Sensing, Intelligence, and the City Operating Picture

5.1You cannot govern what you cannot see

Every instrument described in the following chapters depends on data. Unless a city can see what is happening, in enough detail and fast enough to catch problems before they compound, tools like licensing conditions, pricing mechanisms, enforcement powers, and design standards are useless. Build the sensing layer first.

Each of the five value mechanisms in Chapter 4 has leading indicators that appear well before the damage. Curb dwell times and VMTVMT (Vehicle Miles Traveled)Total distance driven across a fleet or area, the standard measure of road demand. Central to the paper because autonomous fleets are modelled to increase VMT — by around 6 percent on the peer-reviewed central estimate, across a 4 to 60 percent range depending on occupancy and pricing assumptions, with empty repositioning alone able to double city-wide miles. Empty repositioning miles are still miles. (Ch.3.) per passenger trip appear before congestion statistics; parking utilization appears before stranded-asset write-downs; parking revenue trends appear before budget shortfalls; depot load profiles appear before network stress; approval ratings and opposition activity appear before political bans. Detecting the early signals requires a clear operating picture. Without one, the city discovers value destruction when it is too late to intervene.

5.2Data spans across domains

Mobility and curb operations. Vehicle counts by type and occupancy state; pickup and drop-off dwell times and their distribution across curb segments; routing patterns at the level of individual street segments; and curb utilization rates disaggregated by use category. This is the data that enables curb pricing, fleet routing governance, pooling mandate enforcement, and transit interaction management. Waymo’s fleet operates thirteen cameras, four LiDAR units, and six radar sensors per vehicle^EV28EV28What Data Do AV/CAM Fleets Actually Collect?Location, video, LiDAR, radar, biometric, V2X, and passenger data collected continuously; retention, sharing, and consent rules remain fragmented…Open in the Evidence Base →, a suite that generates on the order of 1.4 terabytes of operational data per hour at a 10 Hz cadence (the low end of published estimates for comparable configurations). Cities receive almost none of this.

Energy and grid constraints. Depot charging load profiles, disaggregated by time of day and vehicle type; grid hosting capacityHosting capacityHow much additional load or generation a given point on the electricity network can accept without reinforcement. Publishing it is what turns depot siting from a series of private negotiations into a planning question a city can participate in. (Ch.2.) by substation and feeder circuit; connection queue status and forecasted connection timelines; and demand-response performance during managed charging events. This data sits almost entirely within utility company systems and is not currently shared with transport or planning authorities in any formal way.^EV31EV31Utility Data Sharing: Governance and Accessibility BarriersUtility data siloed by proprietary systems, regulatory gaps, and access inequities; barrier to EV/AV infrastructure planning and grid integrationOpen in the Evidence Base → The consequence is that transport licensing decisions and depot planning approvals are routinely made without any assessment of whether the grid can accommodate the resulting load – an infrastructure planning failure that is invisible until it manifests as a connection delay or a curtailment event.

Land and real estate signals. Parking utilization rates by location and time of day; depot site availability, including brownfield and underused commercial sites with grid access; conversion feasibility indicators for existing parking structures; and land value trends in corridors where CAMCAMConnected and Autonomous Mobility — the commercial deployment of autonomous vehicles operating connected to data and control networks. The paper uses CAM rather than "AV" to foreground the network and governance dimensions that determine how the technology lands in cities. (Ch.1 §1.1; Ch.2 §2.1.) deployment is active or planned. This data is essential for the spatial and investment analysis described in Chapter 7, and for the fiscal transition planning that models the revenue implications of declining parking utilization.

Fiscal and pricing performance. Real-time and trend revenue tracking across parking charges, penalty charge notices, transit fareboxFarebox recoveryThe share of a transit system’s operating cost met by passenger fares. It matters here because it measures exposure: Transport for London’s 58 percent recovery is 58 percent of a budget that a mode shift to autonomous road transport could erode. (Ch.1.) receipts, and any new CAM-related charges or levies. The fiscal value destruction mode requires early detection to allow fiscal planning to respond before the gap has opened in the accounts. A city that is tracking this data monthly can begin building fiscal reserves and replacement mechanisms several years ahead of the peak of disruption. A city that is not tracking it will discover the gap in the annual budget report.

Equity and access. Service coverage maps showing where autonomous fleet operations are active and where they are absent; average fare per trip disaggregated by income quartile of origin area; accessibility compliance rates; wait time distributions by geography; and the spatial distribution of operational externalities such as emissions, noise, and curb conflict. Without this data layer, a city cannot determine whether CAM is reproducing the patterns that ride-hailing established – where, as Chicago’s 2019 shared micromobility pilot demonstrated, 77 percent of trips concentrated in areas already well-served by transit, while lower-income and outer areas received minimal service despite being the communities with the greatest unmet mobility need.

