AI in transportation: 8 use cases, real-life examples, technologies & challenges

AI in transportation: 8 use cases, real-life examples, technologies & challenges

December 20, 2023

Viktoriya Shamkina

by Viktoriya Shamkina,

ML Engineer

Artificial intelligence in transportation can enhance the driving experience, optimize logistics operations, and ensure better management of vehicles, roads, and infrastructures.

Find out how AI experts can help transportation companies apply specific embodiments of this technology to a variety of use cases and address potential adoption challenges.

Table of contents

AI in transportation: market trends & stats

CAGR of the smart transportation market from 2023 to 2028

MarketsandMarkets

the value that autonomous driving is expected to create by 2035

McKinsey

the annual economic value AI in transportation can generate in China alone by 2030

McKinsey

AI use cases in the transportation industry

Transportation companies can implement artificial intelligence in a wide range of tools, corporate functions, and business scenarios. Here are some key applications of AI and related payoffs.

Advanced driver-assistance systems

Many car manufacturers have long started to implement semi-autonomous driving features in their vehicles, such as advanced driver-assistance systems (ADAS), to help perform parking procedures, ensure control of the vehicle in bad weather conditions, and avoid collisions. ADAS solutions rely on AI-powered cameras and sensors designed to identify vehicles, obstacles, pedestrians, or passengers’ facial expressions through computer vision, alert drivers, and even trigger autonomous actions to prevent human errors.
Benefits
AI-based ADAS such as adaptive cruise control, forward-collision warning (FCW), automotive night vision, and traffic sign recognition increase safety for both drivers and pedestrians in the vicinity.

Personal assistants

A different, more interactive way to provide drivers with AI-powered assistance involves the adoption of voice-based devices. These tools tap into natural language processing to understand users' requests and perform a variety of tasks, such as initiating a call, switching radio stations, or providing information on vehicle conditions.
Benefits
Personal assistants help minimize distracting manual interactions with in-vehicle infotainment systems, enhancing the driving experience and improving safety.

Self-driving vehicles

Autonomous vehicles represent the future of ADAS systems, as they rely on AI to completely automate the driving experience. Most embodiments of this technology are still within the realm of prototyping and experimentation. However, companies like Tesla have pioneered self-driving cars and other autonomous vehicles in several scenarios with promising results. In this regard, it's also worth mentioning self-driving taxis, truck platooning systems (the coordinated movement of multiple trucks at close range), and autonomous navigation for container vessels via video object recognition and lidar technologies.
Benefits
Self-driving vehicles in logistics can help mitigate shipping costs. McKinsey, for instance, estimated a 45% reduction in operating costs and savings of up to $125 billion by deploying autonomous trucks at scale in the United States.

Fleet management & route optimization

AI-based solutions can help logistics companies to optimize the supply chain by rationalizing and coordinating fleets of vehicles, ships, and planes. Their operation is based on a blend of GPS, sensors, computer vision-powered cameras, and other interconnected IoT devices deployed to gather data regarding weather, traffic, blockages, or accidents. These tools are then combined with AI-based analytical systems to process such information, identify recurring traffic patterns via machine learning algorithms, and turn data into valuable route recommendations or forecast potential traffic congestion.
Benefits
Optimized routing and fleet management ensures faster deliveries and reduces fuel consumption, resulting in cost savings and more sustainable transportation.

Traffic management & road monitoring

A traffic management system’s operation can be compared to that of fleet management solutions. The idea is to deploy a widespread network of sensors and cameras to oversee road and traffic conditions, identify car crashes via computer vision, and make traffic predictions. This allows authorities to intervene promptly in the event of traffic accidents, speed up road repair and maintenance operations, and optimize traffic light switching based on vehicle density.
Benefits
AI-powered solutions can help reduce traffic jams, waiting times, and carbon emissions while improving road safety and maintenance.

Automatic number plate recognition

These solutions encompass HD cameras mounted on street poles, infrared sensors to ensure 24/7 monitoring, and image processing software to identify vehicle registration plates via OCR (optical character recognition). ANPR systems are useful in a variety of management and security tasks, including journey time analysis to enhance road planning, the identification of vehicles violating road rules, and electronic payments for cashless tolling lanes.
Benefits
ANPR technology facilitates traffic monitoring, law enforcement, and toll management, proving an essential tool for traffic police and other public authorities.

Smart parking

Artificial intelligence facilitates the search for parking with the help of cameras and computer vision, which can be deployed both in indoor car parks and outdoor urban areas. These solutions can be helpful in several ways, including vehicle counting and free slot detection connected with parking availability indicators, license plate matching to detect unauthorized parking, and time tracking for easier ticket billing and payment. AI-powered cameras are also used to identify suspicious activities for parking lot security.
Benefits
Smart parking systems help streamline traffic flow in city centers, mitigate parking queues, and enhance safety in public spaces.

