Multimodal urban mobility and multilayer transport networks

Introduction

• Most trips are short and suited to walking, medium-distance trips to bicycle or bus, and the smallest share of long trips to metro/rail or car.The review argues that large cities need a balanced mix of modes reflecting how far residents travel.Source: Introduction · p. 2
• Multimodal mobility combines transport modes in one trip, naturally or by design, with stations and other interfaces facilitating transfers between services.Combining modes can offer their benefits while avoiding their weaknesses.Source: Introduction · p. 2

Potential and challenges of multimodal mobility

• Multimodal transport can support sustainability, but modes have distinct infrastructure and operating schedules, with differences in density and connectivity.The review describes a modern transport hierarchy that prioritizes walking, cycling, and transit over other forms of transport.Source: Potential and challenges of multimodal… · p. 2, Figure 1 · p. 3
• Travellers must process substantial information to choose journeys, while providers face synchronization challenges across agencies and modes.These difficulties have also created opportunities for mobility applications and research.Source: Potential and challenges of multimodal… · p. 3

Aim and structure of this Review

• The review surveys multimodal mobility and transport infrastructures as multilayer networks, focusing on recent complexity-science approaches and urban research.It uses the primal network approach, representing infrastructure as links and intersections or stops as nodes, and covers network theory, infrastructures, mobility, datasets, tools, and open questions.Source: Aim and structure of this Review · pp. 3–4

• A network or graph consists of nodes and edges, and transport links can connect consecutive stops in L-space or stops on the same route in P-space.The passage notes that the conventional L- and P-space labels should in principle be exchanged because they refer to logical and physical networks, respectively.Source: §1 · p. 4
• Spatial-network weights can encode connection strength, often as the reciprocal of distance or travel time between nodes.The passage gives the forms w_ij = 1/d_ij and w_ij = 1/t_ij.Source: §1 · p. 4
• A multilayer transport network assigns each mode its own adjacency matrix while labeling nodes in the same order across layers.The layers can represent, for example, bus, underground, and urban railway networks.Source: §1 · pp. 4–5
• Inter-layer links can encode the cost, distance, or time required to switch modes at a location, complementing within-layer connections.They may be omitted in measures focused on layer diversity or correlations.Source: §1 · p. 5
• Transport multilayer models have a history in traffic assignment and route choice; their layers are often deliberately distinct, which can make systems fragile to infrastructure failures.The passage describes transport layers as engineered for efficiency rather than redundancy and notes their sensitivity to disruptions.Source: §1 · p. 5

2. Multimodal infrastructures

• Multimodal infrastructures are modeled as layers of mode-specific nodes and links, while their design poses integration challenges across streets, transit, walking, and cycling.Compared with single-layer transport design, multimodal network design has received less study.Source: §2 · p. 6, §2.1 · p. 6

2.1. Modeling multimodal infrastructures

• Congestion-aware multimodal models have informed Santiago metro planning, while simulations found suitable park-and-ride controls can prevent city-center traffic jams.The Santiago application included 13 user classes, 11 transport modes, and 450 zones; park-and-ride controls combine site costs, parking availability, metro fares, and service frequency.Source: §2.1 · pp. 6–7
• Spatially coupled-layer models show that faster-layer speeds can concentrate trips and congestion at transfer nodes, while Randstad transit-layer betweenness tracked passenger flows.The Randstad model combined street, public transport, and land-use layers.Source: §2.1 · pp. 7–8
• Line-level layers capture transfer timing, whereas mode-level superlayers support study of system interdependence and resilience; bus hubs can have high overlap but low multilayer activity.Bicycle-network growth studies report initially decreasing returns until a critical threshold, while existing multilayer growth models omit spatial constraints; the overlap census classifies node reachability across mode subsets and identified six city-profile clusters.Source: §2.1 · pp. 8–9, §2.2 · p. 10

2.2. Characterizing multimodal infrastructures

• Spatial outreach measures reachable-node distances within a travel-cost threshold; increasing metro speed can improve outreach near interchange nodes and increase access to distant areas.Transit-and-pedestrian isochrones have also been used to score public-transport service and area accessibility.Source: §2.2 · p. 11
• Over 40% of Madrid trips use at least one metro link, despite 241 metro nodes versus 4,590 bus nodes, illustrating the metro layer’s role in shortest paths.Betweenness counts shortest paths through a node; interdependence measures shortest paths using multiple layers.Source: §2.2 · p. 12
• As Kuala Lumpur’s rail network grew, multilayer shortest paths could decrease dramatically; subway links can also shift street centrality from avenues and bridges toward metro terminals.Interchanges enable new shortest paths, and theory identifies transfers from less efficient to more efficient layers as a driver of congestion dynamics.Source: §2.2 · pp. 12–13
• Random-walk coverage studies find multimodal systems more resilient to failures than single layers; Singapore’s first significant disruption occurred after 50% of edges were removed.Time-respecting routes account for timetable departures and arrivals, but a comprehensive algorithm for multiple modes in truly time-dependent networks is missing; walking transfer time and plausible mode sequences also matter.Source: §2.2 · pp. 10, 11, 13

