Annex C. AI-readiness and the graph model

C.1.1 This annex sets out the recommended structure of the approach to the AI-readiness of the sustainable development documentation of historic settlements and the recommended mechanism for building and operating the information system of the territory on the basis of Clauses 4–12 of this standard. The provisions of this annex are of a recommendatory character; application of the graph model is not a condition of conformity of the documentation to Clauses 4–12.

C.1.2 The annex defines:

a) the structure for forming the skills of the information system;

b) the composition of the unified graph model of the territory and the rules for populating it;

c) operations on the model in the course of developing the documentation and after its approval;

d) the criteria for recognizing a clause of the documentation as AI-ready;

e) the register of the principles of Clause 4 with stable identifiers (C.9);

f) the correspondence of parts of the model to parts of the master plan and to sources of spatial data (C.10).

C.1.3 This annex does not establish requirements for a specific software platform, implementation language or supplier of software tools. A forecast of the effects of measures that uses the graph model does not replace the assessment in accordance with 7.12. The graph model supplements the cartographic layer and does not replace it.

C.1.4 The recommended approach is as follows. On the basis of Clauses 4–12, the parent skill forms the skills of the information system, including skills based on artificial intelligence. The skills of the clauses form the content of the clauses of the sustainable development documentation and at the same time build the system of graphs of the urban model. The principles of Clause 4 are mapped to a controlling skill that operates at all stages and that ensures maintenance of the register of principles and checking of the consistency of objectives, decisions and measures with the principles. The state of the territory and its problems are described by the graph of the current state of the urban system; measures are described by graphs of measures within the structure of this model; links are established between the parts of the model. The model is updated as the documentation is developed and according to the results of monitoring, and does not remain a textual record fixed at the moment of approval.

C.2.1 The approach consists of three components of one system.

Table C.1 — Components of the approach

A skill within the meaning of this annex is a formalized procedure of the information system, including one based on artificial intelligence, and not a requirement for a supplier of software tools. For any skill of the system, it is defined from which clause and through which skill it is obtained. The purpose of the kinds of skills is established in C.2.2–C.2.4.

C.2.2 The parent skill, on the basis of Clauses 4–12:

a) forms the clause skills;

b) sets the input data, the results and the criteria for checking the clause skills;

c) maintains the register of skills and maintains conformity to the current edition of the standard;

d) forms the system-wide skills (C.5.11).

The parent skill does not create nodes of the graph model of its own.

C.2.3 A clause skill forms the content of one clause of the documentation. Clause skills have a dual purpose: they are a tool for developing the clauses of the sustainable development documentation and they form the functional core of the information system of the territory after the documentation is approved.

A clause skill gives the following results:

a) the text of the clause and its structured results;

b) a contribution to the graph model within the scope of Table C.6: Clauses 4–9 populate the model; Clauses 10–12 work with the model already built;

c) derived skills according to the functionality of the clause.

By the time the documentation is approved, the graph model and a set of skills capable of working with it exist.

C.2.4 A derived skill is formed by a clause skill or by the parent skill and performs operations on the model or on the system:

a) derived skills of a clause skill — according to the functionality of the clause; composition — Table C.6;

b) derived skills of the parent skill — the system-wide skills (C.5.11) and the conformity of the skills to the edition of the standard.

Four operations on the model are performed by derived skills when called by the clause skills (C.6). The model is updated as the documentation is developed and according to the results of monitoring.

C.3.1 The unified graph model (the graph of the urban system) is one model of the territory in which four named parts are distinguished. The parts are not separate models of the city. A link always connects specific nodes and does not connect a part of the model as a whole. Skills add nodes and links as the documentation is developed and do not rewrite the model anew.

C.3.2 The composition of the named parts is given in Table C.2.

Table C.2 — Named parts of the graph model

Clauses 10–12 do not form parts of the model of their own.

C.3.3 Elements that do not form a separate part:

a) principles — nodes of the register within the graph of the current state; composition and number — according to the register of Clause 4;

b) problems — nodes of the graph of the current state, tied to states; there is no separate part "graph of problems";

c) the project solution — a set of graphs of measures formed by the skill of Clause 9; it does not form a fifth part of the model.

C.3.4 For a result of a skill, the fragment of the traceability chain formed at the corresponding stage is checked. The full chain "measure → objective → problem → principle" is mandatory at the acceptance of the linkage of the graphs of measures with the other parts of the model (C.4.5) and at subsequent updating.

