Update to julia 1.0 (#12)

.travis.yml

@@ -4,7 +4,12 @@ os:
   - linux
   - osx
 julia:
-  - 0.6
+  - 1.0
+  - 1.1
+  - nightly
+matrix:
+  allow_failures:
+    - julia: nightly
 notifications:
   email: false
 # uncomment the following lines to override the default test script
@@ -13,6 +18,6 @@ notifications:
 #  - julia -e 'Pkg.clone(pwd()); Pkg.build("DynamicMovementPrimitives"); Pkg.test("DynamicMovementPrimitives"; coverage=true)'
 after_success:
   # push coverage results to Coveralls
-  - julia -e 'cd(Pkg.dir("DynamicMovementPrimitives")); Pkg.add("Coverage"); using Coverage; Coveralls.submit(Coveralls.process_folder())'
+  - julia -e 'using Pkg;import ControlSystems; cd(joinpath(dirname(pathof(DynamicMovementPrimitives)), "..")); Pkg.add("Coverage"); using Coverage; Coveralls.submit(Coveralls.process_folder())'
   # push coverage results to Codecov
-  - julia -e 'cd(Pkg.dir("DynamicMovementPrimitives")); Pkg.add("Coverage"); using Coverage; Codecov.submit(Codecov.process_folder())'
+  - julia -e 'using Pkg;import ControlSystems; cd(joinpath(dirname(pathof(DynamicMovementPrimitives)), "..")); Pkg.add("Coverage"); using Coverage; Codecov.submit(Codecov.process_folder())'

README.md

@@ -1,7 +1,5 @@
 # DynamicMovementPrimitives
 
-[![DynamicMovementPrimitives](http://pkg.julialang.org/badges/DynamicMovementPrimitives_0.4.svg)](http://pkg.julialang.org/?pkg=DynamicMovementPrimitives)
-[![DynamicMovementPrimitives](http://pkg.julialang.org/badges/DynamicMovementPrimitives_0.5.svg)](http://pkg.julialang.org/?pkg=DynamicMovementPrimitives)
 [![DynamicMovementPrimitives](http://pkg.julialang.org/badges/DynamicMovementPrimitives_0.6.svg)](http://pkg.julialang.org/?pkg=DynamicMovementPrimitives)
 [![Build Status](https://travis-ci.org/baggepinnen/DynamicMovementPrimitives.jl.svg?branch=master)](https://travis-ci.org/baggepinnen/DynamicMovementPrimitives.jl)
 
@@ -10,8 +8,7 @@ Provides implementations of Ijspeert et al. 2013 and of Martin Karlsson, Fredrik
 ## Installation
 
 ```julia
-Pkg.add("DynamicMovementPrimitives")
-Pkg.checkout("DynamicMovementPrimitives") # Recommended to get the latest version
+using Pkg; Pkg.add("DynamicMovementPrimitives")
 using DynamicMovementPrimitives
 ```
 
@@ -24,13 +21,13 @@ Nbasis  = 15
 αx      = 1.
 opts    = DMPopts(Nbasis,αx,αz)
 
-y       = [zeros(10);linspace(0,2,1000); 2ones(500)]
+y       = [zeros(10);LinRange(0,2,1000); 2ones(500)]
 T       = length(y)
-t       = linspace(0,10,T)
+t       = LinRange(0,10,T)
 h       = t[2]-t[1] # Sample interval
 y       = [y 0.5y]
-ẏ       = centraldiff(y) /h # Differentiate position to get velocity
-ÿ       = centraldiff(ẏ) /h
+ẏ       = centraldiff(y) ./h # Differentiate position to get velocity
+ÿ       = centraldiff(ẏ) ./h
 dmp     = fit(y,ẏ,ÿ,t,opts)
 
