Intelligent Transportation Systems and Digital Transformation

Introduction and Context

Transportation engineering is being reshaped by the transition from isolated physical infrastructure to connected, data-driven and increasingly automated mobility systems. Roads, railways, vehicles, traffic signals, logistics platforms and travellers now generate continuous streams of operational data. Intelligent Transportation Systems (ITS) use these data, together with communications, computation, sensing and control, to improve safety, mobility, reliability, environmental performance and asset management.

Digital transformation is broader than installing sensors or replacing paper records with software. It involves redesigning transport processes, institutional responsibilities and decision-making around digital information. A mature digital transport organisation integrates data from planning, design, construction, operation and maintenance. It may use geographic information systems (GIS), building information modelling (BIM), Internet of Things (IoT) devices, artificial intelligence (AI), digital twins, automated data collection and connected or autonomous vehicles (CAVs).

ITS projects are socio-technical systems. Their performance depends not only on algorithms and equipment, but also on telecommunications, human behaviour, procurement, standards, institutional capacity, cybersecurity, privacy and public trust. Consequently, a technically impressive system can fail if data are incomplete, responsibilities are unclear or the operating agency cannot maintain the technology.

Learning Objectives

By the end of this chapter, the reader should be able to:

1. Explain the architecture and principal applications of ITS.

2. Select appropriate sensors, IoT devices, GIS and BIM techniques for transport purposes.

3. Distinguish digital models, digital shadows and digital twins.

4. Evaluate automated data collection, AI forecasting, optimisation and predictive maintenance.

5. Explain the role and limitations of connected and autonomous vehicles.

6. Design an evidence-based approach to cybersecurity, privacy and data governance.

7. Critically assess ITS alternatives using technical, economic, social and institutional criteria.

1. ITS as a Systems Engineering Discipline

1.1 Definition and functional architecture

ITS may be defined as the coordinated application of information, communication, sensing, positioning, control and automation technologies to transport systems. A useful functional architecture contains five layers:

• Physical layer: roads, bridges, vehicles, signals, detectors, depots and users.
• Sensing layer: cameras, radar, lidar, inductive loops, Bluetooth receivers, weather stations, weigh-in-motion systems and mobile-device observations.
• Communication layer: fibre, cellular networks, Wi-Fi, satellite links, dedicated short-range communication and other vehicle-to-everything (V2X) channels.
• Data and service layer: databases, GIS, cloud or edge computing, digital twins, analytics and application programming interfaces (APIs).
• Decision and control layer: signal timing, ramp metering, traveller information, incident management, fleet dispatch, maintenance intervention and automated vehicle control.

The layers should not be treated as independent. For example, adaptive signal control requires reliable detection, low-latency communications, a validated control algorithm, an appropriate signal controller and an operating agency able to intervene when the system behaves unexpectedly.

1.2 Common ITS services

Typical services include advanced traffic management, electronic tolling, incident detection, emergency-vehicle priority, public transport priority, automatic fare collection, traveller information, commercial vehicle management, smart parking, road-weather management and electronic work-zone management. Their objectives may conflict. A signal plan that improves bus reliability can increase delay for some private vehicles; a tolling scheme can reduce congestion but raise equity concerns. Evaluation must therefore consider the whole network and affected groups rather than a single performance indicator.

1.3 Performance measures

Useful measures include travel time, delay, travel-time reliability, throughput, queue length, crash frequency and severity, emissions, energy use, public transport punctuality, maintenance response time, data availability and system uptime. Reliability should not be confused with average speed. A route with a 30-minute mean journey may be less useful than one with a 32-minute mean but much smaller variation.

For a set of observed journey times, the sample mean is

x̄ = (1/n) Σ xi

and the sample standard deviation is

s = √[Σ(xi − x̄)²/(n − 1)].

Here, xi is the journey time for observation i and n is the number of observations. These statistics describe the observed sample, not necessarily the entire population. They are sensitive to outliers and should be supplemented by percentiles, such as the 95th-percentile journey time, when reliability is important. Time observations must use consistent units and comparable origins, destinations, time periods and traffic conditions.

2. Sensors, IoT and Automated Data Collection

2.1 Sensor selection

Sensor choice should follow the decision to be supported, not the availability of technology. Important selection criteria include accuracy, coverage, temporal resolution, latency, environmental robustness, calibration needs, power demand, installation cost, cybersecurity exposure and lifecycle cost.

