Merge pull request #18 from JuliaPOMDP/0_6

changes to 0.6

.travis.yml

@@ -2,7 +2,7 @@ language: julia
 os:
   - linux
 julia:
-  - 0.5
+  - 0.6
 notifications:
   email: false
 before_script:

REQUIRE

@@ -1,4 +1,4 @@
-julia 0.5
+julia 0.6
 POMDPs 0.4
 Distributions
 JSON

src/DESPOT.jl

@@ -13,21 +13,21 @@ import POMDPs:
 
 include("history.jl")
 
-abstract DESPOTBeliefUpdate{S,A,O}
+abstract type DESPOTBeliefUpdate{S,A,O} end
 
-typealias DESPOTReward Float64
+const DESPOTReward = Float64
 
-type DESPOTRandomNumber <: POMDPs.AbstractRNG
+mutable struct DESPOTRandomNumber <: POMDPs.AbstractRNG
     number::Float64
 end
 
-type DESPOTParticle{S}
+mutable struct DESPOTParticle{S}
   state::S
   id::Int64
   weight::Float64
 end
 
-type DESPOTBelief{S,A,O}
+mutable struct DESPOTBelief{S,A,O}
     particles::Vector{DESPOTParticle{S}}
     history::History{A,O}
 end
@@ -37,18 +37,18 @@ function rand!(rng::DESPOTRandomNumber, random_number::Array{Float64})
     return nothing
 end
 
-type DESPOTDefaultRNG <: POMDPs.AbstractRNG
+mutable struct DESPOTDefaultRNG <: POMDPs.AbstractRNG
     seed::Array{UInt32,1}
     rand_max::Int64
     debug::Int64
-      
+
     function DESPOTDefaultRNG(seed::UInt32, rand_max::Int64, debug::Int64 = 0)
         this = new()
         this.seed = Cuint[seed]
         this.rand_max = rand_max
         this.debug = debug
         return this
-    end    
+    end
 end
 
 function rand!(rng::DESPOTDefaultRNG, random_number::Array{Float64})
@@ -67,7 +67,7 @@ include("nodes.jl")
 include("utils.jl")
 include("solver.jl")
 
-type DESPOTPolicy{S,A,O,B} <: POMDPs.Policy
+mutable struct DESPOTPolicy{S,A,O,B} <: POMDPs.Policy
     solver::DESPOTSolver{S,A,O,B}
     pomdp ::POMDPs.POMDP{S,A,O}
 end
@@ -79,9 +79,9 @@ create_policy{S,A,O,B}(solver::DESPOTSolver{S,A,O,B}, pomdp::POMDPs.POMDP{S,A,O}
 bounds{S,A,O,B}(bounds::B,
             pomdp::POMDP{S,A,O},
             particles::Vector{DESPOTParticle{S}},
-            config::DESPOTConfig) = 
+            config::DESPOTConfig) =
     error("no bounds() method found for $(typeof(bounds)) type")
-    
+
 init_bounds{S,A,O,B}(bounds::B,
             pomdp::POMDPs.POMDP{S,A,O},
             config::DESPOTConfig) = nothing
@@ -105,7 +105,7 @@ end
 # for any kind of belief besides DESPOTBelief
 function action{S,A,O}(p::DESPOTPolicy{S,A,O}, b)
     N = p.solver.config.n_particles
-    pool = Array(DESPOTParticle{S}, N)
+    pool = Array{DESPOTParticle{S}}(N)
     w = 1.0/N
     for i in 1:N
         pool[i] = DESPOTParticle{S}(rand(p.solver.rng, b), i-1, w)
@@ -130,10 +130,10 @@ end
     @req actions(::P)
     @req generate_sor(::P,::S,::A,::typeof(create_rng(solver.random_streams)))
     @req reward(::P,::S,::A,::S)
-    @req discount(::P)    
+    @req discount(::P)
     @req isterminal(::P,::S)
     as = actions(pomdp)
-    @req iterator(::typeof(as))    
+    @req iterator(::typeof(as))
 end
 
 include("visualization.jl")
@@ -170,5 +170,5 @@ export
     ########## Visualization #######
     blink,
     DESPOTVisualizer
-    
+
 end #module

