested in DeepSeaTreasure v1 environment and doesn't contain a SUMO environment.