Mobility in 2026: from public transport timetables to AI-driven flow

Mobility is about more than timetables and bus schedules. Transport operators, fleet owners, MaaS providers and municipalities generate data every second: GPS tracks, OCPP charging sessions, GTFS-RT updates, vehicle sensors and passenger movements. AI helps turn that data flow into tangible gains: shorter routes, higher vehicle availability, smarter charging and more reliable travel information. It does not replace planners and drivers, but supports decisions that directly affect costs, CO2 emissions and passenger satisfaction.

Route optimisation Demand forecasting Predictive maintenance EV-charging optimisation Computer vision

Discuss your mobility case View applications

Why mobility and AI reinforce each other

The mobility sector has a data flow that lends itself extremely well to machine learning: predictable patterns (commuter flows, delivery runs, charging behaviour) combined with unpredictable disruptions (congestion, weather, events, breakdowns). Traditional planning works for the average day; AI comes into its own precisely when that average falls short.

A transport operator that runs a thousand journeys a day schedules them based on historical averages and manual corrections. That works well until there is an accident on the A2, a driver calls in sick, or an electric bus turns out to have less range than expected. At that point there are thousands of possible rescheduling options, too many for a person to weigh up within five minutes. This is where optimisation algorithms and machine learning come together: solving the problem mathematically under realistic constraints, while historical data teaches the model which solutions work in practice.

Mobility also moves fast. NDOV data and GTFS-RT feeds from operators, OCPP messages from charging infrastructure, RDW vehicle information and number plate recognition together provide a rich real-time context. AI models that combine these streams can spot patterns that remain invisible within individual systems, such as the link between weather forecasts, public transport demand and shared bike availability at a specific time in a specific neighbourhood.

Core areas where AI is transforming the mobility sector

From planning to maintenance, and from charging infrastructure to passenger information, AI touches every layer of the mobility chain. These are the domains where the business case is strongest.

🗺️

Route and network optimisation

The Vehicle Routing Problem (VRP) and its variants, such as Travelling Salesman, Pickup-and-Delivery and Time Windows, form the core of logistics planning. Using OR-Tools, OptaPlanner and custom heuristics, we build route planners that take into account driver rest periods, vehicle capacity, time slots and real-time traffic.

📊

Demand forecasting for MaaS

How many shared bikes does a neighbourhood need at 8:00 on a rainy Tuesday? How many passengers will board line 4 shortly? Forecasting models combine historical trip data, weather forecasts, event calendars and demographic data to predict demand, providing the basis for proactive fleet management and repositioning.

🔧

Fleet predictive maintenance

Vehicle telematics data (engine temperature, oil pressure, brake friction, battery health) predicts faults before they happen. For bus and truck fleets, this means fewer unexpected breakdowns, planned workshop visits and higher vehicle availability. Models learn patterns by vehicle type and usage profile.

⚡

EV charging optimisation

When a fleet electrifies, charging infrastructure becomes the bottleneck. AI-driven charging schedules combine energy prices (dynamic tariffs), grid congestion, predicted routes for tomorrow and battery condition via OCPP. The result: lower energy costs, reduced peak-time load and vehicles that are fully charged when you need them.

💰

Dynamic pricing for shared mobility

For shared cars, e-scooters and bikes, price is determined by supply and demand in real time. Reinforcement learning and demand curves calculate prices per location and time window, maximising utilisation while offering riders fair fares, much like the approach already familiar from ride-hailing.

🚦

Smart city traffic flow

Adaptive traffic lights, dynamic lane closures and real-time diversion advice run on AI that combines data from induction loops, cameras and floating-car data. For municipalities and provinces, we build dashboards and control systems that optimise traffic without manual control at each junction.

Practical applications in the field

Mobility AI is not a future vision. These are realistic projects that can be built with today's technology, right now.

Last-mile delivery with a VRP engine

A delivery service with fifty vans plans thousands of stops every day. A VRP engine based on OR-Tools processes orders, customer time windows, driver shifts and real-time traffic into a daily schedule. During the day, the system re-plans when delays occur or urgent orders come in. The result: fewer kilometres, higher first-time-right rates and better use of existing vehicles.

