Route Optimisation
Digital Transport Solution for Lithuania's Port City / Case
A prototype for simulating public transport route scheduling was developed for a transport provider in Lithuania’s port city Klaipėda.
The aim was to optimize public transport schedules using machine learning models to analyze and predict passenger flows and traffic data. By integrating historical and predictive data, the system improved scheduling, minimizing passenger wait times, reducing traffic congestion, and contributing to a more sustainable urban environment.
Problem
3 ticket zones
31 urban routes
Shuttle taxis: 3 routes in Klaipėda, 2 in Palanga, Kretinga
22 suburban routes
The public transport system in Klaipėda faced several challenges, including long waiting times for passengers at bus stops due to inefficient scheduling and the frequent occurrence of multiple buses arriving at the same time, causing traffic bottlenecks and operational inefficiencies. Overlapping bus routes at key stations led to the inefficient use of vehicles, while delays and worsening traffic conditions further compounded the problem, making the system less efficient.
Bus delays
As one can see from the data below, buses can be delayed or rushed by up to 3 minutes. One can also observe that buses are more likely to be late than early. Although the vast majority of buses arrive on time, some routes are off schedule due to driver errors or traffic jams.
Solution
To overcome the challenges, we developed a machine learning-based simulation model to optimize public transport schedules in Klaipėda by analyzing real-time and historical data. This model provided accurate passenger flow predictions at different times and locations, suggested schedule changes based on predicted passenger numbers and traffic conditions, and optimized the allocation of buses and drivers according to real-time demand. The system integrated historical data from the public transport network, including bus stop locations, traffic incidents, and passenger flows, to train the model. Additionally, a web-based user interface was created to display real-time and historical passenger flow data, predictions, and schedule optimization recommendations.
Technology used
For the proposed solution, we used Python (Flask/Django) for server-side logic. To analyze and predict passenger flows, we considered machine learning models such as Artificial Neural Networks (ANN), Support Vector Classifiers (SVC), and Monte Carlo methods. The user interface was built using React for the frontend, while PostgreSQL was utilized for data storage and integration with weather and traffic condition APIs.
Results
After successful development, the solution was tested on one of the bus routes in the city. Optimized schedules, driven by predictive models, reduced passenger waiting times. The reallocation of buses helped to prevent congestion at key intersections and bus stops, thereby improving traffic flow.
Additionally, by minimizing bus idling and rerouting underutilized vehicles, the solution contributed to reduced carbon emissions, aligning with the municipality’s environmental goals. This prototype demonstrated the potential for scaling the solution to other bus routes in the city, with further applications possible across various cities.
Potential applications in other industries
The machine learning-based simulation model used to optimize public transport scheduling can be adapted for other industries that rely on logistics and resource management. For example, in the healthcare sector, such a model could be used to predict patient flow in hospitals, helping administrators allocate staff and medical resources more efficiently based on real-time and historical data.
Similarly, retail logistics could benefit from predictive models to optimize inventory management and staffing levels, reducing waste and ensuring products are stocked in line with customer demand patterns. In manufacturing, machine learning could predict equipment usage and maintenance needs, ensuring downtime is minimized and production schedules run smoothly.
Client
Klaipėda Transport