 tout,yout,ẏout,xout = solve(dmp,t) # Generate trajectory, see ?solve for options
@@ -59,34 +56,24 @@ plot(dmp2) # Requires Plots.jl, plots the trajectory from solve with default opt
 plot(dmp2,true)
 ```
 
-We test the performance of the 2DOF controller by implementing a custom solver. The solver `euler_disturbance` is the same as the built in solver `euler`, but between `i=500` and `i=700`, we stop the evolution of the physical system by setting `ẏa = 0`.
+We test the performance of the 2DOF controller by implementing a solver callback. Between `t=2.5` and `t=4`, we stop the evolution of the physical system by setting `ẏa = 0` through `u[3] = uprev[3]`.
 ```julia
-function euler_disturbance(time_derivative, state0, t, args...; kwargs...)
-    T = length(t)
-    n = length(state0)
-    res = Matrix{Float64}(T,n)
-    res[1,:] = state0
-    for i in 2:T
-        td = time_derivative(t[i-1],res[i-1,:])*(t[i]-t[i-1])
-        if 400 <= i < 600 # Hold ẏa still
-            td[3] = 0
-        end
-        res[i,:] = res[i-1,:] + td
-    end
-    t,res
-end
+import OrdinaryDiffEq
+condition(u,t,integrator) = 2.5 <= t < 4
+affect!(integrator) = (integrator.u[3] = integrator.uprev[3])
+cb = OrdinaryDiffEq.DiscreteCallback(condition,affect!)
 ```
 
-We can call the solve method with our custom solver and plot the result. It should be clear from the figures that this time, the coupled signal `yc` slows down when there is a nonzero error.
+We can call the solve method with our custom callback and plot the result. It should be clear from the figures that this time, the coupled signal `yc` slows down when there is a nonzero error.
 ```julia
-t,yc,ẏc,x,ya,ẏa,e = solve(dmp2,t, solver=euler_disturbance)
+t,yc,ẏc,x,ya,ẏa,e = solve(dmp2,t, solver=OrdinaryDiffEq.Euler(), callback=cb)
 plot(t,ẏc, lab="\$ẏ_c\$", c=:red, l=(:dash, 3), layout=(2,2), subplot=1)
 plot!(t,yc, lab="\$y_c\$", c=:red, l=(:dash, 3), subplot=2)
 plot!(t,ẏa, lab="\$ẏ_a\$", c=:blue, subplot=1)
 plot!(t,ya, lab="\$y_a\$", c=:blue, subplot=2)
 plot!(t,e, lab="\$e\$", c=:green, subplot=3)
 plot!(t,400 .<= 1:T .< 600, lab="Disturbance", c=:green, subplot=4, fillrange=0)
-t,yc,ẏc,x,ya,ẏa,e = solve(dmp2,t, solver=euler)
+t,yc,ẏc,x,ya,ẏa,e = solve(dmp2,t)
 plot!(t,ẏc, lab="\$ẏ_u\$", c=:black, l=(:dashdot, 3), subplot=1)
 plot!(t,yc, lab="\$y_u\$", c=:black, l=(:dashdot, 3), subplot=2)
 ```

REQUIRE

@@ -1,4 +1,4 @@
-julia 0.6.0
+julia 1.0
 OrdinaryDiffEq
 RecipesBase 0.2.3
-Requires
+Requires

src/DMP2dof.jl

@@ -54,7 +54,7 @@ function _acceleration(d,f, yc,ẏc,x,ya,ẏa,e,g)
     ẏc,ÿc,ẏa,ÿa,ė,ẋ
 end
 
-function solve_canonical(dmp::DMP2dof, t, y0, g, solver)
+function solve_canonical(dmp::DMP2dof, t, y0, g, solver; kwargs...)
     T,n = size(dmp.y)
     αx  = dmp.opts.αx
     τ   = dmp.τ
@@ -82,13 +82,13 @@ function solve_canonical(dmp::DMP2dof, t, y0, g, solver)
         end
         state0 = [y0[i], dmp.ẏ[1,i], y0[i], dmp.ẏ[1,i], 0, 1.]
         prob = OrdinaryDiffEq.ODEProblem(time_derivative,state0,(t[1],t[end]))
-        sol = OrdinaryDiffEq.solve(prob,solver,saveat=t,dt=t[2])
-        yc[:,i]  = sol[1,:]
-        ẏc[:,i]  = sol[2,:]
-        ya[:,i]  = sol[3,:]
-        ẏa[:,i]  = sol[4,:]
-        e[:,i]   = sol[5,:]
-        x[:]     = sol[6,:] # TODO: se till att denna är samma för alla DOF