ANPR for access control and enforcement

Automatic number plate recognition (ANPR) cameras monitor lorry parks, environmental zones and logistics hubs. Computer vision models read number plates in difficult conditions (dirt, rain, oblique angles) and match them against RDW data or your own access list. For logistics operators, this automates truck check-in and check-out; for municipalities, it supports zone enforcement.

Driver behaviour monitoring with camera AI

In-cab cameras with AI detect fatigue, lack of attention and risky driving (harsh braking, sharp steering). For fleet owners, this means lower accident rates and targeted coaching. We build dashboards that show scores per driver, designed with privacy in mind, with data minimisation and a clear division of roles between employer and driver.

Travel information chatbot with public transport data

A conversational AI fed with 9292 data, NDOV feeds and local timetables answers travel questions in natural language: "How do I get from Amersfoort to Utrecht Centraal at half eight tomorrow morning?" or "Is my Connexxion bus still running?" The chatbot integrates into your app, website or WhatsApp and takes pressure off customer service during peak times.

How Appfront approaches mobility AI

We don't build a generic mobility suite for you to configure yourself. Our approach is hands-on: we analyse your operation, identify where AI will have the greatest impact and build a solution that integrates directly with the systems you already use, whether TMS platforms such as Transpas or Adaption, planning tools, telematics platforms, OCPP back-ends or your own fleet management application.

Mobility organisations differ greatly. A regional public transport concessionholder has different needs from a logistics shipper, a MaaS provider or a municipality steering the mobility transition. We tailor AI components to your situation: what data do you have (GTFS-RT, OCPP, telematics, point-of-sale systems), which processes take up the most time, and where is the greatest impact on costs, CO2 or passenger satisfaction?

Our approach is iterative. We start with a proof of concept on a single use case, such as a VRP prototype for one region or a predictive maintenance model for one vehicle type, and expand when the results justify it. No large upfront investment in a platform that may not fit, but step-by-step validation with measurable results.

AI development AI & IoT solutions Predictive analytics dashboard Enterprise AI implementation Custom software development

From first conversation to live mobility AI

Our approach to AI projects in the mobility sector follows four phases. Each phase delivers a concrete result, with no months of analysis without output.

Data assessment

We inventory the available data: telematics feeds, GTFS-RT, OCPP logs, trip records, sensor data and external sources such as NDOV and RDW. We assess quality, frequency and integration options, and determine which use case is most feasible.

Proof of concept

Within a few weeks we build a working prototype: a VRP engine for one route cluster, a forecasting model for one line, or a predictive maintenance pipeline for one vehicle type. Testable against your own historical data.

Integration and production

The validated model is connected to your existing systems: TMS, planning tool, telematics back end, OCPP platform or fleet management. We build APIs, dashboards and interfaces for your planners and drivers.

Monitoring and adjustment

Mobility models become outdated when infrastructure, timetables or fleet mix change. We monitor model performance, retrain when needed and adjust based on new data and feedback from operations.

Technologies and frameworks we use

The choice of technology depends on the use case. For route optimisation we work with Google OR-Tools and custom solvers; for demand forecasting we use gradient boosting (XGBoost, LightGBM) and time-series models (Prophet, neural networks). Predictive maintenance combines classic anomaly detection with deep learning on sensor streams. Computer vision applications, such as ANPR, driver behaviour and vehicle class detection, we build on PyTorch and TensorFlow with optimised models for edge deployment.

For data streams we work with Kafka, Redis Streams or cloud-native equivalents; for geospatial processing with PostGIS, H3 (Uber's hexagonal indexing) and GeoPandas. Standard protocols such as GTFS-RT, NDOV feeds, OCPP for charging infrastructure and RDW APIs are not avoided but embraced: interoperability with existing mobility systems is a core principle.

We choose technology based on proven results in the mobility sector, not on hype. Where a linear regression suffices, we do not build a deep neural network. Where a simple heuristic works as well as a reinforcement learning agent that requires ten times the maintenance, we choose the heuristic.

Python OR-Tools XGBoost PyTorch TensorFlow PostGIS H3 Kafka FastAPI Docker Kubernetes GTFS-RT OCPP RDW API 9292 data

Why Appfront for mobility AI

Mobility is a sector with its own jargon, its own standards and its own operational realities. Our approach takes these into account.