Inductive loops detect changes in inductance caused by vehicles and can provide presence, occupancy and count data. They are mature and accurate but require pavement cutting and are vulnerable to construction damage. Video image detection can classify vehicles and identify incidents over a broad field of view, although performance may degrade because of darkness, rain, glare, occlusion or poor camera positioning. Radar performs relatively well in poor visibility and can estimate speed and range, while lidar provides detailed three-dimensional geometry but may have higher cost and sensitivity to weather and contamination. Bluetooth, Wi-Fi and automatic number-plate recognition can support travel-time estimation or origin–destination studies, but they raise privacy and representativeness issues.

No sensor measures reality perfectly. A detection system should be tested using a reference method. If a detector records TP true positives, FP false positives, FN missed events and TN true negatives, precision is

Precision = TP/(TP + FP),

whereas recall, or sensitivity, is

Recall = TP/(TP + FN).

Precision answers “How many detected events were real?” Recall answers “How many real events were detected?” A safety-critical incident system may prioritise recall, while a public information system may require high precision to avoid unnecessary alerts.

2.2 IoT architecture

An IoT transport installation generally includes a device, local processing, a gateway, a communications network, a platform and an application. Edge computing processes data close to the source. It can reduce latency and bandwidth use, preserve operation during temporary network loss and limit the transmission of identifiable raw data. Cloud computing offers scalable storage and analytics but introduces dependency on connectivity, service providers and configuration security.

A professional deployment requires device identity, secure boot or signed firmware, encryption in transit, key management, time synchronisation, remote health monitoring and a plan for replacement. Devices should report not only measurements but also quality flags, calibration status, battery condition and communication status. A database containing millions of unqualified observations can be less valuable than a smaller, well-documented dataset.

2.3 Data quality and fusion

Transport data commonly suffer from missingness, drift, duplication, inconsistent timestamps, changes in sampling rates and biased spatial coverage. Data fusion combines complementary sources, for example loop counts with probe-vehicle speeds and weather observations. Fusion should preserve provenance: the system should record where data came from, when they were collected, what transformations were applied and how uncertainty was estimated.

Imputation can be useful for analysis but must not silently convert missing data into observed facts. A model trained on data from normal weekdays may perform poorly during holidays, extreme weather, festivals or incidents. Validation should therefore include unusual but operationally important conditions.

3. GIS, BIM and Digital Twins

3.1 GIS for network intelligence

GIS links transport objects to location. It supports network inventory, accessibility analysis, crash mapping, corridor evaluation, route planning, land-use and transport interaction, emergency response and maintenance prioritisation. A transport GIS should distinguish geometric accuracy from attribute accuracy. A road may be located correctly while its speed limit, lane count or pavement condition is obsolete.

Network analysis represents links and nodes, with attributes such as length, capacity, restriction, travel time and direction. Routing algorithms can find minimum-cost paths, but the cost may be time, distance, toll, emissions or a weighted combination. The selected cost function reflects policy. A shortest-time route is not automatically the best route for heavy vehicles, pedestrians or emergency access.

3.2 BIM and the asset lifecycle

BIM is a structured digital representation of built assets and their relationships. In transport infrastructure it can support corridor coordination, clash detection, quantity information, construction sequencing, handover and maintenance. BIM becomes operationally useful when asset objects carry reliable identifiers, specifications, inspection history, warranty information and replacement requirements.

The value of BIM is often lost at handover because design and construction information is not mapped to the operator’s asset-management system. Information requirements should therefore be defined at procurement, including naming conventions, coordinate reference systems, classification, level of information need, exchange formats, cybersecurity controls and responsibility for updates.

3.3 Digital models, shadows and twins

A digital model is a representation that may not exchange data automatically with its physical counterpart. A digital shadow is updated by the physical asset, but information may flow mainly in one direction. A digital twin is a dynamic, connected representation in which data flow between the physical and digital systems supports monitoring, analysis or control. The term is sometimes used loosely; a three-dimensional visualisation alone is not a digital twin.

A useful transport digital twin includes:

• a stable identity for every relevant physical asset;
• spatial and engineering relationships;
• real-time or near-real-time observations;
• historical records and version control;
• models for performance, deterioration or demand;
• visualisation and decision-support functions; and
• governance defining who can change data or issue control commands.

A twin should be developed for a decision need. For example, a bridge twin may support load monitoring and maintenance planning, while a corridor twin may test signal strategies, work zones and demand scenarios. Complexity without a clearly defined decision use increases cost without guaranteeing value.