src/beliefUpdate/beliefUpdateParticle.jl

@@ -1,7 +1,7 @@
 import POMDPs: update, initialize_belief, updater
 using DESPOT
 
-type DESPOTBeliefUpdater{S,A,O} <: POMDPs.Updater
+mutable struct DESPOTBeliefUpdater{S,A,O} <: POMDPs.Updater
     pomdp::POMDP
     num_updates::Int64
     rng::AbstractRNG
@@ -10,15 +10,15 @@ type DESPOTBeliefUpdater{S,A,O} <: POMDPs.Updater
     belief_update_seed::UInt32
     particle_weight_threshold::Float64
     eff_particle_fraction::Float64
-    
+
     #pre-allocated variables (TODO: add the rest at some point)
     n_particles::Int64
     next_state::S
     observation::O
     new_particle::DESPOTParticle{S}
     n_sampled::Int64
     obs_probability::Float64
-    
+
     #default constructor
     function DESPOTBeliefUpdater(pomdp::POMDP{S,A,O};
                                  seed::UInt32 = convert(UInt32, 42),
@@ -28,18 +28,18 @@ type DESPOTBeliefUpdater{S,A,O} <: POMDPs.Updater
                                  eff_particle_fraction::Float64 = 0.05,
                                  next_state::S = S(),
                                  observation::O = O(),
-                                 rng::AbstractRNG=DESPOTDefaultRNG(convert(UInt32, seed $ (n_particles+1)), rand_max)
+                                 rng::AbstractRNG=DESPOTDefaultRNG(convert(UInt32, seed ⊻ (n_particles+1)), rand_max)
                                 )
         this = new()
         this.pomdp = pomdp
-        this.num_updates = 0                               
-        this.belief_update_seed = seed $ (n_particles + 1)       
+        this.num_updates = 0
+        this.belief_update_seed = seed ⊻ (n_particles + 1)
         this.rng = rng
         this.rand_max = rand_max
         this.particle_weight_threshold = particle_weight_threshold
         this.eff_particle_fraction = eff_particle_fraction
         this.n_particles = n_particles
-        
+
         # init preallocated variables
         this.next_state = next_state
         this.observation = observation
@@ -51,34 +51,34 @@ type DESPOTBeliefUpdater{S,A,O} <: POMDPs.Updater
 end
 
 # Special create_belief version for DESPOTBeliefUpdater
-create_belief{S,A,O}(bu::DESPOTBeliefUpdater{S,A,O}) = 
-    DESPOTBelief{S,A,O}(Array(DESPOTParticle{S}, bu.n_particles), History{A,O}())
+create_belief{S,A,O}(bu::DESPOTBeliefUpdater{S,A,O}) =
+    DESPOTBelief{S,A,O}(Array{DESPOTParticle{S}}(bu.n_particles), History{A,O}())
 
-get_belief_update_seed(bu::DESPOTBeliefUpdater) = bu.seed $ (bu.n_particles + 1)
+get_belief_update_seed(bu::DESPOTBeliefUpdater) = bu.seed ⊻ (bu.n_particles + 1)
 
 reset_belief(bu::DESPOTBeliefUpdater) = bu.num_updates = 0
 
 function initialize_belief{S,A,O}(bu::DESPOTBeliefUpdater{S,A,O},
                   state_distribution::ParticleDistribution{S})
-                  
+
     new_belief = create_belief(bu)
     n_particles = length(state_distribution.particles)
-        
+
     # convert to DESPOTParticle type
-    pool = Array(DESPOTParticle{S}, n_particles)
+    pool = Array{DESPOTParticle{S}}(n_particles)
 
     for i in 1:n_particles
         pool[i] = DESPOTParticle{S}(state_distribution.particles[i].state,
                                  i, # id
                                  state_distribution.particles[i].weight)
     end
-    
+
     DESPOT.sample_particles!(new_belief.particles,
                              pool,
                              bu.n_particles,
                              bu.belief_update_seed,
                              bu.rand_max)
-                             
+
     #shuffle!(new_belief.particles) #TODO: uncomment if higher randomness is required
     return new_belief
 end
@@ -87,7 +87,7 @@ end
 function initialize_belief{S}(bu::DESPOTBeliefUpdater{S}, b)
     new_belief = create_belief(bu)
 
-    pool = Array(DESPOTParticle{S}, bu.n_particles)
+    pool = Array{DESPOTParticle{S}}(bu.n_particles)
     w = 1.0/bu.n_particles
     for i in 1:bu.n_particles
         pool[i] = DESPOTParticle{S}(rand(bu.rng, b), i, w)
@@ -109,7 +109,7 @@ updater{S,A,O}(p::DESPOTPolicy{S,A,O}) = DESPOTBeliefUpdater{S,A,O}(p.pomdp,
                                                                     rng=p.solver.rng
                                                                    )
 
-function normalize!{S}(particles::Vector{DESPOTParticle{S}}) 
+function normalize!{S}(particles::Vector{DESPOTParticle{S}})
     prob_sum = 0.0
     for p in particles
         prob_sum += p.weight
@@ -123,12 +123,12 @@ function update{S,A,O}(bu::DESPOTBeliefUpdater{S,A,O},
                 current_belief::DESPOTBelief{S},
                 action::A,
                 obs::O)
-    
+
     updated_belief::DESPOTBelief{S} = create_belief(bu)
     random_number = Array{Float64}(1)
-            
+
     if bu.n_particles != length(current_belief.particles)
-        err("belief size mismatch: belief updater - $(bu.n_particles) particles, belief - $(length(current_belief.particles))")  
+        err("belief size mismatch: belief updater - $(bu.n_particles) particles, belief - $(length(current_belief.particles))")
     end
     updated_belief.particles = []
 
