chama.optimize module

The optimize module contains high-level solvers for sensor placement optimization.

Contents

ImpactFormulation()

Sensor placement based on minimizing average impact across a set of scenarios.

CoverageFormulation()

Sensor placement based on maximizing coverage of a set of entities.

class chama.optimize.ImpactFormulation[source]

Bases: object

Sensor placement based on minimizing average impact across a set of scenarios.

solve(impact=None, sensor=None, scenario=None, sensor_budget=None, use_sensor_cost=False, use_scenario_probability=False, impact_col_name='Impact', mip_solver_name='glpk', pyomo_options=None, solver_options=None)[source]

Solves the sensor placement optimization by minimizing impact.

Parameters:
  • impact (pandas DataFrame) – Impact assessment. Impact is stored as a pandas DataFrame with columns Scenario, Sensor, and Impact. Each row contains a single detection time (or other measure of impact/damage) for a sensor that detects a scenario. The column name for Impact can also be specified by the user using the argument ‘impact_col_name’.

  • sensor (pandas DataFrame) – Sensor characteristics. Contains sensor cost for each sensor. Sensor characteristics are stored as a pandas DataFrame with columns Sensor and Cost. Cost is used in the sensor placement optimization if the ‘use_sensor_cost’ flag is set to True.

  • scenario (pandas DataFrame) – Scenario characteristics. Contains scenario probability and the impact for undetected scenarios. Scenario characteristics are stored as a pandas DataFrame with columns Scenario, Undetected Impact, and Probability. Undetected Impact is required for each scenario. Probability is used if the ‘use_scenario_probability’ flag is set to True.

  • sensor_budget (float) – The total budget available for purchase/installation of sensors. Solution will select a family of sensors whose combined cost is below the sensor_budget. For a simple sensor budget of N sensors, set this to N and the ‘use_sensor_cost’ to False.

  • use_sensor_cost (bool) – Boolean indicating if sensor cost should be used in the optimization. If False, sensors have equal cost of 1.

  • use_scenario_probability (bool) – Boolean indicating if scenario probability should be used in the optimization. If False, scenarios have equal probability.

  • impact_col_name (str) – The name of the column containing the impact data to be used in the objective function.

  • mip_solver_name (str) – Optimization solver name passed to Pyomo. The solver must be supported by Pyomo and support solution of mixed-integer programming problems.

  • pyomo_options (dict) – Keyword arguments to be passed to the Pyomo solver .solve method Defaults to an empty dictionary.

  • solver_options (dict) – Solver specific options to pass through Pyomo to the underlying solver. Defaults to an empty dictionary.

Returns:

A dictionary with the following keys

  • Sensors: A list of the selected sensors

  • Objective: The mean impact based on the selected sensors

  • FractionDetected: The fraction of scenarios that were detected

  • TotalSensorCost: Total cost of the selected sensors

  • Assessment: The impact value for each sensor-scenario pair. The assessment is stored as a pandas DataFrame with columns Scenario, Sensor, and Impact (same format as the input Impact assessment). If the selected sensors did not detect a particular scenario, the impact is set to the Undetected Impact.

create_pyomo_model(impact=None, sensor=None, scenario=None, sensor_budget=None, use_sensor_cost=False, use_scenario_probability=False, impact_col_name='Impact')[source]

Returns the Pyomo model.

See ImpactFormulation.solve() for more information on parameters.

Returns:

Pyomo ConcreteModel ready to be solved

add_grouping_constraint(sensor_list, select=None, min_select=None, max_select=None)[source]

Adds a sensor grouping constraint to the sensor placement model. This constraint forces a certain number of sensors to be selected from a particular subset of all the possible sensors.

The keyword argument ‘select’ enforces an equality constraint, while ‘min_select’ and ‘max_select’ correspond to lower and upper bounds on the grouping constraints, respectively. You can specify one or both of ‘min_select’ and ‘max_select’ OR use ‘select’

Parameters:
  • sensor_list (list of strings) – List containing the string names of a subset of the sensors

  • select (positive integer or None) – The exact number of sensors from the sensor_list that should be selected

  • min_select (positive integer or None) – The minimum number of sensors from the sensor_list that should be selected

  • max_select (positive integer or None) – The maximum number of sensors from the sensor_list that should be selected

solve_pyomo_model(sensor_budget=None, mip_solver_name='glpk', pyomo_options=None, solver_options=None)[source]

Solves the Pyomo model created to perform the sensor placement.

See ImpactFormulation.solve() for more information on parameters.

create_solution_summary()[source]

Creates a dictionary representing common summary information about the solution from a Pyomo model object that has already been solved.