4. AI, Forecasting and Optimisation

4.1 Forecasting methods

Transport forecasting can use statistical, machine-learning and hybrid approaches. Time-series models may represent trend, seasonality and autocorrelation. Regression models provide interpretable relationships with weather, events, land use or economic indicators. Tree-based methods and neural networks can capture nonlinear interactions, while graph-based deep-learning models can represent spatial relationships among network links.

The selection must reflect the decision horizon. A short-horizon speed forecast may benefit from high-frequency sensor data; a long-term demand forecast requires land-use, demographic and policy variables. A model that predicts well in ordinary conditions may be unsafe for rare incidents. Evaluation should use a time-based holdout rather than randomly mixing future observations into training data.

For observed values yi and predictions ŷi, mean absolute error is

MAE = (1/n) Σ |yi − ŷi|,

and root mean square error is

RMSE = √[(1/n) Σ(yi − ŷi)²].

Both have the same units as the predicted quantity. RMSE gives greater influence to large errors, whereas MAE is easier to interpret as an average absolute error. Neither establishes causation, fairness or operational usefulness. A forecast should also be assessed for calibration, bias by location or user group, robustness to missing data and consequences of errors.

4.2 Optimisation and control

Optimisation selects decisions from feasible alternatives according to an objective and constraints. A simplified signal-control problem may minimise a weighted function of vehicle delay, pedestrian delay, stops and emissions, subject to cycle-length, phase-order, intergreen and safety constraints. The weights are policy choices, not neutral technical facts.

Traffic systems may use fixed-time, actuated, adaptive or predictive control. Fixed-time plans are predictable and robust where demand patterns are stable. Actuated control responds to detector inputs. Adaptive systems update timings according to measured conditions. Predictive control can incorporate forecasts, but model error and computation time may reduce benefits. Any automated controller requires fallback modes, bounds on permissible actions and monitoring for unstable behaviour.

AI should assist professional judgement rather than eliminate accountability. Explainability is especially important where a recommendation affects emergency response, access for disabled users, enforcement, road pricing or allocation of maintenance funds. Human review thresholds should be specified before deployment.

5. Predictive Maintenance and Asset Management

Traditional maintenance may be reactive or based on fixed intervals. Predictive maintenance estimates the probability or timing of failure using condition data, usage, environment and historical interventions. It can reduce unplanned disruption and direct inspections to assets with the greatest risk.

A risk-based framework commonly considers:

Risk = Probability of failure × Consequence of failure.

This is a decision-support relationship, not a universal physical law. Probability and consequence must be defined consistently, and the product can conceal important differences between safety, economic, environmental and social consequences. A low-probability bridge failure may warrant action because its consequence is catastrophic.

Condition indicators may include roughness, cracking, deflection, corrosion, vibration, bearing temperature, drainage performance and electrical faults. Models can estimate remaining useful life, but predictions are uncertain because deterioration is affected by loading, climate, workmanship, interventions and measurement error. Inspection and engineering review remain necessary, particularly for safety-critical assets.

A sound workflow is: establish the asset register; define failure modes; collect and validate condition data; develop deterioration or survival models; rank interventions; schedule work subject to budget and network constraints; and compare predicted with actual outcomes. The system should learn from completed work, including cases in which an algorithm’s recommendation was rejected.

6. Connected and Autonomous Vehicles

Connected vehicles exchange information with other vehicles, infrastructure, networks and pedestrians. Messages may describe position, speed, acceleration, hazard warnings, signal phase and timing or road-work conditions. Connectivity can support collision warnings, cooperative merging, transit priority, platooning and improved situational awareness.

Autonomous driving involves perception, localisation, prediction, planning and control. Automation levels describe the distribution of driving responsibility, but terminology must be interpreted carefully because a system may automate some operational conditions while requiring a human fallback in others. A vehicle that performs well on a mapped motorway may not be capable of safe operation in mixed traffic, poor weather or unstructured work zones.

Infrastructure implications include lane marking quality, digital mapping, roadside units, consistent signage, maintenance of communications, curb management and standardised work-zone information. CAVs may improve safety and capacity, but benefits are not automatic. Empty repositioning trips, induced travel, cybersecurity attacks, inequitable access and interaction with pedestrians and cyclists may create new risks. Transport planning should evaluate scenarios rather than assume universal adoption or perfect automation.

