Merge pull request #1 from findmyway/master

Add Space & AbstractEnv into base

Author: jbrea

.gitignore

@@ -0,0 +1,6 @@
+*.jl.cov
+*.jl.*.cov
+*.jl.mem
+deps/deps.jl
+
+Manifest.toml

.travis.yml

@@ -0,0 +1,30 @@
+## Documentation: http://docs.travis-ci.com/user/languages/julia/
+language: julia
+os:
+  - linux
+  - osx
+julia:
+  - 1.0
+  - nightly
+notifications:
+  email: false
+git:
+  depth: 99999999
+
+## uncomment the following lines to allow failures on nightly julia
+## (tests will run but not make your overall status red)
+#matrix:
+#  allow_failures:
+#  - julia: nightly
+
+## uncomment and modify the following lines to manually install system packages
+#addons:
+#  apt: # apt-get for linux
+#    packages:
+#    - gfortran
+#before_script: # homebrew for mac
+#  - if [ $TRAVIS_OS_NAME = osx ]; then brew install gcc; fi
+
+## uncomment the following lines to override the default test script
+script:
+  - julia -e 'using Pkg; Pkg.activate(pwd()); Pkg.test()'

Project.toml

@@ -1,6 +1,10 @@
-authors = ["Johanni Brea <[email protected]>"]
 name = "ReinforcementLearningBase"
 uuid = "9b2b9cba-ac73-11e8-02b1-9f0869453fc0"
+authors = ["Johanni Brea <[email protected]>", "Jun Tian <[email protected]>"]
 version = "0.1.0"
 
-[deps]
+[extras]
+Test = "8dfed614-e22c-5e08-85e1-65c5234f0b40"
+
+[targets]
+test = ["Test"]

README.md

@@ -0,0 +1,5 @@
+# ReinforcementLearningBase.jl
+
+[![Build Status](https://travis-ci.com/JuliaReinforcementLearning/ReinforcementLearningBase.jl.svg?branch=master)](https://travis-ci.com/JuliaReinforcementLearning/ReinforcementLearningBase.jl)
+
+ReinforcementLearningBase.jl holds the common types and utility functions to be shared by other components in ReinforcementLearning ecosystem.

REQUIRE

@@ -0,0 +1 @@
+julia 1.0

src/ReinforcementLearningBase.jl

@@ -1,5 +1,6 @@
 module ReinforcementLearningBase
 
-greet() = print("Hello World!")
+include("spaces/space.jl")
+include("abstractenv.jl")
 
 end # module

src/abstractenv.jl

@@ -0,0 +1,17 @@
+export AbstractEnv
+abstract type AbstractEnv end
+
+"Get current state of an environment"
+function getstate end
+
+"Reset an environment to the initial state"
+function reset! end
+
+"Take an action in an environment and return the current state"
+function interact! end
+
+"Get the action space of an environment"
+function actionspace end
+
+"Plot the current state of an environment"
+function plotenv end

src/spaces/abstractspace.jl

@@ -0,0 +1,15 @@
+abstract type AbstractSpace end
+
+"""
+    sample(s::AbstractSpace)
+
+Get a random sample from `s`.
+"""
+function sample end
+
+"""
+    occursin(x, s::AbstractSpace)
+
+Return wheather `x` is a valid sample in space `s`.
+"""
+function occursin end

