Optimize new plumbing (#4112)

Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
Co-authored-by: Yohann Paris <github@yohannparis.com>

packages/client/hmi-client/src/components/workflow/ops/optimize-ciemss/optimize-ciemss-operation.ts

@@ -130,8 +130,7 @@ export const OptimizeCiemssOperation: Operation = {
  {
  type: 'policyInterventionId',
  label: 'Interventions',
- acceptMultiple: false,
- isOptional: true
+ acceptMultiple: false
  }
  ],
  outputs: [{ type: 'simulationId' }],
@@ -168,11 +167,11 @@ export async function getOptimizedInterventions(optimizeRunId: string) {
  const simulation = await getSimulation(optimizeRunId);
  const simulationIntervetions =
  simulation?.executionPayload.fixed_static_parameter_interventions ?? [];
- const policyInterventions = simulation?.executionPayload?.policy_interventions;
- const interventionType = policyInterventions.selection ?? '';
- const paramNames: string[] = policyInterventions.param_names ?? [];
- const paramValue: number[] = policyInterventions.param_values ?? [];
- const startTime: number[] = policyInterventions.start_time ?? [];
+ const optimizeInterventions = simulation?.executionPayload?.optimize_interventions;
+ const interventionType = optimizeInterventions.intervention_type ?? '';
+ const paramNames: string[] = optimizeInterventions.param_names ?? [];
+ const paramValue: number[] = optimizeInterventions.param_values ?? [];
+ const startTime: number[] = optimizeInterventions.start_time ?? [];
 
  const policyResult = await getRunResult(optimizeRunId, 'policy.json');
 

packages/client/hmi-client/src/components/workflow/ops/optimize-ciemss/tera-optimize-ciemss.vue

@@ -320,9 +320,10 @@ import {
  ModelParameter,
  OptimizeRequestCiemss,
  CsvAsset,
- PolicyInterventions,
+ OptimizeInterventions,
  OptimizeQoi,
- InterventionPolicy
+ InterventionPolicy,
+ Intervention
 } from '@/types/Types';
 import { logger } from '@/utils/logger';
 import { chartActionsProxy, drilldownChartSize, nodeMetadata } from '@/components/workflow/util';
@@ -404,6 +405,14 @@ const isSaveDisabled = computed<boolean>(() =>
  isSaveDatasetDisabled(props.node.state.postForecastRunId, useProjects().activeProject.value?.id)
 );
 
+const activePolicyGroups = computed(() =>
+ props.node.state.interventionPolicyGroups.filter((ele) => ele.isActive === true)
+);
+
+const inactivePolicyGroups = computed(() =>
+ props.node.state.interventionPolicyGroups.filter((ele) => ele.isActive === false)
+);
+
 const menuItems = computed(() => [
  {
  label: 'Save as a new model configuration',
@@ -450,11 +459,13 @@ const outputs = computed(() => {
 const isRunDisabled = computed(() => {
  if (
  !props.node.state.constraintGroups?.at(0)?.targetVariable ||
- props.node.state.interventionPolicyGroups.length === 0
+ props.node.state.interventionPolicyGroups.length === 0 ||
+ activePolicyGroups.value.length <= 0
  )
  return true;
  return false;
 });
+
 const selectedOutputId = ref<string>();
 
 const outputViewSelection = ref(OutputView.Charts);
@@ -573,23 +584,36 @@ const runOptimize = async () => {
  const startTime: number[] = [];
  const listInitialGuessInterventions: number[] = [];
  const listBoundsInterventions: number[][] = [];
- props.node.state.interventionPolicyGroups.forEach((ele) => {
+ const initialGuess: number[] = [];
+ const objectiveFunctionOption: string[] = [];
+
+ activePolicyGroups.value.forEach((ele) => {
  paramNames.push(ele.intervention.appliedTo);
  paramValues.push(ele.intervention.staticInterventions[0].value);
  startTime.push(ele.startTime);
+ initialGuess.push(ele.initialGuessValue);
+ objectiveFunctionOption.push(ele.objectiveFunctionOption);
  listInitialGuessInterventions.push(ele.initialGuessValue);
  listBoundsInterventions.push([ele.lowerBoundValue]);
  listBoundsInterventions.push([ele.upperBoundValue]);
  });
  const interventionType = props.node.state.interventionPolicyGroups[0].optimizationType;
 
- const optimizeInterventions: PolicyInterventions = {
+ // These are interventions to be optimized over.
+ const optimizeInterventions: OptimizeInterventions = {
  interventionType,
  paramNames,
  startTime,
- paramValues
+ paramValues,
+ initialGuess,
+ objectiveFunctionOption
  };
 