+        sol = OrdinaryDiffEq.solve(prob,solver;saveat=t,dt=t[2], kwargs...)
+        yc[:,i]  = sol(t,idxs=1)
+        ẏc[:,i]  = sol(t,idxs=2)
+        ya[:,i]  = sol(t,idxs=3)
+        ẏa[:,i]  = sol(t,idxs=4)
+        e[:,i]   = sol(t,idxs=5)
+        x[:]     = sol(t,idxs=6) # TODO: se till att denna är samma för alla DOF
     end
     t,yc,ẏc,x,ya,ẏa,e
 end

src/DynamicMovementPrimitives.jl

@@ -1,5 +1,5 @@
 module DynamicMovementPrimitives
-using Requires, RecipesBase
+using LinearAlgebra, RecipesBase
 import OrdinaryDiffEq
 
 export DMP, DMPopts, centraldiff,fit, solve, force, acceleration, solve_canonical, kernel_vector, plotdmp, plotdmpphase, euler
@@ -113,7 +113,7 @@ function fit(y,ẏ,ÿ,t,opts,g=y[end,:][:])
         else # LWR
             for j = 1:Nbasis
                 sTΓ = ξ[:,i].*Ψ[:,j]
-                w[j,i] = vecdot(sTΓ,ft[:,i])/vecdot(sTΓ,ξ[:,i])
+                w[j,i] = dot(sTΓ,ft[:,i])/dot(sTΓ,ξ[:,i])
             end
         end
     end
@@ -124,7 +124,7 @@ function force_single(d::AbstractDMP,x::Number, i)
     # ODE Point case
     y0 = _1(d)
     Ψ  = kernel_vector(x, d.c, d.σ2)
-    f = vecdot(Ψ,d.w[:,i]) * x*(d.g[i]-y0[i])
+    f = dot(Ψ,d.w[:,i]) * x*(d.g[i]-y0[i])
 end
 
 function force_single(d::AbstractDMP,x::Number)
@@ -145,7 +145,7 @@ function force_multiple(d::AbstractDMP,x::Number, i)
     # ODE Point case
     y0 = _1(d)
     Ψ  = kernel_vector(x, d.c[:,i], d.σ2[:,i])
-    f  = vecdot(Ψ,d.w[:,i]) * x*(d.g[i]-y0[i])
+    f  = dot(Ψ,d.w[:,i]) * x*(d.g[i]-y0[i])
 end
 
 function force_multiple(d::AbstractDMP,x::AbstractVector)
@@ -155,7 +155,7 @@ function force_multiple(d::AbstractDMP,x::AbstractVector)
     f = zeros(n)
     for i = 1:n
         Ψ    = kernel_matrix([x[i]], d.c[:,i], d.σ2[:,i])
-        f[i] = vecdot(Ψ,d.w[:,i]) * x[i].*(d.g[i]-y0[i])
+        f[i] = dot(Ψ,d.w[:,i]) * x[i].*(d.g[i]-y0[i])
     end
     f
 end
@@ -209,7 +209,7 @@ function acceleration(d::DMP, y::AbstractMatrix,ẏ::AbstractMatrix,x::AbstractV
     (d.opts.αz*(d.opts.βz*(g'.-y)-d.τ*ẏ)+f)/d.τ^2
 end
 
-function solve_canonical(dmp::DMP, t, y0, g, solver)
+function solve_canonical(dmp::DMP, t, y0, g, solver; kwargs...)
     T,n = size(dmp.y)
     αx  = dmp.opts.αx
     τ   = dmp.τ
@@ -227,15 +227,15 @@ function solve_canonical(dmp::DMP, t, y0, g, solver)
         end
         state0 = [dmp.ẏ[1,i]; y0[i]; 1.]
         prob = OrdinaryDiffEq.ODEProblem(time_derivative,state0,(t[1],t[end]))
-        sol = OrdinaryDiffEq.solve(prob,solver,saveat=t,dt=t[2])
+        sol = OrdinaryDiffEq.solve(prob,solver;saveat=t,dt=t[2], kwargs...)
         z[:,i] = sol[1,:]
         y[:,i] = sol[2,:]
         x      = sol[3,:] # TODO: se till att denna är samma för alla DOF
     end
     t,y,z,x
 end
 
-function solve_position(dmp, t, y0, g, solver)
+function solve_position(dmp, t, y0, g, solver; kwargs...)
     T,n = size(dmp.y)
     αx  = dmp.opts.αx
     τ   = dmp.τ
@@ -250,14 +250,14 @@ function solve_position(dmp, t, y0, g, solver)
         end
         state0  = [0; y0[i]]
         prob = OrdinaryDiffEq.ODEProblem(time_derivative,state0,(t[1],t[end]))
-        sol = OrdinaryDiffEq.solve(prob,solver,saveat=t,dt=t[2])
+        sol = OrdinaryDiffEq.solve(prob,solver;saveat=t,dt=t[2], kwargs...)
         z[:,i] = sol[1,:]
         y[:,i] = sol[2,:]
     end
     t,y,z,y