See ImpactFormulation.solve() for more information on the solution summary.

Returns:

Dictionary containing a summary of results.

class chama.optimize.CoverageFormulation[source]

Bases: object

Sensor placement based on maximizing coverage of a set of entities. An ‘entity’ can represent geographic areas, scenarios, or scenario-time pairs.

solve(coverage, sensor=None, entity=None, sensor_budget=None, use_sensor_cost=None, use_entity_weight=False, redundancy=0, coverage_col_name='Coverage', mip_solver_name='glpk', pyomo_options=None, solver_options=None)[source]

Solves the sensor placement optimization by maximizing coverage.

Parameters:
  • coverage (pandas DataFrame) – Coverage data. Coverage is stored as a pandas DataFrame with columns Sensor and Coverage. Each row contains a list of entities that are covered by single sensor. The column name for Coverage can also be specified by the user using the argument ‘coverage_col_name’.

  • sensor (pandas DataFrame) – Sensor characteristics. Contains sensor cost for each sensor. Sensor characteristics are stored as a pandas DataFrame with columns Sensor and Cost. Cost is used in the sensor placement optimization if the ‘use_sensor_cost’ flag is set to True.

  • entity (pandas DataFrame) – Entity characteristics. Contains entity weights. Entity characteristics are stored as a pandas DataFrame with columns Entity and Weight. Weight is used if the ‘use_entity_weight’ flag is set to True.

  • sensor_budget (float) – The total budget available for purchase/installation of sensors. Solution will select a family of sensors whose combined cost is below the sensor_budget. For a simple sensor budget of N sensors, set this to N and the ‘use_sensor_cost’ to False.

  • use_sensor_cost (bool) – Boolean indicating if sensor cost should be used in the optimization. If False, sensors have equal cost of 1.

  • use_entity_weight (bool) – Boolean indicating if entity weights should be used in the optimization. If False, entities have equal weight.

  • redundancy (int) – Redundancy level. A value of 0 means only one sensor is required to covered an entity, whereas a value of 1 means two sensors must cover an entity before it considered covered.

  • coverage_col_name (str) – The name of the column containing the coverage data to be used in the objective function.

  • mip_solver_name (str) – Optimization solver name passed to Pyomo. The solver must be supported by Pyomo and support solution of mixed-integer programming problems.

  • pyomo_options (dict) – Keyword arguments to be passed to the Pyomo solver .solve method. Defaults to an empty dictionary.

  • solver_options (dict) – Solver specific options to pass through Pyomo to the underlying solver. Defaults to an empty dictionary.

Returns:

A dictionary with the following keys

  • Sensors: A list of the selected sensors

  • Objective: The mean coverage based on the selected sensors

  • FractionDetected: the fraction of entities that are detected

  • TotalSensorCost: Total cost of the selected sensors

  • EntityAssessment: a dictionary whose keys are the entity names, and values are a list of sensors that detect that entity

  • SensorAssessment: a dictionary whose keys are the sensor names, and values are the list of entities that are detected by that sensor

create_pyomo_model(coverage, sensor=None, entity=None, sensor_budget=None, use_sensor_cost=False, use_entity_weight=False, redundancy=0, coverage_col_name='Coverage')[source]

Returns the Pyomo model.

See CoverageFormulation.solve() for more information on parameters.

Returns:

Pyomo ConcreteModel ready to be solved

add_grouping_constraint(sensor_list, select=None, min_select=None, max_select=None)[source]

Adds a sensor grouping constraint to the sensor placement model. This constraint forces a certain number of sensors to be selected from a particular subset of all the possible sensors.

The keyword argument ‘select’ enforces an equality constraint, while ‘min_select’ and ‘max_select’ correspond to lower and upper bounds on the grouping constraints, respectively. You can specify one or both of ‘min_select’ and ‘max_select’ OR use ‘select’

Parameters:
  • sensor_list (list of strings) – List containing the string names of a subset of the sensors

  • select (positive integer or None) – The exact number of sensors from the sensor_list that should be selected

  • min_select (positive integer or None) – The minimum number of sensors from the sensor_list that should be selected

  • max_select (positive integer or None) – The maximum number of sensors from the sensor_list that should be selected

solve_pyomo_model(sensor_budget=None, mip_solver_name='glpk', pyomo_options=None, solver_options=None)[source]

Solves the Pyomo model created to perform the sensor placement.

See CoverageFormulation.solve() for more information on parameters.

create_solution_summary()[source]

Creates a dictionary representing common summary information about the solution from a Pyomo model object that has already been solved.

See CoverageFormulation.solve() for more information on the solution summary.

Returns:

Dictionary containing a summary of results.