This study addresses the challenges of optimizing urban transport systems in cities characterized by hybrid infrastructures, where historic preservation coexists with modern mobility demands. Focusing on Lviv, Ukraine-a mid-sized European city with a UNESCO-listed core and rapidly expanding periphery-we propose four novel AI-driven models to achieve context-aware optimization: a Dynamic Zonal Optimization Model (DZOM)[1] that enforces adaptive traffic policies across heritage, transition, and modern zones; a decentralized edge-cloud computing framework (DECENTRA)[2] leveraging tram networks for low-latency incident response; a Multimodal Mobility Graph (MMG)[3] integrating reinforcement learning to minimize intermodal transfer delays; and a privacy-preserving Crowdsourced Congestion Forecasting (CCF)[4] system using federated learning. The research employs a simulation-based methodology, validating models through SUMO and Aimsun platforms calibrated with 2023 traffic data from Lviv. Key results demonstrate a 32% reduction in peak-hour congestion, an 18% decrease in CO₂ emissions, and a 24% increase in tram ridership following system integration. The DZOM [1] reduced pedestrian wait times in heritage zones by 28%, while the MMG[3] cut average intermodal transfer delays by 43% during peak tourism events. The CCF[4] system achieved an 89% congestion prediction accuracy with a strict privacy budget (ε = 0.29), addressing GDPR concerns absent in conventional CCTV-based approaches. This work contributes to transport science by introducing a scalable framework for cities balancing heritage constraints with modernization pressures. Unlike prior studies focused on megacities, our models prioritize decentralized data processing, geospatial adaptability, and citizen privacy-critical factors for mid-sized European urban centers. The demonstrated annual fuel savings of 1.2 million liters and improved multimodal coordination provide a replicable blueprint for sustainable mobility in analogous regions.
Rapid urbanization and escalating freight demands have strained traditional transport systems, necessitating innovative solutions. Global urban populations are projected to reach 68% by 2050, exacerbating congestion, pollution, and inefficiencies. Smart automation and AI emerge as pivotal tools, leveraging real-time data and adaptive algorithms to optimize networks. This article examines their applications, benefits, and challenges, offering a roadmap for stakeholders in academia, industry, and policy.
routes, addressing urban growth challenges. Recent data confirms urban populations will reach 68% by 2050, increasing pressure on transport systems. Cities like Singapore and Los Angeles are leading with AI-driven solutions, while the COVID-19 pandemic has highlighted the need for resilient systems.
IoT sensors and connected devices enable real-time traffic monitoring, with Singapore’s adaptive
[
There are the following approaches for the Smart Automation and AI-Driven Optimization in
Machine learning (ML) is used to predict travel demand by analyzing data such as traffic patterns and weather conditions. Examples include:
2.1.3.
Reinforcement learning (RL) optimizes routing by learning from trial and error, rewarding efficient choices:
To tailor approaches and models effectively for our region, we must first conduct a thorough
analysis of local transportation patterns and infrastructure capabilities. By collecting and
examining region-specific data, we can identify unique challenges and opportunities that may not
be present in other areas. Additionally, collaborating with local stakeholders and authorities can
provide valuable insights and support the implementation of customized solutions. By leveraging
advanced analytics and machine learning techniques, we can develop predictive models that are
finely tuned to regional nuances, ultimately enhancing the efficiency and effectiveness of transport
networks in our area.
Objective: Reduce congestion in historic zones while balancing pedestrian-transport priorities [
Optimization Criteria:
Simulation:
Method: Multi-objective genetic algorithm: = 0.5 ∗ (ℎ⁄)
()
+ 0.3 ∗
+ 0.2 ∗
()
,
(1)
100 10 60
Coefficients reflect municipal priorities; normalization scales heterogeneous units to [
Case Study: During the 2023 "Dreamland Festival," DZOM rerouted 15 tour buses to peripheral lots, reducing Rynok Square congestion from 95 to 60 veh/km. Pedestrian access time to attractions dropped from 20 to 8 min.
= 10 ∗ () + 5 ∗ () − 3 ∗ (), : ; : 2 ;
: . 29 min 1.3 Case Study: During the 2023 Lviv Jazz Fest, MMG diverted 30% of tram users to bike-sharing hubs, reducing overcrowding by 35%. 3.3. Crowdsourced Congestion Forecasting (CCF) Objective: Predict congestion with ≥85% accuracy under GDPR compliance. Constraints: Forecast Accuracy (MAE) – < 4.5 min.
Privacy Budget (ε) – ε < 0.5.
Simulation: Data: Anonymized GPS traces from 30,000 users.
Architecture: Federated LSTM networks with differential privacy:
MMG created pop-up bike lanes, reducing access time from 25 to 10 min. CCF preempted a vul. Hnatyuk congestion event 20 min in advance.
3.5.
Singapore’s smart mobility ecosystem leverages AI-powered traffic cameras and the ERP 2.0
congestion pricing system, reducing peak-hour delays by 25% and saving SGD 150 million annually
in lost productivity as of 2024 [
from GPS-equipped taxis and IoT sensors, enabling dynamic traffic light tuning that cuts average
commute times by 10% [
10% in 2024 using Singapore’s AI traffic tools, while Kuala Lumpur targets 15% delay reductions by
2026 [
of hybrid cities, this study provides a replicable blueprint for sustainable mobility in regions where historic preservation and modernization coexist.