In the transportation sector, artificial intelligence (AI) techniques are increasingly being utilized to overcome several challenges, including environmental degradation, safety concerns, carbon dioxide emissions, and travel demand, in a more effective and efficient manner. The availability of a substantial amount of qualitative and quantitative data in this digital era has made the use of AI more plausible in transportation. This article discusses the importance and application of AI in transportation.
Image credit: Creative Stock Studio/Shutterstock
Importance of AI in Transportation
Transportation problems can translate into major challenges when the behavior of the user and system becomes difficult to predict and model the travel patterns. These challenges primarily emerge due to the steady growth in urban and rural traffic as a result of a growing population, specifically in developing countries.
AI techniques can play a crucial role in addressing these challenges in transportation systems as they can model a relationship between the effect and cause of various real-life scenarios by combining the available data with probabilities and assumptions, leading to a better analysis compared to conventional techniques.
Several studies have been performed to investigate the use of AI techniques, such as artificial neural networks (ANN), fuzzy logic model (FLM), bee colony optimization (BCO), ant colony optimizer (ACO), artificial immune system (AIS), simulated annealing (SA), and genetic algorithms (GA), to realize a more reliable and cost-effective transportation system with less impact on the environment and people.
A good understanding of the relationships between data and AI and between transportation system variables and characteristics is necessary for the successful application of AI in transport. Transport authorities can use these technologies to effectively relieve congestion, ensure more reliable travel, and improve the productivity and economics of their crucial assets.
In transportation, AI applications are being developed and implemented in corporate decision-making, managing, and planning, improving public transport, and in autonomous and connected vehicles. AI is employed to better utilize the precise detection and prediction models for improved forecasting of traffic conditions, volume, and incidents.
Improvements in public transport, which is a sustainable mode of mobility, using AI can contribute to sustainability, while AI in autonomous/connected vehicles can reduce the number of accidents on roads, specifically highways.
Major Applications of AI
Advanced Safety Systems: Advanced object recognition and detection algorithms, coupled with AI-powered collision avoidance systems, can significantly improve pedestrian and vehicle safety. Additionally, AI can be employed to identify potential maintenance issues in infrastructure and vehicles, which is necessary for proactive maintenance that minimizes the risk of accidents due to mechanical failures.
Sustainable Solutions for Transportation: AI can effectively contribute to sustainable transportation solution development. Machine learning (ML) algorithms can promote eco-friendly driving behaviors, facilitate the integration of autonomous and electric vehicles into transportation networks, and optimize energy consumption. Moreover, AI-powered routing and logistics systems can also be used to select eco-friendly options to promote sustainable mobility and reduce carbon emissions.
Personalized Mobility Services: AI techniques can play a crucial role in offering personalized mobility services/customized transportation solutions, such as multimodal journey planning and on-demand ride-sharing, to users by analyzing individual travel patterns, preferences, and real-time data to improve the overall travel experience of users, enhance accessibility to public transportation systems, and reduce the dependence on private vehicles.
Intelligent Connectivity and Infrastructure: AI can enable intelligent infrastructure systems that can communicate with vehicles and provide information in real time. For instance, smart traffic lights can adjust signal timings dynamically based on traffic conditions to optimize the traffic flow. Additionally, AI-powered vehicle-to-vehicle (V2V) communication and vehicle-to-infrastructure (V2I) systems can enhance efficiency, coordination, and safety on the roads.
Enhanced Traffic Management and Prediction: AI methods can be utilized to effectively manage and predict traffic patterns. These techniques provide up-to-date and accurate traffic predictions by analyzing substantial amounts of real-time data obtained from different sources, including social media, global positioning systems (GPS), and sensors.
Precise traffic predictions can enable better traffic management, including optimized traffic signal control, congestion prevention, and dynamic routing, leading to more efficient and streamlined transportation systems.
Fully Autonomous Vehicles: Fully autonomous vehicle development has gained significant attention in recent years. AI algorithms, coupled with deep learning techniques and advanced sensor technologies, can enable vehicles to handle every driving task without human intervention and navigate complex traffic scenarios, leading to enhanced road safety and reduced traffic congestion.