Mobility domain expertise

We understand the difference between a fixed timetable and a dynamic timetable, between NDOV and GTFS-RT, between OCPP 1.6 and 2.0. We translate that knowledge into models that work in your operation, not academic demos nobody uses in practice.

Integration with your stack

Whether you work with Transpas, Adaption, your own TMS, telematics from Geotab or Webfleet, a charging infrastructure back end or a Translink integration, we build integrations that fit into your existing workflow. No parallel system requiring duplicate data entry.

From POC to production

Many AI projects stall after the prototype. We guide the entire journey: from data assessment through proof of concept to production deployment and ongoing maintenance. One partner across the whole chain, with no handover friction between consultant and builder.

Get in touch Microservices architecture Smart API integrations Building a mobile app

Mobility data, privacy and compliance

Vehicle tracking, number plate recognition and driver monitoring directly involve personal data. We build mobility AI that complies with the GDPR, sector-specific guidelines and realistic security requirements.

GDPR-compliant processing

Vehicle location data, number plate images and driver profiles are only processed for their original purpose. We implement pseudonymisation, retention policies and data minimisation. Where possible, training data is anonymised before models learn from it.

Responsible driver monitoring

Camera AI aimed at drivers is a sensitive subject. We build solutions that respect data minimisation: storing only events rather than continuous footage, clear agreements about who sees what, and works council involvement where needed. We don't sell surveillance; we improve safety.

Hosting and data sovereignty

Operational mobility data does not leave the Netherlands or the EU without your explicit choice. We host with Dutch cloud providers or in your own environment. No unnecessary transfer to US clouds for sensitive passenger or driver data.

Resilient under disruption

Mobility systems run 24/7, and a crash at rush hour is unacceptable. We design for failover, graceful degradation and clear alerting. An AI component that fails must not halt scheduling; there is always a fallback to rule-based planning.

IT operations optimisation High-availability systems Platform engineering Information security policy

Frequently asked questions about AI in mobility

What are the most practical AI applications for a fleet owner?

For most fleets we start with route optimisation (VRP) or predictive maintenance. Both deliver measurable improvements within a few months without you having to overhaul your existing processes entirely. EV charging optimisation becomes interesting once a substantial part of the fleet is electric.

What data is needed for a good AI solution in mobility?

At a minimum, historical trip records or telematics data. For MaaS and public transport, GTFS or NDOV data is essential; for charging infrastructure, OCPP logs; for fleets, sensor and maintenance data. The more high-quality historical data, the more accurate the model. We help you inventory and make available the sources you have.

Do you work with standard mobility protocols such as GTFS-RT, OCPP and NDOV?

Yes. Our engineers have experience with GTFS and GTFS-RT for public transport, OCPP for charging infrastructure (1.6 and 2.0), and NDOV feeds for travel information. We don't build proprietary silos; interoperability with existing mobility standards is a core principle.

Can a VRP engine really plan faster than an experienced planner?

For smaller problems (up to a few dozen stops), an experienced planner is often just as good. With hundreds to thousands of stops across multiple vehicles, with time windows and restrictions, a VRP solver wins by a wide margin, not because planners fail, but because the search space simply becomes too large for people. The planner remains crucial for exceptions and strategic decisions.

How do you handle privacy with driver monitoring and number plate recognition?

Strictly compliant with the GDPR, with data minimisation as the starting point. For driver monitoring, we store only aggregated scores and high-risk events, never continuous footage. For ANPR, we link number plates only to a legitimate access or enforcement basis. Clear DPIAs (data protection impact assessments) and processor agreements are standard.

What determines the investment in a mobility AI project?

The complexity of the use case, the availability and quality of data, the number of integrations (TMS, telematics, planning, OCPP) and the required scale. A PoC for a single line or route cluster is a different project from a fleet-wide predictive maintenance platform. We always start with a focused PoC to validate feasibility before a larger investment follows.

How long does it take for an AI solution to become operational?

A proof of concept is usually ready within a few weeks. The time to production depends on integration requirements, data quality and the scale of your operation. We work in iterative sprints, so you see results along the way and can steer the project as it progresses.

Ready to use AI for your mobility operation?

Discuss your case with us. We will analyse where AI delivers the most value for your fleet, line, charging infrastructure or urban mobility — with no obligation and no commitment.

Schedule a conversation