C.3.5 The core of the territorial digital twin is formed by the graph of the current state, geographic information layers and the skills. The formation of the territorial digital twin and its recommended composition are given in 11.6.5.

C.4.1 The model describes the territory by two layers of nodes.

Table C.3 — Layers of nodes

The value of a domain node that relates to a specific territorial unit is set by an edge of the kind "territorial assignment" between the domain node and the spatial node (C.4.2, item e). The propagation of influence proceeds along the links of the territorial units of the given node and of the linked nodes, and not between domains as a whole.

A node of the domain layer is designated by the domain code and the name of the state, for example INFRA.energy_efficiency. Links are built between nodes.

For a domain node, at least the following are recorded: the current value of the indicator as a characteristic of the node, the unit of measurement, the data source and the date of update. The target value is a characteristic of a node of the graph of the vision and target states; that node is linked to the domain node by an edge between parts of the model.

A problem node is tied to the domain and spatial states that it describes. A hypothesis of a cause is formalized as a hypothetical edge between states, with an assessment of probability.

The vision graph may introduce nodes that do not yet exist in the current state of the territory. When project solutions are adopted, such nodes are included in the graph of the current state and receive base values.

C.4.1.1 The composition of the nodes of each domain is determined by the depth of analysis and the availability of data. The examples are of an informative character.

Table C.4 — Domains of the graph of the current state

Examples of links between nodes of two domains: INFRA.energy_efficiency → BUDGET.operating_costs; INFRA.water_supply → SOC.health; HER.okn_condition → TOUR.attractiveness. The kind and the attributes of a link are established in C.4.2.

C.4.2 An edge is a semantic dependence or a relation between two nodes, and not a joint mention in the text. A link without a recorded justification is not entered into the model.

The groups of links coincide with the permissible values of the attribute "kind of link":

a) normative constraints — the kind "normative constraint": legal requirements and conditions of protection of cultural heritage objects and of the subject of protection of the historic settlement. Blocking links have a probability of 1.0; the attributes "strength of influence" and "attenuation" are not applied to them;

b) links between parts of the model — the kind "link between parts of the model": edges between nodes of different named parts (C.4.3);

c) domain dependences — the kinds "cause and effect", "observed dependence", "unverified hypothesis": links between the states of nodes of the domains. Hypothetical links are subject to checking against monitoring data;

d) spatial links — the kind "spatial proximity": planning, transport, economic or protection-related connectedness of territorial units, determined by domain characteristics;

e) territorial assignment — the kind "territorial assignment": an edge between a domain node and a territorial unit that sets the value of a characteristic of the domain node for that unit. The attributes "strength of influence" and "attenuation" do not apply.

An edge is established on the basis of studies, results of analysis or observations that show how one node influences another. Socio-cultural, socio-economic, urban-planning and environmental studies, and studies in the field of cultural heritage, are suitable, including those carried out in developing the documentation under this standard: a dependence that has been found becomes an edge with an indication of the source.

The hierarchy of grounds is: a normative requirement — observation data and studies — an expert assessment — an unverified hypothesis. Confidence is expressed by the attribute "probability": it is the higher, the better the quality of the evidence base and the closer the conditions of the study are to the given territory; with several independent confirmations the probability increases. The longer the chain of links, the lower the probability of the result and the weaker the effect that arrives. Links with a high probability require recalculation of indicators, those with a medium probability require checking, and those with a low probability are taken into account for reference. Influence between territorial units decreases with distance; the rate of decrease differs by domain (C.4.4).

For domain dependences and spatial links, the seven attributes of Table C.5 are recorded. For a blocking link, a territorial assignment and links between parts of the model that do not transmit quantitative influence, the kind, the direction, the justification and the date of update are mandatory. For a blocking link, a probability of 1.0 is recorded in addition. The attributes "strength of influence" and "attenuation" are not applied to the links indicated.