src/spaces/boxspace.jl

@@ -0,0 +1,34 @@
+"A box in R^n."
+struct BoxSpace{T <: Number,N} <: AbstractSpace
+    low::Array{T,N}
+    high::Array{T,N}
+end
+
+size(s::BoxSpace) = size(s.low)
+
+"""
+    BoxSpace(low::Number, high::Number, size::Tuple{Vararg{Int}}=(1,))
+
+```julia
+BoxSpace(-1, 1)
+BoxSpace(-1, 1, (2,3))
+```
+"""
+BoxSpace(low::Number, high::Number, size::Tuple{Vararg{Int}}=(1,)) = BoxSpace(fill(low, size), fill(high, size))
+
+"""
+    BoxSpace(low::Array{<:Number}, high::Array{<:Number})
+
+```julia
+BoxSpace([0, 0, 0], [1, 2, 3])
+```
+"""
+BoxSpace(low::Array{<:Number}, high::Array{<:Number}) = size(low) == size(high) && BoxSpace(low, high)
+
+==(x::BoxSpace, y::BoxSpace) = x.low == y.low && x.high == y.high
+
+sample(s::BoxSpace) = map((l, h) -> l + rand() * (h - l), s.low, s.high)
+
+occursin(xs::Array{<:Number}, s::BoxSpace) = size(s) == size(xs) && 
+    all(map((a, b, c) -> a ≤ b ≤ c, s.low, xs, s.high))
+occursin(x::Number, s::BoxSpace) = all(map((l, h) -> l ≤ x ≤ h, s.low, s.high))

src/spaces/discretespace.jl

@@ -0,0 +1,9 @@
+struct DiscreteSpace <: AbstractSpace
+    n::Int
+    offset::Int
+end
+
+size(d::DiscreteSpace) = 1
+sample(d::DiscreteSpace) = rand(d.offset : d.n + d.offset - 1)
+occursin(x::Int, d::DiscreteSpace) = d.offset ≤ x < d.offset + d.n
+==(x::DiscreteSpace, y::DiscreteSpace) = x.n == y.n && x.offset == y.offset

src/spaces/multibinaryspace.jl

@@ -0,0 +1,11 @@
+struct MultiBinarySpace <: AbstractSpace
+    size::Tuple{Vararg{Int}}
+end
+
+MultiBinarySpace(sz::Vararg{Int}) = MultiBinarySpace(sz)
+
+size(s::MultiBinarySpace) = s.size
+
+sample(s::MultiBinarySpace) = rand(Bool, s.size...)
+occursin(x::Array{Bool}, s::MultiBinarySpace) = size(s) == size(x)
+==(x::MultiBinarySpace, y::MultiBinarySpace) = x.size == y.size

src/spaces/multidiscretespace.jl

@@ -0,0 +1,13 @@
+struct MultiDiscreteSpace{N} <:AbstractSpace
+    counts::Array{Int, N}
+    offset::Int
+end
+
+size(s::MultiDiscreteSpace) = size(s.counts)
+
+"To compat with Python, here we start with 0"
+sample(s::MultiDiscreteSpace) = map(x -> rand(s.offset : x + s.offset -1), s.counts)
+occursin(x::Int, s::MultiDiscreteSpace) = all(e -> s.offset ≤ x < e + s.offset, s.counts)
+occursin(xs::Array{Int}, s::MultiDiscreteSpace) = size(s) == size(xs) &&
+    all(map((e, x) -> s.offset ≤ x < e + s.offset , s.counts, xs))
+==(x::MultiDiscreteSpace, y::MultiDiscreteSpace) = x.counts == y.counts && x.offset == y.offset

src/spaces/space.jl

@@ -0,0 +1,26 @@
+import Base:occursin, size, ==
+export AbstractSpace,
+       BoxSpace,
+       DiscreteSpace,
+       MultiBinarySpace,
+       MultiDiscreteSpace,
+       sample
+
+include("abstractspace.jl")
+include("boxspace.jl")
+include("discretespace.jl")
+include("multibinaryspace.jl")
+include("multidiscretespace.jl")
+
+# Tuple Support
+sample(s::Tuple{Vararg{<:AbstractSpace}}) = map(sample, s)
+occursin(a::Tuple, b::Tuple{Vararg{<:AbstractSpace}}) = length(a) == length(b) &&
+    all(map((x, y) -> occursin(x, y), a, b))
+
+# Dict Support
+sample(s::Dict{String}) = Dict(map((k, v) -> (k, sample(v)), s))
+occursin(a::Dict{String}, b::Dict{String}) = length(a) == length(b) &&
+    all(p -> haskey(a, p.first) ? 
+            occursin(a[p.first], p.second) :
+            false,
+        b)

test/runtests.jl

@@ -0,0 +1,4 @@
+using ReinforcementLearningBase
+using Test
+
+include("space.jl")

