ERROR: BoundsError: attempt to access 3-element Array{Float64,1} at index [4]

[danortega2014]

Hello, and thank you for reading my question. I am trying to make a 3x3 matrix pomdp game, and I am receiving the following error:

ERROR: BoundsError: attempt to access 3-element Array{Float64,1} at index [4]
Stacktrace:
 [1] setindex! at .\array.jl:847 [inlined]
 [2] initialize_belief(::BeliefUpdaters.DiscreteUpdater{pomdp3x3}, ::SparseCat{Array{Int64,1},Array{Float64,1}}) at C:\Users\danor\.julia\packages\BeliefUpdaters\Wzj7W\src\discrete.jl:106
 [3] simulate(::StepSimulator, ::pomdp3x3, ::POMDPPolicies.AlphaVectorPolicy{pomdp3x3,Int64}, ::BeliefUpdaters.DiscreteUpdater{pomdp3x3}, ::SparseCat{Array{Int64,1},Array{Float64,1}}, ::Int64) at C:\Users\danor\.julia\packages\POMDPSimulators\oQHQe\src\stepthrough.jl:25 (repeats 2 times)
 [4] simulate at C:\Users\danor\.julia\packages\POMDPSimulators\oQHQe\src\stepthrough.jl:21 [inlined]
 [5] simulate at C:\Users\danor\.julia\packages\POMDPSimulators\oQHQe\src\stepthrough.jl:20 [inlined]
 [6] #stepthrough#30 at C:\Users\danor\.julia\packages\POMDPSimulators\oQHQe\src\stepthrough.jl:212 [inlined]
 [7] top-level scope at REPL[58]:1

Here is my code:

using POMDPs,POMDPModelTools

Base.@kwdef struct pomdp3x3 <: POMDP{Int64, Int64, Int64} # POMDP{State, Action, Observation
    discount_factor::Float64 = 0.95 # discount
end

POMDPs.states(pomdp::pomdp3x3) = [1,2,3]
POMDPs.stateindex(pomdp::pomdp3x3, s::Int64) = s 
POMDPs.actions(::pomdp3x3) = [1,2,3]
POMDPs.actionindex(pomdp::pomdp3x3, a::Int64) = a 

function POMDPs.transition(pomdp::pomdp3x3, s::Int64, a::Int64)
    return SparseCat([1, 2, 3], [.10, .30, .60])
end 

POMDPs.observations(::pomdp3x3) = [1,2,3,4,5,6,7,8,9]
POMDPs.obsindex(::pomdp3x3, o::Int) = o

function POMDPs.observation(pomdp::pomdp3x3, a::Int64, sp::Int64)
    if a == 1  &&  sp == 1
        return SparseCat([1,2,3,4,5,6,7,8,9], [.05,.05,.05,.05,.05,.05,.05,.05,.60]) 
    elseif a== 1 && sp == 2
        return SparseCat([1,2,3,4,5,6,7,8,9], [.05,.05,.05,.05,.05,.05,.05,.60,.05]) 
    elseif a==1 && sp==3
        return SparseCat([1,2,3,4,5,6,7,8,9], [.05,.05,.05,.05,.05,.05,.60,.05,.05]) 
    elseif a == 2  &&  sp == 1
        return SparseCat([1,2,3,4,5,6,7,8,9], [.05,.05,.05,.05,.05,.60,.05,.05,.05])  
    elseif a== 2 && sp == 2
        return SparseCat([1,2,3,4,5,6,7,8,9], [.05,.05,.05,.05,.60,.05,.05,.05,.05]) 
    elseif a==2 && sp==3
        return SparseCat([1,2,3,4,5,6,7,8,9], [.05,.05,.05,.60,.05,.05,.05,.05,.05]) 
    elseif a== 3 && sp == 1
        return SparseCat([1,2,3,4,5,6,7,8,9], [.05,.05,.60,.05,.05,.05,.05,.05,.05]) 
    elseif a==3 && sp==2
        return SparseCat([1,2,3,4,5,6,7,8,9], [.05,.60,.05,.05,.05,.05,.05,.05,.05]) 
    else 
        return SparseCat([1,2,3,4,5,6,7,8,9], [.60,.05,.05,.05,.05,.05,.05,.05,.05]) 
    end
end


function POMDPs.reward(pomdp::pomdp3x3, s::Int64, a::Int64)
    if a == 1  &&  s == 1
        return 0.0
    elseif a== 1 && s == 2
        return 0.0 
    elseif a== 1 && s == 3
        return 0.0 
    elseif a== 2 && s == 1
        return 0.0 
    elseif a== 2 && s == 2
        return 0.0 
    elseif a== 2 && s == 3
        return 0.0 
    elseif a== 3 && s == 1
        return 0.0 
    elseif a== 3 && s == 2
        return 0.0 
    elseif a== 3 && s == 3
        return 0.0 
    end
end

POMDPs.initialstate(::pomdp3x3) = SparseCat([1,2,3,4,5,6,7,8,9], [.11,.11,.11,.11,.11,.11,.11,.11,.12])
POMDPs.discount(pomdp::pomdp3x3) = pomdp.discount_factor


m = pomdp3x3()

using QMDP
solver = QMDPSolver()
policy = POMDPs.solve(solver, m)

using POMDPSimulators


rsum = 0.0
for (s,b,a,o,r) in stepthrough(m, policy, "s,b,a,o,r", max_steps=10)
    println("s: $s, b: $([pdf(b,s) for s in states(m)]), a: $a, o: $o")
    global rsum += r
end
println("Undiscounted reward was $rsum.")

I think the indexing issue is related to either stateindex or actionindex, but I am not sure what to do. I appreciate any suggestions, thanks!

[lassepe]

Your initialstate distribution is a SparseCat over states 1:9 while state space defined in states allows only states 1:3.

[danortega2014]

Thank you!