Table C.5 — Attributes of a link

C.4.3 A node of one graph is linked to specific nodes of another graph, and not to the graph as a whole:

a) an objective — to the domain nodes and the problems of the graph of the current state at which it is directed; the objective is measured by the values of the characteristics of the linked domain nodes;

b) a target state — to the same domain nodes and problems of the graph of the current state;

c) a target state and an element of the vision — to nodes of the graph of objectives;

d) a measure after the linkage under C.4.5 — to anchors on the graph of the current state, including nodes of actual states, problems and principles; to the objectives that it implements; where necessary — to nodes of the vision. A measure node may be linked to any node of the other parts of the model;

e) a problem — to the principle that has been infringed; a measure and an objective — to the principles that are implemented.

C.4.4 The model is tied to the territory through the division of the settlement into territorial units with integral characteristics of the domains: within the boundaries of a unit, the values of the domain nodes describe its state as a whole.

There are several levels of division; they correspond to the depth of analysis:

a) the municipality (the territory of the master plan) — strategic indicators, the budget, target benchmarks;

b) the settlement (inhabited locality);

c) the territory of the historic settlement — a zone of special requirements (5.3.1);

d) large territorial formations — programs for cultural heritage objects, tourist routes, inter-district links;

e) the smallest units of detailed analysis — planting, transport, noise;

f) an object.

In addition, functional zones, territorial zones and zones with special conditions for the use of territories, including protection zones of cultural heritage objects and safeguard zones, the territories of priority projects and the territories of integrated development, are recorded on the spatial layer.

An indicator is recorded at the level that corresponds to the depth of analysis and is aggregated to the higher levels. This annex does not establish the method of constructing the grid of units.

The matrix of spatial weights sets the topological basis for the propagation of influence between territorial units: adjacency, a transport link, belonging to one protection zone.

The domain attenuation coefficient sets how quickly the influence of a change in a given domain decreases with distance. In some domains the influence propagates far; in others it attenuates quickly. The coefficients are calibrated against monitoring data.

Between the same territorial units, different connectedness and a different radius of influence apply for different domains.

C.4.5 The graph of measures uses the same kinds of nodes and links as the graph of the current state: expected changes are expressed in the same characteristics of domain nodes, the same objects and the same territorial units. The graph of measures is formed in two stages.

Stage 1 is performed by the skill of Clause 9 at the project level. Each measure is formed as a separate graph: a set of nodes and edges that describe the measure itself, including the composition of the works, the expected changes in the values of domain nodes, the normative constraints of protection and the spatial assignment. At this stage the graph of the measure is not linked to the nodes of the other parts of the model. The project solution is a set of such graphs.

Stage 2 is performed by the derived skill for the forecast of effects, formed by the skill of Clause 9, when the forecast of effects is made. The graph of the measure is linked to the nodes of the other parts of the model. The measure acts as a node and is linked:

a) to nodes of the graph of the current state — nodes of actual states, problems and principles: how the measure affects them; anchors are selected on individual nodes, and not on a domain as a whole;

b) to the nodes of the graph of objectives that the measure implements;

c) to nodes of the graph of the vision — where necessary.

A measure node may be linked to any node of the other parts of the model. The mandatory minimum at linkage is: anchors on the graph of the current state and the objectives. Indirect effects are calculated by cascading along the edges of the graph of the current state, including by territorial units (C.4.2, C.4.4). The normative constraints of heritage protection that apply to the measure are recorded as blocking links. Implementation attributes are assigned by the skill of Clause 10, and implementation statuses by the skill of Clause 11.

The same graph of measures is used in two modes: before implementation — with the expected changes; after implementation — with the actual changes. Expected and actual values are stored separately; the actual result does not replace the forecast. The result of the linkage of stage 2 is used in the assessment of the effects of the measure: before implementation — as a forecast; after implementation — when the forecast is compared with the actual result.

C.5.1 The content of Clauses 4–12 is not reproduced by this Annex. The contribution of a skill to the model and the skills formed by it are given in Table C.6. The order of performing the skills and of populating the model is established in C.5.2.

Table C.6 — Mapping of clauses to skills and the model

C.5.2 The order of the clauses of the documentation is mandatory. The order of populating the model is applied only when this Annex is applied. The controlling skill of Clause 4 operates at all stages when this Annex is applied.