 end
 
-function solve_time(dmp, t, y0, g, solver)
+function solve_time(dmp, t, y0, g, solver; kwargs...)
     T,n     = size(dmp.y)
     αx      = dmp.opts.αx
     τ       = dmp.τ
@@ -272,7 +272,7 @@ function solve_time(dmp, t, y0, g, solver)
         end
         state0  = [0; y0[i]]
         prob = OrdinaryDiffEq.ODEProblem(time_derivative,state0,(t[1],t[end]))
-        sol = OrdinaryDiffEq.solve(prob,solver,saveat=t,dt=t[2])
+        sol = OrdinaryDiffEq.solve(prob,solver;saveat=t,dt=t[2], kwargs...)
         z[:,i] = sol[1,:]
         y[:,i] = sol[2,:]
     end
@@ -292,13 +292,13 @@ The default solver is `solver=ode54`, a faster alternative is `solver=euler`
 
 see also `plotdmp`
 """
-function solve(dmp::AbstractDMP, t = dmp.t; y0 = _1(dmp), g = dmp.g, solver=OrdinaryDiffEq.Tsit5())
+function solve(dmp::AbstractDMP, t = dmp.t; y0 = _1(dmp), g = dmp.g, solver=OrdinaryDiffEq.Tsit5(), kwargs...)
     if dmp.opts.sched_sig == :position
-        return solve_position(dmp, t, y0, g, solver)
+        return solve_position(dmp, t, y0, g, solver; kwargs...)
     elseif dmp.opts.sched_sig == :time
-        return solve_time(dmp, t, y0, g, solver)
+        return solve_time(dmp, t, y0, g, solver; kwargs...)
     end
-    solve_canonical(dmp, t, y0, g, solver)
+    solve_canonical(dmp, t, y0, g, solver; kwargs...)
 end
 
 

src/demo_two_link.jl

@@ -20,7 +20,7 @@ Nbasis = 20
 opts = DMPopts(Nbasis,αx,αz, :canonical)
 
 T = size(y,1)
-t = linspace(0,T,T)
+t = LinRange(0,T,T)
 τ = t[end]/3
 h = t[2]-t[1]
 qd = centraldiff(q) / h

src/two_link.jl

@@ -1,4 +1,5 @@
 module TwoLink
+using RecipesBase
 
 export torque, forward_kin, inverse_kin, inverse_kin_up, inverse_kin_down, traj, connect_points, acceleration, time_derivative, time_derivative!, inertia, ⊕
 ⊕(a::Tuple{Number,Number}, b::Tuple{Number,Number}) = (a[1]+b[1], a[2]+b[2])
@@ -122,7 +123,7 @@ function forward_kin(q)
     p1, p
 end
 
-function forward_kin(jtraj::Matrix)
+function forward_kin(jtraj::AbstractMatrix)
     N = size(jtraj,2)
     p = zeros(eltype(jtraj), 2, N)
     for i = 1:N
@@ -166,7 +167,7 @@ end
 
 `p` is of size N × 2
 """
-function inverse_kin(ctraj::Matrix, dir = :up)
+function inverse_kin(ctraj::AbstractMatrix, dir = :up)
     if dir == :up
         kin_fun = inverse_kin_up
     elseif dir == :down
@@ -190,7 +191,7 @@ Move from `q0` to `q1` during total time `t` using quadratic blends.
 """
 function traj(q0,q1,t)
     tf = maximum(t)
-    V = (q1-q0)/tf * 1.5
+    V = (q1-q0) ./tf * 1.5
     traj(q0,q1,t, V)
 end
 
@@ -218,9 +219,9 @@ function traj(q0,q1,t, V)
     tb = (q0 - q1 + V*tf)/V
     a = V/tb
 
-    p = Array{eltype(q0)}(size(t))
-    pd = Array{eltype(q0)}(size(t))
-    pdd = Array{eltype(q0)}(size(t))
+    p = Array{eltype(q0)}(undef, size(t))
+    pd = Array{eltype(q0)}(undef, size(t))
+    pdd = Array{eltype(q0)}(undef, size(t))
 
     for (i,t) = enumerate(t)
         if t <= tb
@@ -276,12 +277,12 @@ function connect_points(points,ni)
     p', pd', pdd'
 end
 
-
-function plotworkspace(N=1000)
+@userplot Plotworkspace
+@recipe function plotworkspace(p::Plotworkspace, N=1000)
     q = [(2π*rand(),2π*rand()) for n in 1:N]
     p = getindex.(forward_kin.(q), 2)
     p1,p2 = getindex.(p,1), getindex.(p,2)
-    scatter(p1,p2)
+    seriesstyle := :scatter
     p1,p2
 end