Predictive maintenance

This approach relies on machine learning-based anomaly detection to predict failures before they actually occur. A machine learning system can be trained to understand the normal operation of vehicles and infrastructures by "feeding" it with data relating to their standard performance and collected via sensors. Once the system has learned to recognize the ideal operating patterns of the mechanical and electronic components powering a vehicle or an entire station, it can also detect any outlier that might be a sign of impending breakdown.

Benefits
Private and public transport organizations leveraging predictive maintenance can improve vehicle reliability, reduce maintenance costs, speed up repair procedures, and cut fleet reserves deployed to avoid service interruptions.

Have an AI project in mind?

Turn to Itransition

Real-life examples of AI in transportation

Subaru’s driver monitoring system

    Subaru’s driver monitoring system

    Like many other automotive companies adopting AI, Subaru equips its vehicles with a driver monitoring system powered by computer vision, which provides a variety of support features to enhance safety and comfort. The solution can scan and recognize the driver's face and alert them when detecting signs of drowsiness or distraction. It can also adjust the seat position and interface settings autonomously based on the current driver.

    Waymo’s robotaxi

      Waymo’s robotaxi

      California-based development company Waymo opened its robotaxi ride-hailing service to US customers in 2019, representing the first commercial service to operate without an onboard backup driver. The self-driving system has been tested through 10 million miles of simulation encompassing complex and adversarial scenarios. However, the service is still in an early program stage, as the driverless rides only happen in a controlled geofenced environment.

      Heathrow Airport’s monitoring solution

        Heathrow Airport’s monitoring solution

        To improve air traffic control and deal with the infamous London weather, Heathrow Airport has implemented Aimee, an AI solution powered by neural networks. This system designed for the air transport industry can process data collected via high-definition cameras and help controllers supervise arrivals and departures in low visibility scenarios. It also facilitates controller-pilot communication by handling departure clearance requests via natural language processing. Once deployed at full capacity, this tool should enhance the airport's landing capacity by 20%, reducing the risk of flight delays.
        Heathrow Airport’s monitoring solution

        Image title: Aimee’s advanced object detection
        Image source: searidgetech.com — Aimee: AI framework

        Surtrac traffic management system

          Surtrac traffic management system

          An AI-based monitoring system Surtrac, developed by Rapid Flow Technologies, has been deployed in Pittsburgh, US, to collect data with smart cameras, adjust traffic lights in real time, and thus facilitate traffic flow. The monitoring devices in each intersection operate independently, handling their own local traffic and replanning second by second. The solution has led to a 25% reduction in travel times, 30% reduction in the number of stops, and a 20% cut in emissions.

          SNCF’s predictive maintenance system

            SNCF’s predictive maintenance system

            France's national railway company SNCF adopted a predictive analytics solution to spot potential asset malfunctions (including pantographs at risk of wear), anticipate maintenance needs, and thus ensure power supply to its trains across 32,000 km of network. According to the company, predictive maintenance will also lead to a 30% reduction in accidents related to train switches, ensuring superior economic performance, service reliability, and passenger safety.

            Related technologies used in transportation

            Transportation and logistics organizations can count on a vast toolkit of AI-related technologies which can be combined depending on the field of application.

            Machine learning

            A branch of AI focusing on self-learning algorithms that can improve their performance through experience. It can be divided into supervised and unsupervised learning based on the approach to algorithm training. Applications of ML include:

            Computer vision

            to acquire visual inputs, identify people or objects, and alert the driver or trigger appropriate actions

            Example of use
            Enabling ADAS and automated visual inspection

            Natural language processing

            to replicate human communication and thus unlock realistic driver-machine interactions

            Example of use
            Powering voice-based driving assistants

            Data mining

            to aggregate big data, identify patterns, outliers, or dependencies between variables, and extract valuable insights

            Example of use
            Detecting anomalies to predict engine failures

            Internet of Things

            Extensive networks of interconnected devices (cameras, infrared sensors, etc.) collecting information from vehicles and infrastructures through the internet, Bluetooth, or other communications technologies.

            Example of use

            Gathering traffic data to optimize routes

            Itransition will guide you through AI implementation

            Get in touch

            Challenges & tips for adopting AI in transportation

            Despite the multiple perks of adopting AI in the transportation sector, companies willing to deploy this technology can face a variety of technical challenges.

            Issue

            Recommendation

            Integration

            The multi-layered, interconnected architecture of a typical AI solution for transportation implies that its components, including IoT devices and data analytics software, should be able to exchange data. However, such elements can use different communication protocols or technologies and handle multiple data formats, including streams of real-time data. So when poorly integrated, AI systems will base their analyses on fragmented sources and inconsistent or outdated data, delivering inaccurate predictions.

            The multi-layered, interconnected architecture of a typical AI solution for transportation implies that its components, including IoT devices and data analytics software, should be able to exchange data. However, such elements can use different communication protocols or technologies and handle multiple data formats, including streams of real-time data. So when poorly integrated, AI systems will base their analyses on fragmented sources and inconsistent or outdated data, delivering inaccurate predictions.