test/space.jl

@@ -0,0 +1,76 @@
+@testset "Space" begin
+
+@testset "BoxSpace" begin
+    @test occursin(0.5, BoxSpace(0, 1)) == true
+    @test occursin(0.0, BoxSpace(0, 1)) == true
+    @test occursin(1.0, BoxSpace(0, 1)) == true
+    @test occursin(-1.0, BoxSpace(0, 1)) == false
+    @test occursin(-Inf, BoxSpace(0, 1)) == false
+    @test occursin([0.5], BoxSpace(0, 1)) == true
+
+    @test occursin([0, 0], BoxSpace([-1, -2], [1, 2])) == true
+    @test occursin([0, 3], BoxSpace([-1, -2], [1, 2])) == false
+    @test occursin([0, 0], BoxSpace([-1, -2], [1, 2])) == true
+end
+
+@testset "DiscreteSpace" begin
+    @test occursin(0, DiscreteSpace(10, 0)) == true
+    @test occursin(5, DiscreteSpace(10, 0)) == true
+    @test occursin(10, DiscreteSpace(10, 0)) == false
+end
+
+@testset "MultiBinarySpace" begin
+    @test occursin([true false; true false], MultiBinarySpace(2,2)) == true
+    @test occursin([true false], MultiBinarySpace(2,2)) == false
+end
+
+@testset "MultiDiscreteSpace" begin
+    @test occursin(0, MultiDiscreteSpace([2,3,2], 0)) == true
+    @test occursin(1, MultiDiscreteSpace([2,3,2], 0)) == true
+    @test occursin(2, MultiDiscreteSpace([2,3,2], 0)) == false
+    @test occursin([1,1,1], MultiDiscreteSpace([2,3,2], 0)) == true
+    @test occursin([0,0,0], MultiDiscreteSpace([2,3,2], 0)) == true
+    @test occursin([3,3,3], MultiDiscreteSpace([2,3,2], 0)) == false
+end
+
+@testset "Space Tuple" begin
+    @test occursin(([0.5], 5, [true true; true true], [1, 1]),
+                   (BoxSpace(0,1), DiscreteSpace(5, 0), MultiBinarySpace(2,2), MultiDiscreteSpace([2,2], 0))) == false
+    @test occursin(([0.5], 0, [true true; true true], [1, 1]),
+                   (BoxSpace(0,1), DiscreteSpace(5, 0), MultiBinarySpace(2,2), MultiDiscreteSpace([2,2], 0))) == true
+    @test occursin((), 
+                   (BoxSpace(0,1), DiscreteSpace(5, 0), MultiBinarySpace(2,2), MultiDiscreteSpace([2,2], 0))) == false
+end
+
+@testset "Space Dict" begin
+    @test occursin(
+        Dict(
+            "sensors" => Dict(
+                "position" => [-10, 0, 10],
+                "velocity" => [0.1, 0.2, 0.3],
+                "front_cam" => (rand(10, 10, 3), rand(10, 10, 3)),
+                "rear_cam" => rand(10,10,3)),
+            "ext_controller" => [2, 1, 1],
+            "inner_state" => Dict(
+                "charge" => 35,
+                "system_checks" => rand(Bool, 10),
+                "job_status" => Dict(
+                    "task" => 3,
+                    "progress" => 23))),
+        Dict(
+            "sensors" =>  Dict(
+                "position"=> BoxSpace(-100, 100, (3,)),
+                "velocity"=> BoxSpace(-1, 1, (3,)),
+                "front_cam"=> (BoxSpace(0, 1, (10, 10, 3)),
+                               BoxSpace(0, 1, (10, 10, 3))),
+                "rear_cam" => BoxSpace(0, 1, (10, 10, 3))),
+            "ext_controller" => MultiDiscreteSpace([5, 2, 2], 0),
+            "inner_state" => Dict(
+                "charge" => DiscreteSpace(100, 0),
+                "system_checks" => MultiBinarySpace(10),
+                "job_status" => Dict(
+                    "task" => DiscreteSpace(5, 0),
+                    "progress" => BoxSpace(0, 100))))) == true
+end
+
+end