Order of the documentation:

preliminary objective-setting → 6 → 7 → refinement of objectives based on the results of the analysis → 8 → 9 → 10 → 11 → 12

Order of populating the model:

principles register (4) → basis of the graph of the current state (6) → graph of the current state: links, problems, hypotheses (7) → graph of objectives and links between parts of the model (refinement of objectives) → vision and target states (8) → graphs of measures (9) → implementation attributes (10) → execution statuses and the digital environment (11) → updating and refinement of edges (12)

Preliminary objective-setting precedes Clause 6: the preliminary system of objectives and the scope of application are entered in the graph of objectives without binding to the graph of the current state. Refinement of objectives based on the results of the analysis is performed after Clauses 6 and 7: the objectives are refined and linked with the states and problems of the graph of the current state.

C.5.3 The skill of Clause 5 is performed in two stages that coincide with the order in C.5.2. Preliminary objective-setting is performed before Clauses 6 and 7 and forms the context, the interested parties, the boundaries, the preliminary system of objectives and the assignment for the collection of source data; the preliminary system of objectives and the scope of application are entered in the graph of objectives. Refinement of objectives based on the results of the analysis is performed after Clauses 6 and 7, refines the measurable objectives and establishes links, between the parts of the model, with the states and problems of the graph of the current state. Each objective is checked by the skill of Clause 4.

C.5.4 The skill of Clause 6 is performed after preliminary objective-setting. The source data are the assignment of 5.8 and the data of 6.2–6.8. The skill forms the baseline and the basis of the graph of the current state: domain states with a source, a date of updating and a binding to territorial units (C.4.1, C.4.4). Synchronization of the skill of Clause 6 relates to the source data of the baseline. Assessment under 7.12 is not performed by the skill of Clause 6.

C.5.5 The skill of Clause 7 is performed after Clause 6 and before the refinement of objectives based on the results of the analysis. Assessment under 7.12 is taken into account as normative constraints of protection. The preliminary system of objectives is not reopened.

C.5.6 The skill of Clause 8 is performed after the refinement of objectives based on the results of the analysis. The skill forms the graph of the vision and target states, including nodes of the image of the future that do not yet exist in the current state of the territory, and forms derived skills: cascading of the vision, calculation of the target gap and updating of the vision on the grounds of 8.10.2.

C.5.7 The skill of Clause 9 forms the project solution as a set of graphs of measures (stage 1 under C.4.5). The derived skill of the impact forecast performs the linkage of stage 2. Clause 9 creates a named part of the model — the graphs of measures.

C.5.8 The skill of Clause 10 assigns implementation attributes to the graphs of measures: stage, responsible party, source and confirmation of financing, planned status. The skills formed by Clause 10 are performed at the planning stage.

C.5.9 The skill of Clause 11 updates the actual execution statuses on the graphs of measures and provides the digital environment. When the digital twin is formed, the skill ensures the functioning of its core under C.3.5. The skill forms the derived skill of operational data exchange with external systems and invokes the skills formed by Clause 10 at execution control points.

C.5.10 The skill of Clause 12 conducts monitoring of the achievement of objectives and of impacts. It closes the feedback loop to Clauses 5, 8, 9, 10 and 11: in the event of deviations, the objectives, the vision, the decisions and the plan are revised.

C.5.11 System-wide skills are formed by the parent skill. Management of the sequence of performing the clauses is included in the functions of the parent skill (C.2.2) and is not formalized as a separate skill. Conformity of the skills to the edition of the standard is a derived skill of the parent (Table C.6).

a) control of the integrity of the model — a check of the cascade of dependencies, the normative constraints of protection, the completeness of the traceability chain, the consistency of resource attributes and the admissibility of the parts of the model. The skill does not create nodes. Control of the chain "measure → objective → problem → principle" is performed by the derived skill of Clause 4; the resource balance of the plan — by the derived skill of Clause 10;

b) composition and units of the indicators of Annex B — assignment to domain nodes of the code, the unit and the level of the indicator as characteristics of the node, common to objectives, measures and monitoring. Collection and calculation of the values are performed by the derived skill of Clause 12. An indicator is not formalized as a separate node (C.4.1);

c) access and publication — a role model, machine-readable open data and public versions of the documents of the clauses. Execution statuses and the core of the digital twin are formed by the skill of Clause 11; exchange with external systems — by its derived skill.

C.6.1 Four operations are performed by derived skills upon invocation by the skills of the clauses. The operations are not assigned to a single clause: the skill of Clause 9 forms the means of the impact forecast and of the assessment of implementation; the skills of Clauses 10–12 invoke them at control points of implementation and monitoring.

