How AI can transform urban transport into sustainable MaaS systems

Artificial intelligence is emerging as a transformative tool for reshaping urban transportation, with new research highlighting both its potential and limitations. A team of scholars has conducted a systematic review of AI’s role in advancing sustainable and intelligent mobility solutions.

Their article, “Artificial Intelligence Applications for Smart and Sustainable Mobility as a Service Concept: A Systematic Literature Review,” published in Future Transportation, examines how machine learning, deep learning, and big data techniques are being applied to the concept of Mobility as a Service (MaaS). By reviewing 49 peer-reviewed studies published between 2020 and 2024, the authors provide one of the most comprehensive analyses to date of how AI is shaping the future of mobility ecosystems.

How is AI being applied across different MaaS integration levels?

The review structures AI applications around four levels of MaaS integration: basic, intermediate, advanced, and full.

At the basic level, machine learning algorithms and optimization models dominate. These tools are used for route planning, ride-sharing optimization, and the integration of basic multimodal options. Their focus is on improving efficiency while reducing costs for providers and users.

The intermediate level introduces greater sophistication, with demand forecasting, booking systems, and pricing optimization supported by long short-term memory networks, convolutional models, and ensemble learning. These approaches aim to align supply with fluctuating demand patterns while offering more accurate pricing strategies.

At the advanced level, deep learning and reinforcement learning are applied for real-time personalization, dynamic pricing, and multimodal coordination. These models enable transport systems to adapt quickly to changing conditions, providing personalized travel options and supporting seamless coordination across different modes of transport.

The full integration level extends beyond operational efficiency, incorporating blockchain, federated learning, and privacy-preserving AI. These technologies address security, transparency, and data protection while promoting equity and sustainability. The review emphasizes that full integration is still in its early stages, with practical implementation limited to pilot projects and research prototypes.

What challenges and gaps remain in AI-driven MaaS?

The study also sheds light on the significant barriers to the widespread adoption of AI in mobility ecosystems. A major concern is governance and inclusivity. Most existing MaaS–AI applications are concentrated in developed economies, leaving emerging markets underexplored. Without greater geographic diversity in research and implementation, global inequalities in mobility innovation are likely to persist.

Data governance also presents a challenge. AI systems require vast amounts of data, raising issues around privacy, ownership, and consent. While techniques like federated learning offer potential solutions, regulatory frameworks remain underdeveloped. Inconsistent standards risk undermining trust among users and service providers.

Another gap lies in explainability. Many advanced AI models operate as “black boxes,” producing outputs that are difficult to interpret. This lack of transparency can limit trust, particularly when systems are tasked with sensitive decisions such as pricing, service prioritization, or accessibility.

Finally, sustainability and equity are recurring concerns. While AI can improve efficiency and reduce environmental impacts, there is limited research on how these technologies can be deployed to serve disadvantaged groups or reduce carbon emissions in a socially just way. The review calls for a stronger focus on designing MaaS systems that are not only technologically advanced but also socially inclusive and environmentally sustainable.

What directions should future research and policy take?

The study recommends several directions for future research and policy. A priority is advancing explainable AI methods that make decision-making transparent and understandable to stakeholders. This will be essential for building trust and ensuring accountability in MaaS applications.

The study also emphasizes the importance of next-generation computing technologies. Edge computing, which processes data closer to the source, could reduce latency and improve real-time decision-making. Quantum computing, still at a nascent stage, offers the potential to solve complex optimization problems that are beyond the capacity of current systems.

From a policy perspective, the review calls for comprehensive governance frameworks to regulate data use, protect privacy, and ensure fair access. These frameworks must be adaptive, keeping pace with rapidly evolving AI technologies while safeguarding user rights.

The authors stress the need for cross-sector collaboration. Building sustainable MaaS ecosystems will require partnerships between governments, private companies, and research institutions. Pilot projects in diverse contexts, particularly in emerging economies, could generate critical insights into how AI-driven MaaS can be adapted to different infrastructural and cultural environments.