@@ -145,20 +145,20 @@ function update{S,A,O}(bu::DESPOTBeliefUpdater{S,A,O},
         else
             rng = bu.rng
         end
-        
+
         bu.next_state = POMDPs.generate_s(bu.pomdp, p.state, action, rng)
 
         #get observation distribution for (s,a,s') tuple
         od = POMDPs.observation(bu.pomdp, p.state, action, bu.next_state)
-        
+
         bu.obs_probability = pdf(od, obs)
-        
+
         if bu.obs_probability > 0.0
-            bu.new_particle = DESPOTParticle(bu.next_state, p.id, p.weight * bu.obs_probability)            
+            bu.new_particle = DESPOTParticle(bu.next_state, p.id, p.weight * bu.obs_probability)
             push!(updated_belief.particles, bu.new_particle)
         end
     end
-    
+
     normalize!(updated_belief.particles)
 
     if length(updated_belief.particles) == 0
@@ -173,7 +173,7 @@ function update{S,A,O}(bu::DESPOTBeliefUpdater{S,A,O},
             #Pick a random particle from the current belief state as the initial state
             rand!(resample_rng, random_number)
             particle_number = ceil(bu.n_particles * random_number[1])
-            
+
             next_state = POMDPs.rand(resample_rng, states(bu.pomdp)) #generate a random state
             od = POMDPs.observation(bu.pomdp,
                                current_belief.particles[particle_number].state,
@@ -191,7 +191,7 @@ function update{S,A,O}(bu::DESPOTBeliefUpdater{S,A,O},
     end
 
     # Remove all particles below the threshold weight
-    viable_particle_indices = Array(Int64,0)
+    viable_particle_indices = Array{Int64}(0)
     for i in 1:length(updated_belief.particles)
         if updated_belief.particles[i].weight >= bu.particle_weight_threshold
             push!(viable_particle_indices, i)
@@ -213,21 +213,17 @@ function update{S,A,O}(bu::DESPOTBeliefUpdater{S,A,O},
     num_eff_particles = 1./num_eff_particles
     if (num_eff_particles < bu.n_particles * bu.eff_particle_fraction) ||
         (length(updated_belief.particles) < bu.n_particles)
-        resampled_set = Array(DESPOTParticle{S}, bu.n_particles)
-        sample_particles!(resampled_set, 
+        resampled_set = Array{DESPOTParticle{S}}(bu.n_particles)
+        sample_particles!(resampled_set,
                           updated_belief.particles,
                           bu.n_particles,
                           bu.belief_update_seed,
                           bu.rand_max)
         updated_belief.particles = resampled_set
     end
-    
+
     # Finally, update history
     add(updated_belief.history, action, obs)
-    
+
     return updated_belief
 end
-
-
-
-

src/config.jl

@@ -1,4 +1,4 @@
-type DESPOTConfig
+mutable struct DESPOTConfig
   # Maximum depth of the search tree
   search_depth::Int64
   # Random-number seed
@@ -21,10 +21,10 @@ type DESPOTConfig
 #   particle_weight_threshold::Float64
 #   eff_particle_fraction::Float64
   tiny::Float64 # tiny number
-  max_trials::Int64 
-  rand_max::Int64 
+  max_trials::Int64
+  rand_max::Int64
   debug::UInt8
-  
+
   # construct empty
   function DESPOTConfig()
    this = new()

src/history.jl

@@ -1,13 +1,13 @@
-type History{A,O}
+mutable struct History{A,O}
     actions::Vector{A}
     observations::Vector{O}
-    History() = new(Array(A, 0), Array(O, 0))
+    History{A,O}() where {A,O} = new(Array{A}(0), Array{O}(0))
 end
 
 function add{A,O}(history::History{A,O},
             action::A,
             obs::O)
-    
+
     push!(history.actions, action)
     push!(history.observations, obs)
 end

src/lowerBound/DESPOTDefaultLowerBound.jl

@@ -1,14 +1,14 @@
 import DESPOT:
         lower_bound
 
-type DESPOTDefaultLowerBound
+mutable struct DESPOTDefaultLowerBound
     #placeholder for now
 end
 
 function init_bound{S,A,O}(ub::DESPOTDefaultLowerBound,
                     pomdp::POMDP{S,A,O},
                     config::DESPOTConfig)
-    error("Function init_bound for $(typeof(ub)) has not been implemented yet")                    
+    error("Function init_bound for $(typeof(ub)) has not been implemented yet")
 end
 
 function lower_bound{S,A,O}(lb::DESPOTDefaultLowerBound,