3. Multimodal mobility

• Multilayer mobility research remains limited by difficulty collecting comprehensive cross-mode data, while much empirical work focuses on public transit and schedules rather than real-time conditions.The review centers on models and recent empirical findings based on passively collected data.Source: §3 · p. 14, §3.2 · p. 16

3.1. Modeling urban mobility

• The review focuses on mode choice and route assignment, where models often assume rational, homogeneous travelers with unlimited knowledge despite evidence of cognitive limits.It surveys random walks, travel-time minimization, and capacity-aware agent-based approaches; more realistic multilayer routing models remain a research need.Source: §3.1 · pp. 14–16
• Interconnected layers can improve random-walk exploration and resilience to random failures, while exploration depends on layer topology, interlayer connections, and strategy.Random walks model agents using local connectivity rather than global shortest-route knowledge.Source: §3.1 · pp. 14–15

3.2.1. Public transport systems dynamics

• In the UK, an average 23% of travel time is lost in connections on trips with more than one mode; across several UK urban transport systems, synchronization efficiency follows the same scaling relation with path length.Source: §3.2.1 · p. 17
• As planned metro speed increases in Riyadh, global congestion falls but congestion rises locally near key stations.The analysis evaluates the planned metro's effects as a function of the speed ratio between street and metro systems.Source: §3.2.1 · p. 19
• For multilayer transport networks, information needed to travel between points exceeds the identified 8-bit cognitive limit in 80% of cases, indicating navigational difficulty.In single-layer metro networks, the information requirement had an upper bound of about 8 bits, corresponding to approximately 250 route connections.Source: §3.2.1 · p. 19

3.2.2. Individual multimodal behaviour

• In Shenzhen, behavioral coupling between bus and metro layers is associated with long trips from single-mode nodes, but not with the layers' speed ratio.The study measures coupling from passenger trips rather than the network's multimodal shortest paths; passengers may choose unimodal routes even when congestion makes multimodal travel preferable.Source: §3.2.2 · p. 20, Figure 4 · p. 20
• Aggregated travel distances show a power-law tail, while single-mode distance distributions can be lognormal; combining modalities may generate the power law.A model with velocity changes along individual trajectories also produces a power-law tail, and travel speed increases with distance as v ∼ r^α, where α ≈ 0.5.Source: §3.2.2 · p. 21

Data

• OpenStreetMap supports building street, subway, and railway networks, including multiplex networks linked where transport modes intersect.Applications described include the Randstad region and fifteen cities at different development stages.Source: Data · p. 21
• Public-transport datasets capture schedules and network structure, from a temporal Great Britain network to GTFS feeds collected across 667 locations.A collection of 25 urban networks also includes pedestrian layers and distinguishes public-transport modes.Source: Data · p. 22
• The Helsinki dataset uses door-to-door travel times and provides matrices for 2013, 2015, and 2018, enabling comparisons across years.It includes walking, cycling, and public transportation options.Source: Data · p. 22
• GeoLife provides GPS trajectories for 178 users over more than four years, with 69 users labeling transport modes.The labeled trajectories have supported research on mobility behavior across modes.Source: Data · p. 22

Tools

• Available tools cover graph analysis, infrastructure extraction, public-transport feeds, mobility trajectories, and multilayer networks.Examples include OSMnx and Pyrosm for infrastructure, transitfeed and Peartree for GTFS, movingpandas and scikit-mobility for mobility, and muxViz and MAMMULT for multilayer analysis.Source: Tools · pp. 22–23
• Most multilayer analysis tools are general-purpose rather than designed specifically for multimodal transportation, motivating transport-specific open-source libraries.The reviewed tools are freely available and open to edits, collaboration, and improvement.Source: Tools · p. 24

5. Conclusions

• High-quality, fine-grained, appropriately anonymized mobility data remain difficult to access openly, and commercial silos can obstruct transparent research.The review calls for governments, academia, and industry to promote open datasets and tools.Source: §5 · p. 24
• Integrating shared mobility with public and private transport is a research priority for real-time, user-centered planning, forecasting, and service management.The review connects these frameworks with aims including increased safety and reduced congestion and emissions.Source: §5 · p. 24
• Research on multimodal mobility has benefited from interdisciplinary work, and the review expects that character to continue.It identifies urban planning, geoinformatics, computer science, and physics among the contributing fields, and anticipates deep learning may reveal new features of multimodal mobility.Source: §5 · pp. 24–25
• Understanding multimodal mobility and its infrastructure is important for sustainable urban transport, which relies on public transport modes.The review frames this understanding as one contribution toward addressing climate change associated in considerable part with car-centric transport monocultures.Source: §5 · p. 25