            Ensure communication between the components of your AI solution by configuring application programming interfaces (APIs). You can leverage tools from major cloud providers, such as Amazon API Gateway or Azure API Management, to streamline this task. In some situations, however, a middleware architecture is required, such as an ESB, to convert different protocols. With the help of cloud data integration tools like AWS Glue or Azure Data Factory, you can also integrate heterogeneous data from multiple sources via ETL processes and consolidate it into a unified data storage. This may be, for instance, a time-series database, a NoSQL database, or a data lake.

            Connectivity

            The operation of transportation-oriented AI systems is, by their very nature, geographically distributed over long distances. This means the IoT sensors collecting and transmitting real-time information to a data analytics platform should rely on stable networks, even during potential load spikes. Otherwise, lost connection and high latency can result in inaccurate analysis.

            The operation of transportation-oriented AI systems is, by their very nature, geographically distributed over long distances. This means the IoT sensors collecting and transmitting real-time information to a data analytics platform should rely on stable networks, even during potential load spikes. Otherwise, lost connection and high latency can result in inaccurate analysis.

            Consider adopting IoT sensors loosely coupled to other systems based on a publish-subscribe pattern, as they exchange data in real time when a stable connection is available. If the vehicle enters a blind spot (such as a tunnel) the devices memorize new data points in an offline message queue and transmit them as the connection is restored. You can also leverage edge computing to shift some of the computing workload from centralized data centers to the devices, minimizing network dependency and latency risk. Finally, you can use bandwidth-efficient communication protocols, including the MQTT.

            Data analysis

            The real-time analysis of massive data volumes collected from vast transport networks and fleets of vehicles requires significant computing power. Furthermore, the algorithms used to process this data require training on massive data sets to produce an AI model that will deliver accurate analyses and forecasts. Even after training, however, the model can be less reliable than expected. This could happen due to overfitting, if the model was overtrained on a certain data set and performed poorly with other data, or due to a model drift when its predictive power degrades because of progressive changes in input variables and their relationships.

            The real-time analysis of massive data volumes collected from vast transport networks and fleets of vehicles requires significant computing power. Furthermore, the algorithms used to process this data require training on massive data sets to produce an AI model that will deliver accurate analyses and forecasts. Even after training, however, the model can be less reliable than expected. This could happen due to overfitting, if the model was overtrained on a certain data set and performed poorly with other data, or due to a model drift when its predictive power degrades because of progressive changes in input variables and their relationships.

            ML services from top cloud providers, including Amazon SageMaker and Azure Machine Learning, deliver built-in algorithms, pre-trained ML models, and scalable processing power to speed up your solution’s deployment and complement your in-house computing resources. You can also adopt a masterless cluster architecture, which enables you to scale up and down the processing nodes serving your network of IoT devices based on the current workload. Regarding model reliability, a common practice involves splitting modeling into training, validation, and test sets to mitigate overfitting. As for model drift, you can keep track of it by regularly monitoring metrics like the Population Stability Index and perform multiple retraining iterations to fine-tune the model with new data.

            Reimagine transportation with AI

            In recent years, according to McKinsey, transportation has ranked among the industries benefiting from the highest adoption rate of AI technology. This should come as no surprise, as advancements in AI have proved capable of making us travel and move goods faster, safer, and cleaner than ever before. However, implementation efforts should always be carried out with the complex and sometimes inscrutable nature of artificial intelligence in mind, especially when it’s deployed in wide, sprawling ecosystems such as transport networks. To streamline the adoption of AI while overcoming its potential drawbacks, consider relying on Itransition’s team of experienced developers and consultants.

            Artificial intelligence consulting

            Service

            Artificial intelligence consulting

            Explore our AI consulting services, along with related technologies, use cases by industry, and implementation best practices.

            AI in the automotive industry: use cases, success stories & adoption guidelines

            Insights

            AI in the automotive industry: use cases, success stories & adoption guidelines

            Explore key use cases, payoffs, and real-life examples of AI in the automotive industry, along with common adoption challenges and tips to address them.

            Machine learning in logistics: technology breakdown & 10 use cases

            Insights

            Machine learning in logistics: technology breakdown & 10 use cases

            Find out how machine learning is transforming logistics and supply chain, including its top use cases, benefits, technologies used, and implementation tips.

            RPA in the automotive industry: key use cases, examples & best practices

            Insights

            RPA in the automotive industry: key use cases, examples & best practices

            Learn how to improve the efficiency of your automotive business processes and reduce operational costs with robotic process automation implementation.

            Automotive ERP: key functionality, integrations & software options

            Service

            Automotive ERP: key functionality, integrations & software options

            Itransition presents an overview of ERP solutions for the automotive industry and their benefits, features, key integrations, best software & selection factors.

            Cloud business intelligence system for vehicle manufacturers

            Case study

            Cloud business intelligence system for vehicle manufacturers

            Find out how Itransition migrated a BI suite to the cloud and delivered brand-new cloud business intelligence tools for the automotive industry.