This directory contains various data structures that are expected to be returned by the content server. I document them by the URL endpoint that the server will be answering in the sections below. All of these requests are made with an HTTP GET request, which means the server should not carry any state. This implies the server will be "dumb" and never remember requests, sessions, or prior optimization results.

Returns the configuration for the visualization as JSON, including the starting reward values, policy values, and transition function values.

Path

/initialize

Parameters:

none

Returns:

{
  "reward": [{"name": "NAME", "current_value": 11111, "max": 999999, "min": 0, "units": "$"},...]
  "transition": [{"name": "NAME", "current_value": 11111, "max": 999999, "min": 0, "units": "$"},...]
  "policy": [{"name": "NAME", "current_value": 11111, "max": 999999, "min": 0, "units": "$"},...]
}

The name is both the display name for the parameter and the identifier when sent in subsequent requests to the web server.

The description is the long form description of the parameter value.

The current value is the value the parameter should be initialized to in the interface.

The max and the min values are the maximum and minimum values the visualization should allow for that parameter.

The visualization sends the policy it wants trajectories from along with any specifiable transition function parameters.

Path

/trajectories?policy[POLICY_PARAM1]=X&...&transition[TRANSITION_FUNCTION_PARAM1]=I&...

Parameters

policy[PARAM]: a list of policy parameters and their values.
transition[PARAM]: a list of transition parameters and their values.

Returns

Gives the trajectories object.

[
  [{"variable name":"variable value", ...}, {"variable name":"variable value", ...}],
  [{"variable name":"variable value", ...}, {"variable name":"variable value", ...}],
  ...
]

The visualization requests all the hidden state variables for a particular state. For wildfire this is the complete landscape object.

Path

/state?event=ID

Parameters

event: an identifier for the particular event to be shown. This identifier should be included in the trajectories and it could include any of the elements required to re-generate the hidden state variables.

Returns

TODO: specify this.

Optimize a new policy from the currently specified policy

Path

/optimize?policy[POLICY_PARAM1]=X&...&transition[TRANSITION_FUNCTION_PARAM1]=I&...&reward[PARAM]

(post current rewards, transition function, and current policy return same object as get trajectories)

Parameters

policy[PARAM]: a list of policy parameters and their values.
transition[PARAM]: a list of transition parameters and their values. reward[PARAM]: a list of reward parameters and their values.

Returns:

{
  "reward": [{"name": "NAME", "current_value": 11111, "max": 999999, "min": 0, "units": "$"},...]
  "transition": [{"name": "NAME", "current_value": 11111, "max": 999999, "min": 0, "units": "$"},...]
  "policy": [{"name": "NAME", "current_value": 11111, "max": 999999, "min": 0, "units": "$"},...]
}

The visualization should then request trajectories from the returned policy.