Compliance and auditability. Operator adherence to licensing conditions across all domains: safety incident reporting rates and completeness; data-sharing obligation fulfillment; service-level performance against coverage and accessibility commitments; and fleet compliance with routing, occupancy, and operational constraints. The Cruise incident in San Francisco was a compliance failure – specifically, a data disclosure failure – before it was a safety failure.^EV07EV07Cruise San Francisco: Scale Without Aligned Institutional ControlCruise SF robotaxi expansion outpaced regulatory oversight; October 2023 pedestrian drag incident exposed misreporting to CPUC, triggering permit…Open in the Evidence Base → Without independent audit rights, a disclosure failure only surfaces at the point of regulatory crisis.

5.3The minimum viable operating picture

A city does not need to build all six of these data capabilities to full specification before it can begin effective governance. It needs a minimum viable operating picture (MVOPMVOP (Minimum Viable Operating Picture)The minimum sensing and intelligence capability a city needs to govern CAM credibly: the smallest combination of data feeds, schemas, and analytical capacity that lets the city see what is happening across its territory and act on it. Sequenced over time toward a full operating picture. (Ch.5 §5.3.)) – the smallest set of data capabilities that enables the most critical instruments described in Chapters 6 and 7 to function. MVOPMVOP (Minimum Viable Operating Picture)The minimum sensing and intelligence capability a city needs to govern CAM credibly: the smallest combination of data feeds, schemas, and analytical capacity that lets the city see what is happening across its territory and act on it. Sequenced over time toward a full operating picture. (Ch.5 §5.3.) makes the investment tractable: a city with limited resources and limited political bandwidth can sequence its sensing investment in a way that delivers the most governance value first and builds towards the full operating picture over time.

The technical infrastructure for building an MVOP largely exists. The Mobility Data SpecificationMDS (Mobility Data Specification)An open standard for how mobility operators report trip and vehicle data to cities, originating with Los Angeles and now widely adopted for micromobility. The paper treats MDS 2.1 as mature enough that no city need invent its own schema. (Ch.4.) at version 2.1, initially developed for micromobility and now being extended to autonomous vehicles^EV3430EV3430MDS 2.1 adds AV-specific incident reporting, real-time policy push, and enforcement data structures — moving the governance toolkit closer to autonomous-vehicle oversightOMF MDS 2.1, 1,000+ day cycle, final review March 2026; incidents endpoint, ADS disengagements, remote takeovers, sub-second geofence push; first MDS…Open in the Evidence Base →, provides a standardized protocol for real-time trip reporting, vehicle status updates, and event data. It is used by more than 200 cities and public agencies across 21 countries – Los Angeles, Portland, and Louisville among the first – and its data model is well understood and documented.^EV3280EV3280MDS: Open-Source Digital Infrastructure for Right-of-Way Governance at Global ScaleMDS open standard, LADOT-originated, OMF-stewarded Nov 2019; 6 APIs, v2.0.2 June 2025; 200+ cities, 21 countries, 50+ MSPs — regulatory relationships…Open in the Evidence Base → Segment-level spatial referencing – pioneered by SharedStreets and carried forward, since that project went dormant, by Overture’s Global Entity Reference SystemGERS (Global Entity Reference System)The Overture Maps Foundation’s shared identifier system for physical places, carrying forward the role once played by proprietary map IDs. Enables privacy-preserving aggregate pattern analysis, because places can be matched without exchanging trip-level records. (Ch.3.) (GERS) – allows city-level aggregate pattern analysis without exposing individual trip data to privacy risk^EV29EV29Privacy-Preserving Data Standards for Fleet Data SharingAnonymisation, consent frameworks, and data-minimisation protocols emerging as core standards for cross-fleet sharing; regulatory alignment still…Open in the Evidence Base →. The regulatory floor for data collection and use is established in GDPRGDPRThe EU and UK General Data Protection Regulation. Relevant throughout the Data domain because mobility data is frequently personal data, and the lawful basis for a city to collect, retain and re-disclose it has to be established rather than assumed. (Ch.3.), the UK Data Protection Act 2018, and comparable legislation across most developed jurisdictions. The governance instruments for requiring data sharing – operator licensing conditions, concession agreements, data access requirements – are available in every jurisdiction that can license commercial vehicle operations.

A contemporary US worked example demonstrates how this is built in practice. Austin’s Transportation and Public Works department, working with NACTO and the Open Mobility Foundation, was an active contributor to the development of the Mobility Data Specification that several major North American cities now use to regulate shared micromobility. Austin requires that an operator’s license to operate within the city is conditional on the operator’s commitment to provide data to the city in the MDS-compliant format.

The same move – data sharing as a condition of license – is available for CAM. The technical standards are already mature. The unresolved questions are which CAM-specific data classes belong in the condition (vehicle telemetry, curb dwell, occupancy state, remote assistance operator (RAO) intervention logs), at what aggregation, with what audit rights, and under what conditions of secondary disclosure. These are license-design questions, not technical questions – and they are best answered once, through the standards body that produced MDS, rather than re-derived bilaterally by every city that licenses a fleet.