7. Cybersecurity, Privacy and Data Governance

7.1 Cybersecurity

ITS cybersecurity protects confidentiality, integrity, availability, authenticity and safety. Threats include ransomware, spoofed positioning, compromised roadside units, malicious software updates, denial-of-service attacks, unauthorised access to traffic cabinets and manipulation of public information. Safety and availability can be more important than confidentiality in a signal-control system, although personal data require strong confidentiality protections.

A lifecycle security approach includes asset discovery, threat modelling, risk assessment, secure architecture, least-privilege access, network segmentation, patch management, vulnerability disclosure, logging, backup, incident response and recovery exercises. Legacy controllers may not support modern controls; compensating measures such as isolation, monitored gateways and carefully restricted maintenance access may be required. Cybersecurity should be funded as an operational requirement, not treated as a one-time purchase feature.

7.2 Privacy and responsible data use

Location traces, number plates, mobile identifiers and video analytics can reveal movements and associations. Agencies should apply purpose limitation, data minimisation, retention limits, access controls, de-identification where appropriate and transparent public notices. De-identification is not automatically irreversible, especially when datasets can be linked. Privacy impact assessments should precede systems that collect identifiable or sensitive information.

Responsible AI requires attention to bias, explainability, contestability, accessibility and human oversight. A model trained mainly on central urban roads may systematically perform worse in rural areas. A predictive policing or enforcement application may reproduce historical inequalities. Governance should identify who is accountable for errors and provide a mechanism to challenge consequential decisions.

7.3 Interoperability and institutional governance

Open standards, documented APIs and common reference systems reduce vendor lock-in and permit integration. Data governance should specify ownership, stewardship, quality responsibilities, metadata, permitted uses, sharing conditions, retention, audit rights and incident reporting. Ownership of raw data is not sufficient; an agency must also have the capability and legal authority to use, protect and update it.

8. Economic and Professional Evaluation

An ITS business case should compare alternatives over their full lifecycle. Costs include equipment, communications, software licences, integration, training, cybersecurity, energy, calibration, data storage, maintenance, replacement and decommissioning. Benefits may include reduced crashes, travel-time savings, reliability, emissions reductions, operating savings and improved accessibility.

Benefit–cost analysis can be expressed as

BCR = Present value of benefits / Present value of costs.

Present values require a stated base year, analysis period and discount rate. Benefits and costs occurring in different years should not be compared without discounting. A positive BCR does not prove that a project is equitable or affordable, and non-monetised safety, privacy and distributional effects should be reported separately. Sensitivity analysis should vary demand, technology life, adoption, maintenance cost, benefit magnitude and cyber incident assumptions.

Procurement should avoid specifying a fashionable technology without specifying the operational outcome. Outcome-based requirements may define accuracy, latency, availability, interoperability, privacy, safety and support obligations. Pilot projects need pre-defined success criteria and a credible path to scale; otherwise they become permanent demonstrations.

9. Illustrative Applied Case: Digital Management of an Urban Corridor

Background

The following is an illustrative case, not a report of a particular real project. A metropolitan authority manages a 12-km corridor containing a bus route, signalised intersections, schools, retail access and a frequently congested bridge. Complaints concern unreliable bus journeys, incident-related queues and delayed maintenance of signals and pavement.

Problem and analysis

The authority initially relied on periodic manual counts and separate maintenance databases. It could measure average travel time but could not distinguish recurring congestion from incident effects. Signal faults were often discovered through public complaints. Asset records used inconsistent identifiers, and a proposed camera system raised privacy concerns.

A needs assessment identified four decisions: detect incidents rapidly; improve bus reliability; prioritise maintenance; and provide trustworthy public information. The authority established baseline measures for journey-time percentiles, bus schedule adherence, incident detection time, signal availability and pavement condition.

Approach and solution

A corridor GIS was linked to a BIM-derived asset register. Each signal, detector, cabinet, pavement segment and bridge component received a persistent identifier. Radar and video detection were installed at selected locations, with edge processing used to transmit traffic counts and incident classifications rather than unnecessary raw video. Probe-vehicle data supplemented fixed sensors.

An IoT platform monitored detector health, cabinet temperature, power and communications. A digital twin displayed current conditions and linked observations to asset history. A forecasting model predicted near-term queue growth, while an adaptive signal strategy gave conditional priority to buses within safety and coordination constraints. Maintenance staff received risk-ranked work orders, but bridge and pavement interventions remained subject to engineering inspection.

The authority adopted a governance plan covering retention, access, encryption, audit logs, model validation and public communication. A fallback signal plan was retained for communication failure. Performance was evaluated against the baseline and against a comparable corridor rather than relying only on before-and-after averages.