Table C.7 — Operations with the model

C.6.2 On the basis of monitoring data, the probabilities of hypothetical links are updated according to the number of confirmations, taking into account the quality of the evidence base and the transferability of conclusions to the given territory. Confirmed hypotheses pass into established dependencies; unconfirmed ones are lowered or excluded. Domain attenuation coefficients are calibrated in the same way. Refinements are recorded in the log. Normative blocking links are not lowered.

C.7.1 A clause of the documentation is recognized as AI-ready if the general conditions are met simultaneously:

a) the results of the clause are presented as structured data and are reflected in the graph model to the extent provided for the given clause (C.5.1);

b) links between elements of the clause are indicated explicitly as edges, including links with specific nodes of other parts of the model;

c) the values of domain nodes are bound to territorial units of the level corresponding to the depth of analysis;

d) if the clause sets objectives, they are linked with the domain nodes whose values are used in monitoring;

e) the skill of the clause has been formed and forms the derived skills provided for it.

C.7.2 Additionally, where the corresponding subject matter is present in the clause:

a) graphs of measures have been prepared for the measures;

b) where cultural heritage objects are affected, the assessment under 7.12 has been taken into account.

C.7.3 The specific content and composition of the parts of the graph model may be refined for the given historic settlement, taking into account the availability of source data, the tasks being addressed and the depth of analysis. In doing so, the four named parts of the model and the structure of skill formation are retained.

C.8.1 AI-readiness — a property of the documentation and data under this standard whereby the procedures of the clauses are performed by skills of an information system, and the results are described by a single graph model of the territory.

C.8.2 Skill — a formalized procedure of an information system, including one based on artificial intelligence, with defined source data, results and checking rules, formed on the basis of a clause of the standard or created by another skill.

C.8.3 Structure of skill formation — an organization in which the parent skill forms the skills of the clauses and its own derived skills, and the skills of the clauses form the derived skills of the clause (C.2.2–C.2.4).

C.8.4 Parent skill — a skill that forms the skills of Clauses 4–12, maintains their register and maintains conformity to the current edition of the standard (C.2.2).

C.8.5 Skill of a clause — a skill that forms the content of one clause and has results under C.2.3.

C.8.6 Derived skill — a skill formed by a skill of a clause or by the parent skill and performing operations with the model or with the system: updating, cascading, impact forecast, assessment of implementation, monitoring, traceability control, data exchange (C.2.4).

C.8.7 Graph model of the territory (graph of the urban system) — a single model of four named parts (C.3.2).

C.8.8 Graph of the current state — a part of the model: the current state of the territory, problems and hypotheses of causes, the principles register.

C.8.9 Graph of sustainable development objectives — a part of the model: objectives, tasks, values, mission, interested parties.

C.8.10 Graph of the vision and target states — a part of the model: the image of the future and the desired values of indicators by horizons.

C.8.11 Graph of measures — a part of the model; formed in two stages under C.4.5.

C.8.12 Link between parts of the model — an edge between nodes of different named parts of the model.

C.8.13 Territorial unit of analysis — a node of the spatial layer: a part of the territory with integral characteristics of the domains, delineated at the level corresponding to the depth of analysis.

C.8.14 Spatial weights matrix — a table of the weights of links between territorial units that sets the topological basis for the propagation of influence.

C.8.15 Domain attenuation coefficient — a parameter that can be calibrated and that determines how quickly the influence of a change in the given domain decreases with distance.

C.8.16 Principles register — a structured list of the principles of Clause 4 with stable identifiers, a direction of sustainable development and an implementation mechanism.

C.8.17 Structured data — tables, registers and other machine-readable data sets processed without manual decoding of continuous text. "Results of a clause" means such data, and not only narrative text.

C.10.1 In mode A, parts of the graph model correspond to parts of the master plan. The correspondence does not carry into the master plan the materials that, under 5.1.4, are not included in it.

C.10.2 The sources of the spatial layer are the unified electronic cartographic basis, the state funds of spatial data, the unified state register of cultural heritage objects, the Unified State Register of Real Estate and the state information system for urban-planning activity (6.9.1).

C.10.3 Structured results included in the master plan are prepared for export to the federal state information system for territorial planning and for transfer to the federal state information system "Unified Digital Platform 'National Spatial Data System'" (11.6.2).