Los Angeles’ Express Park program used real-time occupancy sensor data to drive dynamic parking pricing, reducing average parking search duration by 37 percent and raising parking revenue by 16 percent in the areas measured, while improving the quality of the curb environment. San Francisco’s SFpark program achieved target curb occupancy levels 31 percent more often than its static-pricing baseline.^EV30EV30City Curb Sensing and Parking Data PlatformsMunicipal curb sensors and parking data platforms enable real-time occupancy visibility, dynamic pricing, and AV integration at urban kerbsidesOpen in the Evidence Base → In the UK, AppyWay’s integrated curbside management platform – which combines real-time occupancy sensing, traffic regulation order digitization, and automated enforcement into a single operational system – has been deployed in several UK local authorities^EV30EV30City Curb Sensing and Parking Data PlatformsMunicipal curb sensors and parking data platforms enable real-time occupancy visibility, dynamic pricing, and AV integration at urban kerbsidesOpen in the Evidence Base → and provides a practical template for the kind of curb sensing capability that an MVOP requires.

The energy layer differs from the other five domains in two ways that have direct implications for what data a city needs and where it must go to get it. The first difference is temporal resolution. Knowing the total daily charging load from a depot is insufficient for grid management; what matters is the hourly – and in some cases, half-hourly – distribution of that load, because the constraint is peak demand on the local distribution network, not average demand over the day. The MVOP standard for energy data should therefore specify hourly load profiles as the minimum temporal resolution, not daily totals.

The second difference is spatial resolution. Grid hosting capacity is not a city-wide average; it is a property of specific substation and feeder circuit areas, which can vary dramatically across short distances in dense urban environments. A depot site that is within the hosting capacity of one feeder circuit may be 200 meters from a site that would require a network upgrade costing several hundred thousand dollars. City transport and planning departments do not currently have access to this data in a usable form. Distribution network operatorsDNO (Distribution Network Operator)The regulated monopoly that owns and operates the local electricity distribution network in a UK region. Depot siting decisions fall in DNO areas rather than city boundaries, so a single operator’s fleet plan can require coordinated review across several networks. (Ch.3.) hold it – in some cases at substation level, in some cases aggregated to broader areas – and disclosure practices vary by jurisdiction and by operator. Building the energy layer of the operating picture requires a formal data-sharing agreement with the relevant distribution network operator, not merely a licensing condition on fleet operators who do not hold the data that matters most.

The ownership and access question is therefore different for energy data than for any other domain in the operating picture. Fleet operators can be required, as a condition of licensing, to share trip data, occupancy data, and operational performance data. But grid hosting capacity data is not in the operator’s control; it is in the distribution network operator’s systems. In the UK, the energy regulator Ofgem’s network charging and access reform program includes requirements for DNOs to publish hosting capacity maps at higher spatial resolution than previously available – a policy change that creates an opening for city planning authorities to build formal data access relationships. In the United States, the Federal Energy Regulatory CommissionFERC (Federal Energy Regulatory Commission)The US regulator for interstate electricity transmission and wholesale markets. Its interconnection queue reform has increased visibility of where and when grid capacity will actually become available — the constraint that governs depot siting. (Ch.5.)’s (FERC) interconnection queue reform has increased visibility of transmission-level capacity, but distribution-level hosting capacity remains inconsistently disclosed.

Cities that are serious about governing the energy domain of CAM deployment need to work with their distribution network operator, understand what hosting capacity data is available, and build the inter-agency relationship to access and use it before depot applications begin arriving. This chapter does not provide the corresponding governance of the city itself as data holder – retention, purpose limitation, secondary disclosure, and independent oversight – which needs the same specificity as the rest of the toolkit; it is recorded as an open commitment in section 8.6

5.4From operating picture to simulation

A mature operating picture does more than describe what is happening now. It enables the simulation of what is likely to happen next – and therefore the design of governance interventions in advance of the conditions they will need to address. A city with a well-developed operating picture can answer the governance planning questions that are most consequential: How would curb demand change if a fleet of 300 additional vehicles entered a particular corridor? Under what occupancy ratio and routing conditions does induced VMT begin to exceed road capacity in each area? At what fleet size does a specific depot location’s charging load exceed the grid hosting capacity of its substation? How will the operating environment change if we see rapid adoption of personal autonomous vehicles? Licensing decisions, curb allocations, and depot approvals have to answer these questions, and they can only be answered well with a simulation layer on top of the operating picture.

There are already starting points that can be built uponTransport for London’s strategic transport model, VISSIM and AIMSUN traffic simulation platforms, and the growing ecosystem of mobility simulation tools built on open data standards all provide starting points. What they require is input data – the real-time fleet behavior, charging load, and curb utilization data that the operating picture provides. Simulation is valuable for plausibility testing, not precision.