Outcome

In this illustrative scenario, the main benefit was not a single algorithm but integration. Operations staff could see whether an apparent congestion problem was caused by demand, a crash, a detector failure or a signal fault. Bus priority improved reliability during eligible conditions, while pedestrian safety constraints prevented unrestricted green extension. Maintenance planning became more proactive because condition, failure and work-order information were connected.

The evaluation also identified limitations: sensors performed less reliably during heavy rain, the prediction model was less accurate during unusual events, and data integration required continuing staff effort. These findings led to recalibration, additional event features, revised fallback procedures and a decision not to automate certain safety-critical controls.

Learning points

1. Begin with operational decisions and measurable outcomes.

2. Integrate asset identity and data quality before deploying advanced analytics.

3. Combine different sensors and preserve uncertainty and provenance.

4. Keep human oversight and safe fallback modes.

5. Treat privacy, cybersecurity and maintenance as core design requirements.

6. Evaluate distributional effects, not only corridor averages.

10. Challenges, Limitations and Contemporary Developments

Digital transformation can increase technical debt when agencies accumulate incompatible platforms, unsupported devices and undocumented interfaces. High initial investment may favour wealthy corridors and widen regional inequality. Data gaps can make rural, informal, disabled or low-connectivity users invisible. Algorithms may optimise measurable flow while neglecting safety, access or public space.

Emerging developments include low-cost edge AI, vehicle-to-infrastructure cooperative perception, federated learning, synthetic data, 5G and future communications, open mobility data platforms, automated inspection using drones and robotics, and climate-resilient digital twins. These developments should be assessed cautiously. Greater connectivity increases the attack surface; synthetic data may not reproduce rare hazards; and automated inspection may identify defects without understanding structural significance.

Climate change strengthens the case for real-time flood, heat, wildfire, landslide and storm monitoring. Digital systems can support evacuation and resilient operations, but sensors and communications themselves require backup power, physical protection and disaster recovery. Digital transformation is therefore part of infrastructure resilience, not a substitute for robust physical design.

Best-Practice Framework

A professional ITS programme should:

1. define the transport problem and affected users;

2. establish measurable safety, mobility, environmental and equity objectives;

3. map processes, assets, data flows and institutional responsibilities;

4. select the minimum technology capable of supporting the decision;

5. specify data quality, interoperability, cybersecurity and privacy requirements;

6. pilot under normal and abnormal conditions;

7. validate algorithms using independent, time-appropriate and subgroup-aware tests;

8. provide human oversight, fail-safe behaviour and manual fallback;

9. budget for operations, maintenance, training and replacement; and

10. monitor outcomes, audit models and retire systems that no longer provide value.

Chapter Summary

Intelligent Transportation Systems integrate sensing, communications, computation and control to improve transport performance. Digital transformation extends this integration across the asset and service lifecycle through GIS, BIM, IoT platforms, automated data collection and digital twins. Effective systems depend on sensor suitability, data quality, interoperability and operational capability.

AI can support forecasting, optimisation and predictive maintenance, but performance must be assessed with appropriate error measures, time-based validation, uncertainty analysis and human oversight. Connected and autonomous vehicles create opportunities for safety and efficiency while introducing infrastructure, equity and cybersecurity challenges. Cybersecurity, privacy, responsible AI and data governance are not optional additions; they are essential conditions for trustworthy deployment. Finally, ITS alternatives should be evaluated through lifecycle, economic, social, environmental and institutional criteria, with safe fallback arrangements and continuous evaluation.

Key Terms

Intelligent Transportation Systems (ITS): Coordinated use of information, communication, sensing and control technologies in transport.

Internet of Things (IoT): Network of physical devices that sense, communicate and sometimes act on data.

Edge computing: Processing data near the point of collection.

GIS: System for storing, analysing and visualising geographically referenced information.

BIM: Structured digital representation of built assets and their relationships across the lifecycle.

Digital twin: Dynamic, connected digital representation of a physical asset or system.

V2X: Vehicle-to-everything communication, including vehicle, infrastructure, network and pedestrian links.

Connected vehicle: Vehicle capable of exchanging data with external systems.

Autonomous vehicle: Vehicle performing part or all of the driving task through automated systems.

Precision: Proportion of detected events that are true events.

Recall: Proportion of true events detected by a system.

Data provenance: Record of data origin, transformation and history.

Predictive maintenance: Maintenance planned using predicted condition or failure risk.

