Required Interface

reset!(env::AbstractEnv)

Reset env to its initial state and return nothing.

This is a required function that must be provided by every AbstractEnv.

actions(env::AbstractEnv)

Return a collection of all of the actions available for AbstractEnv env. If the environment has multiple agents, this function should return the union of all of their actions.

This is a required function that must be provided by every AbstractEnv.

This function is a static property of the environment; the value it returns should not change based on the state.

actions(env::AbstractEnv, player_index)

Return a collection of all the actions available to a given player.

This function is a static property of the environment; the value it returns should not change based on the state.

observe(env::AbstractEnv)

Return an observation from the environment for the current player.

This is a required function that must be provided by every AbstractEnv.

r = act!(env::AbstractEnv, a)

Take action a and advance AbstractEnv env forward one step, and return rewards for all players.

This is a required function that must be provided by every AbstractEnv.

If the environment has a single player, it is acceptable to return a scalar number. If there are multiple players, it should return a container with all rewards indexed by player number.

Example

Single Player

function act!(env::MyMDPEnv, a)
    env.state += a + randn()
    return env.s^2
end

Two Player

function act!(env::MyMDPEnv, a)
    env.positions[player(env)] += a   # In this game, each player has a position that is updated by his or her action
    rewards = in_goal.(env.positions) # Rewards are +1 for being in a goal region, 0 otherwise
    return rewards                    # returns a vector of rewards for each player
end

terminated(env::AbstractEnv)

Determine whether an environment has finished executing.

If terminated(env) is true, no further actions should be taken and it is safe to assume that no further rewards will be received.