AI Techniques in Transportation
Several AI techniques can transform the traffic sensors on the road into smart agents that automatically detect accidents and predict future traffic conditions. For instance, ANNs, including supervised methods such as radial basis network (RBN) and probabilistic neural network (PNN), and unsupervised methods, such as cluster analysis, can be utilized for predicting traffic conditions, traffic incident detection, and road planning.
Raster algorithms, such as GA, can be used in urban design networks as they can effectively solve complex optimization problems. The spatial relationship between land-use planning and transportation can be modeled using a parallel neural network system.
Although both GA and ANN can be employed to model the safety management plan of cities, ANN displays better performance compared to GA. An ant colony algorithm can be used to design an optimal vehicle path and address the vehicle routing problem.
AIS and GA can be utilized to precisely identify the most suitable modification for an existing road network structure. The real-time optimization of traffic control policies embedded in large-scale intelligent transport systems (ITS) can be realized using deep reinforcement learning.
Fuzzy methods and GA can be utilized to automatically control traffic signal systems at intersections. ANNs can also be used for future traffic congestion prediction and traffic signal control based on the microscopic simulated data. Incident detection on the road network can be achieved using the classification neural network algorithm.
Challenges of Using AI
The development of AI-based solutions for an efficient transportation system is extremely difficult, which has limited the application of AI to specific ITS applications such as data analysis and future mobility predictions.
Additionally, the on-time predictions and accuracy of AI techniques using data collected through conventional methods, such as actuators, sensors, and loop detectors, are unreliable. Thus, transitioning from classical data collection methods to advanced AI-based technologies is crucial for higher AI accuracy and more reliable predictions, as these technologies can provide easily deployable and novel data mining tools.
In transportation, forecasting long and short-term traffic flow under adverse weather conditions and unexpected events is typically challenging. Existing AI techniques cannot accurately forecast the traffic flow under such conditions. Thus, incident and weather-responsive algorithms and prediction schemes must be developed to obtain accurate forecasts.
Moreover, the high cost of developing and maintaining smart technologies owing to their complexity, lack of transparency and privacy of AI-driven technologies, and the computational complexity of AI techniques are the other general limitations of AI in transportation.
Recent Studies
Although traffic flow forecasting is a critical problem in the ITS, the complex dynamic spatial-temporal dependency and the correlation between different locations on the road increase the challenges to ensure security in transportation.
Most of the existing forecasting frames use graph convolution to model the dynamic spatial-temporal correlation of vehicle transportation data without considering the semantic similarity between nodes, resulting in lower accuracy. Additionally, the traffic data cannot be captured easily as it does not strictly follow periodicity.
In a paper recently published in the journal IEEE Transactions on Intelligent Transportation Systems, researchers proposed a conditional random field augmented spatial-temporal graph convolutional network (CRFAST-GCN), a multi-branch spatial-temporal attention GCN, to address the existing challenges.
Initially, the multi-scale long- and short-term dependencies were captured through three identical branches, and then CRFAST-GCN was introduced to capture the semantic similarity. Finally, the attention mechanism was exploited to capture the periodicity.
Two real-world datasets were used for model evaluation. Performance analysis demonstrated that the proposed CRFAST-GCN can effectively handle the complex spatial-temporal dynamics and achieve 50% improvement over the baselines.
References and Further Reading
Diao, C., Zhang, D., Liang, W., Li, K. -C., Hong, Y., Gaudiot, J. -L. (2023). A Novel Spatial-Temporal Multi-Scale Alignment Graph Neural Network Security Model for Vehicles Prediction. IEEE Transactions on Intelligent Transportation Systems, 24, 1, 904-914. https://doi.org/10.1109/TITS.2022.3140229.
Abduljabbar, R., Dia, H., Liyanage, S., Bagloee, S. A. (2018). Applications of Artificial Intelligence in Transport: An Overview. Sustainability, 11(1), 189. https://doi.org/10.3390/su11010189
Bharadiya, J. P. (2023). Artificial Intelligence in Transportation Systems A Critical Review. American Journal of Computing and Engineering, 6, 34-45. https://doi.org/10.47672/ajce.1487.