Type Mismatch for observation state

[Bala-1998]

Please check this file for how I try to create and then use POMCPOW - https://github.com/Bala-1998/pomdp_jl/blob/main/discrete_drive_pomdp/drive_test_jl.ipynb

I have defined a structure for the observation as:

struct driveObservation
h::Int
r::Int
a::Int
end

and I have defined the observation model as this -

function POMDPs.observation(pomdp::drivePOMDP, act::Int64, state::driveState)
h, r, z, phi = state.h, state.r, state.z, state.phi
a = act
p_stop = human_policy(h, r, z)

# If the action to stop is taken by the robot, the human stops at a probability of p_stop this distribution needs to be returned for both stop action and go action.
if a == 0
    if r == 0 && h == 0
        return SparseCat([driveObservation(0,0,a), driveObservation(1,0,a)], [p_stop, 1-p_stop])
    else
        return Deterministic(1.0e-9)
    end
elseif a == 1
    if r  == 0 && h == 0
        return SparseCat([driveObservation(0,1,a), driveObservation(1,1,a)], [p_stop, 1-p_stop])
    else
        return Deterministic(1.0e-9)
    end
else
    return Deterministic(1.0e-9)
end

end

But when I try to solve using the pomcpow library, I get an error like this-
I get the following stack trace which has to do with driveObservation - MethodError: Cannot convert an object of type Float64 to an object of type driveObservation

Closest candidates are:
convert(::Type{T}, !Matched::T) where T
@ Base Base.jl:84
driveObservation(::Any, !Matched::Any, !Matched::Any)
@ Main ~/Fall23/aizen/pomdp_test/drive_test_jl.ipynb:2

Stacktrace:
[1] push!(a::Vector{driveObservation}, item::Float64)
@ Base ./array.jl:1118
[2] insert_obs_node!(t::BasicPOMCP.POMCPTree{Int64, driveObservation}, pomdp::drivePOMDP, ha::Int64, sp::driveState, o::Float64)
@ BasicPOMCP ~/.julia/packages/BasicPOMCP/ba47d/src/BasicPOMCP.jl:183
[3] simulate(p::POMCPPlanner{drivePOMDP, BasicPOMCP.SolvedPORollout{RandomPolicy{Random._GLOBAL_RNG, drivePOMDP, NothingUpdater}, NothingUpdater, Random._GLOBAL_RNG}, Random._GLOBAL_RNG}, s::driveState, hnode::BasicPOMCP.POMCPObsNode{Int64, driveObservation}, steps::Int64)
@ BasicPOMCP ~/.julia/packages/BasicPOMCP/ba47d/src/solver.jl:94
[4] search(p::POMCPPlanner{drivePOMDP, BasicPOMCP.SolvedPORollout{RandomPolicy{Random._GLOBAL_RNG, drivePOMDP, NothingUpdater}, NothingUpdater, Random._GLOBAL_RNG}, Random._GLOBAL_RNG}, b::SparseCat{Vector{driveState}, Vector{Float64}}, t::BasicPOMCP.POMCPTree{Int64, driveObservation}, info::Dict{Symbol, Any})
@ BasicPOMCP ~/.julia/packages/BasicPOMCP/ba47d/src/solver.jl:32
[5] action_info(p::POMCPPlanner{drivePOMDP, BasicPOMCP.SolvedPORollout{RandomPolicy{Random._GLOBAL_RNG, drivePOMDP, NothingUpdater}, NothingUpdater, Random._GLOBAL_RNG}, Random._GLOBAL_RNG}, b::SparseCat{Vector{driveState}, Vector{Float64}}; tree_in_info::Bool)
@ BasicPOMCP ~/.julia/packages/BasicPOMCP/ba47d/src/solver.jl:6
[6] action_info
@ BasicPOMCP ~/.julia/packages/BasicPOMCP/ba47d/src/solver.jl:1 [inlined]
[7] action(p::POMCPPlanner{drivePOMDP, BasicPOMCP.SolvedPORollout{RandomPolicy{Random._GLOBAL_RNG, drivePOMDP, NothingUpdater}, NothingUpdater, Random._GLOBAL_RNG}, Random._GLOBAL_RNG}, b::SparseCat{Vector{driveState}, Vector{Float64}})
@ BasicPOMCP ~/.julia/packages/BasicPOMCP/ba47d/src/solver.jl:19
[8] top-level scope
@ ~/Fall23/aizen/pomdp_test/drive_test_jl.ipynb:2

How Do I solve this method error mismatch?

[zsunberg]

You should return Deterministic(::driveObservation) instead of Deterministic(::Float64) from observation. Does that make sense?

[Bala-1998]

Thank you! I was able to solve the issue.