AI and IoT: The future of intelligent transportation systems

Traffic congestion can make roads unsafe and your commute expensive, considering fuel costs and person-hours wasted on the roads. Increasing the number of roads or widening existing roads may not be a viable solution, given infrastructure costs. In the future, the way to manage traffic congestion would be by an intelligent analysis of traffic data.

Machine learning (ML) is a data-driven solution that can take up this challenge by learning to extract patterns out of historical data subsequently to automate a system. Intelligent transportation System (ITS) uses ML to improve the transportation system’s safety, efficiency, and comfort.

Figure 1: Intelligent transportation

Why Intelligent Transportation:

• Traffic safety:
  

  Artificial intelligence (AI), in conjunction with the Internet of Things (IoT), can help governments enforce traffic management and reduce vehicle violations.

• Road condition analysis:
  

  Traffic analysis involves extensive analysis of data extracted from public transportation, cars, and pedestrians. The data is then streamed into mobile phones allowing drivers to plan their routes.

• Technology law enforcement:
  

  AI image recognition enables law enforcement to reduce traffic congestion, implement illegal bans, and reduce road safety accidents. Smart bans improve law enforcement accuracy.

ML for Intelligent Transportation

Data is the primary commodity harvested from ITS. ML has the inherent ability to discover knowledge from data. ML-enabled features like regression, clustering, classification, prediction and decision-making enhance ITS and serve as the building blocks of most IT applications.

ML pipeline

Figure 2 depicts the ML pipeline. This pipeline consists of several steps, such as data preprocessing, feature extraction, and modelling. The ML pipeline aims to model ITS elements that ITS tasks can utilise. For example, modelling vehicle mobility is useful for prediction tasks and classifying transportation infrastructure models from images can be used in perception.

Figure 2: ML Pipeline

ML driven perception in ITS

ML-driven perception for image processing is primarily presented through a vision-based lens. Since perception tasks apply to several ITS topics and scopes, the final ITS topic must be narrowed down considering the related work, the ML approaches used, and the role played by ML. The four major tasks covered are road, vehicles, users, and networking.

Table 1

ML driven prediction in ITS

ML approaches have achieved excellent performance on ITS prediction problems. Table 2 groups the ITS topics related to prediction tasks and also presents the ML approaches and the role of ML in each topic.

Table 2

ML driven management in ITS

This section introduces ML-driven ITS management from two perspectives: ITS infrastructure and resource management. Table 3 displays the related work.

Table 3

Conclusion

ITS combines rapidly evolving technologies into various platforms for a wide range of advanced applications. The timely acquisition, processing, and analysis of large volumes of data have become an essential cornerstone for several applications' effective deployment and run-time operation. Advances in ML are thus critical enabling technologies for driving an ITS revolution.