+ // These are interventions to be considered but not optimized over.
+ const fixedStaticParameterInterventions: Intervention[] = _.cloneDeep(
+ inactivePolicyGroups.value.map((ele) => ele.intervention)
+ );
+
  // TODO: https://github.com/DARPA-ASKEM/terarium/issues/3909
  // The method should be a list but pyciemss + pyciemss service is not yet ready for this.
  const qoi: OptimizeQoi = {
@@ -605,7 +629,8 @@ const runOptimize = async () => {
  start: 0,
  end: knobs.value.endTime
  },
- policyInterventions: optimizeInterventions,
+ optimizeInterventions,
+ fixedStaticParameterInterventions,
  qoi,
  riskBound: props.node.state.constraintGroups[0].threshold, // TODO: https://github.com/DARPA-ASKEM/terarium/issues/3909
  initialGuessInterventions: listInitialGuessInterventions,
@@ -620,9 +645,11 @@ const runOptimize = async () => {
  }
  };
 
+ // InferredParameters is to link a calibration run to this optimize call.
  if (inferredParameters.value) {
  optimizePayload.extra.inferredParameters = inferredParameters.value[0];
  }
+
  const optResult = await makeOptimizeJobCiemss(optimizePayload, nodeMetadata(props.node));
  const state = _.cloneDeep(props.node.state);
  state.inProgressOptimizeId = optResult.simulationId;

packages/client/hmi-client/src/types/Types.ts

@@ -727,8 +727,8 @@ export interface EnsembleSimulationCiemssRequest {
 export interface OptimizeRequestCiemss {
  modelConfigId: string;
  timespan: TimeSpan;
- policyInterventions?: PolicyInterventions;
- fixedStaticParameterInterventions?: string;
+ optimizeInterventions?: OptimizeInterventions;
+ fixedStaticParameterInterventions?: Intervention[];
  stepSize?: number;
  qoi: OptimizeQoi;
  riskBound: number;
@@ -803,16 +803,18 @@ export interface OptimizeExtra {
  solverMethod?: string;
 }
 
-export interface OptimizeQoi {
- contexts: string[];
- method: string;
-}
-
-export interface PolicyInterventions {
+export interface OptimizeInterventions {
  interventionType: string;
  paramNames: string[];
  paramValues?: number[];
  startTime?: number[];
+ objectiveFunctionOption?: string[];
+ initialGuess?: number[];
+}
+
+export interface OptimizeQoi {
+ contexts: string[];
+ method: string;
 }
 
 export interface TimeSpan {

packages/server/src/main/java/software/uncharted/terarium/hmiserver/models/simulationservice/OptimizeRequestCiemss.java

@@ -8,9 +8,10 @@
 import lombok.experimental.Accessors;
 import software.uncharted.terarium.hmiserver.annotations.TSModel;
 import software.uncharted.terarium.hmiserver.annotations.TSOptional;
+import software.uncharted.terarium.hmiserver.models.simulationservice.interventions.Intervention;
 import software.uncharted.terarium.hmiserver.models.simulationservice.parts.OptimizeExtra;
+import software.uncharted.terarium.hmiserver.models.simulationservice.parts.OptimizeInterventions;
 import software.uncharted.terarium.hmiserver.models.simulationservice.parts.OptimizeQoi;
-import software.uncharted.terarium.hmiserver.models.simulationservice.parts.PolicyInterventions;
 import software.uncharted.terarium.hmiserver.models.simulationservice.parts.TimeSpan;
 
 @Data
@@ -25,10 +26,11 @@ public class OptimizeRequestCiemss implements Serializable {
 
  @TSOptional
  // https://github.com/DARPA-ASKEM/pyciemss-service/blob/main/service/models/operations/optimize.py#L80
- private PolicyInterventions policyInterventions;
+ private OptimizeInterventions optimizeInterventions;
 
  @TSOptional
- private UUID fixedStaticParameterInterventions;
+ @JsonAlias("fixed_static_parameter_interventions")
+ private List<Intervention> fixedStaticParameterInterventions;
 
  @JsonAlias("step_size")
  @TSOptional

packages/server/src/main/java/software/uncharted/terarium/hmiserver/models/simulationservice/parts/PolicyInterventions.java → packages/server/src/main/java/software/uncharted/terarium/hmiserver/models/simulationservice/parts/OptimizeInterventions.java

@@ -11,7 +11,7 @@
 @Accessors(chain = true)
 @TSModel
 // Interventions applied by the user within the optimization box.
-public class PolicyInterventions {
+public class OptimizeInterventions {
  // This denotes whether the intervention is on a start date, or a parameter value.
  // https://github.com/DARPA-ASKEM/pyciemss-service/blob/main/service/models/operations/optimize.py#L99
  private String interventionType;
@@ -27,6 +27,14 @@ public class PolicyInterventions {
  @JsonAlias("start_time")
  private List<Integer> startTime;
 
+ @TSOptional
+ @JsonAlias("objective_function_option")
+ private List<String> objectiveFunctionOption;
+
+ @TSOptional
+ @JsonAlias("initial_guess")
+ private List<Double> initialGuess;
+
  @Override
  public String toString() {
  return " { Parameter Names: " + this.paramNames + " start time: " + startTime.toString() + " } ";