Digital model: Representation not necessarily connected automatically to its physical counterpart.

Digital shadow: Representation updated by the physical system, generally in one direction.

Interoperability: Ability of systems to exchange and use information effectively.

V2I: Vehicle-to-infrastructure communication.

Data governance: Rules and responsibilities for data quality, access, use, protection and retention.

Model drift: Decline in model performance as operating conditions or data relationships change.

Review Questions

Short Answer Questions

1. Identify the five functional layers of an ITS architecture.

2. Distinguish precision from recall in automated incident detection.

3. Explain why a three-dimensional visualisation is not necessarily a digital twin.

4. State two cybersecurity controls appropriate for roadside IoT devices.

5. Give two limitations of AI-based transport forecasting.

Long Answer Questions

1. Explain how GIS, BIM and digital twins can be integrated across the transport asset lifecycle.

2. Discuss the principles and practical requirements of predictive maintenance for bridges and pavements.

3. Evaluate the potential benefits and risks of connected and autonomous vehicles for urban transport.

4. Describe a complete data-governance framework for an ITS programme collecting location and video data.

5. Explain how an agency should evaluate an adaptive signal-control project using technical, economic and social criteria.

Analytical / Application Questions

1. A corridor authority has frequent detector failures and unreliable congestion alerts. Develop a diagnostic and improvement plan covering sensor validation, communications, data quality, cybersecurity and performance measurement.

2. An AI model has MAE of 4 minutes on ordinary weekdays but 12 minutes during severe storms. Assess whether it should be used for public travel-time alerts and propose a validation and risk-control strategy.

3. Compare two ITS procurement options: a proprietary integrated platform and an open, modular architecture. Analyse lifecycle cost, interoperability, cybersecurity, accountability, scalability and vendor dependence before recommending a strategy.

Multiple Choice Questions with Answers

Question 1

Which ITS layer contains cameras, radar and inductive loops?

A. Physical layer

B. Sensing layer

C. Data and service layer

D. Decision and control layer

Correct Answer: B. Sensing layer

Question 2

What does recall measure in an incident-detection system?

A. The proportion of alerts that are correct

B. The average time between detected incidents

C. The proportion of true incidents that are detected

D. The proportion of vehicles classified correctly

Correct Answer: C. The proportion of true incidents that are detected

Question 3

Which characteristic most directly distinguishes a digital twin from a static digital model?

A. Use of three-dimensional graphics

B. Connection and data exchange with the physical system

C. Storage in a cloud platform

D. Use of a geographic coordinate system

Correct Answer: B. Connection and data exchange with the physical system

Question 4

Which error measure gives greater influence to large prediction errors?

A. Mean absolute error

B. Median absolute deviation

C. Root mean square error

D. Percentage of missing observations

Correct Answer: C. Root mean square error

Question 5

Why is edge computing useful in some ITS applications?

A. It eliminates the need for sensors

B. It can reduce latency and bandwidth requirements

C. It guarantees that algorithms are unbiased

D. It removes all cybersecurity risks

Correct Answer: B. It can reduce latency and bandwidth requirements

Question 6

Which statement about predictive maintenance is most accurate?

A. It eliminates the need for engineering inspection

B. It schedules every asset at a fixed interval

C. It uses condition and risk information to support intervention decisions

D. It is applicable only to vehicles

Correct Answer: C. It uses condition and risk information to support intervention decisions

Question 7

What is a principal purpose of network segmentation in ITS cybersecurity?

A. To increase camera resolution

B. To prevent or limit movement of an attacker between systems

C. To replace encryption

D. To improve pavement strength

Correct Answer: B. To prevent or limit movement of an attacker between systems

Question 8

Which measure is especially relevant to travel-time reliability?

A. Only the maximum speed recorded

B. A journey-time percentile such as the 95th percentile

C. The number of database tables

D. The camera field of view

Correct Answer: B. A journey-time percentile such as the 95th percentile

Question 9

Why should AI transport models be tested on time-based holdout data?

A. To ensure future information is not improperly used during training

B. To increase the number of sensors automatically

C. To remove the need for calibration

D. To guarantee zero prediction error

Correct Answer: A. To ensure future information is not improperly used during training

Question 10

Which is an example of data governance?

A. Choosing a vehicle colour

B. Defining data access, retention and quality responsibilities

C. Increasing tyre pressure

D. Changing a road’s lane width without analysis

Correct Answer: B. Defining data access, retention and quality responsibilities