NLP seems to be doing pretty well

julia/test/inner_loop/nlp_solver.jl

@@ -1,36 +1,48 @@
 @testset "Phase" begin
 
+  using PlotlyJS: savefig
+
   println("Testing NLP solver")
 
-  # We'll start by testing the mission_guess -> vector function
-  vec = Vector(test_mg)
-  @test typeof(vec) == Vector{Float64}
-
-  # Now we go in the other direction
-  flybys = [ p.planet for p in test_mg.phases ]
-  guess = Mission_Guess(vec, test_mg.sc, test_mg.start_mass, flybys)
-  @test typeof(guess) == Mission_Guess
-  @test guess.sc == test_mg.sc
-  @test guess.start_mass == test_mg.start_mass
-  @test guess.launch_date == test_mg.launch_date
-  @test guess.launch_v∞ == test_mg.launch_v∞
-  @test guess.converged == test_mg.converged
-  for i in 1:length(guess.phases)
-    @test guess.phases[i].planet == test_mg.phases[i].planet
-    @test guess.phases[i].v∞_in == test_mg.phases[i].v∞_in
-    @test guess.phases[i].v∞_out == test_mg.phases[i].v∞_out
-    @test guess.phases[i].tof == test_mg.phases[i].tof
-    @test guess.phases[i].thrust_profile == test_mg.phases[i].thrust_profile
-  end
-
-  # Now we test an example run of the basic "inner function"
+  # Test the optimizer for a one-phase mission
+  # The lambert's solver said this should be pretty valid
+  launch_window = [DateTime(1992,11,1), DateTime(1992,12,1)]
+  latest_arrival = DateTime(1993,6,1)
   leave, arrive = DateTime(1992,11,19), DateTime(1993,4,1)
+  test_leave = DateTime(1992,11,12)
   earth_state = state(Earth, leave)
   venus_state = state(Venus, arrive)
   v∞_out, v∞_in, tof = Thesis.lamberts(Earth, Venus, leave, arrive)
-  phase = Phase(Venus, v∞_in, v∞_in, tof, zeros(20,3))
-  guess = Mission_Guess(bepi, 12_000., leave, v∞_out, [phase])
-  g = solve_mission(guess)
-  println(g)
+  # We can get the thrust profile and tof pretty wrong and still be ok
+  phase = Phase(Venus, 1.1v∞_in, v∞_in, 0.9*tof, 0.1*ones(20,3))
+  guess = Mission_Guess(bepi, 12_000., test_leave, 0.9*v∞_out, [phase])
+  m = solve_mission(guess, launch_window, latest_arrival, verbose=true)
+  @test typeof(m) == Mission
+  @test m.converged == true
+
+  # Now we can plot the results to check visually
+  p = plot(m, title="NLP Test Solution")
+  savefig(p,"../plots/nlp_test_1_phase.html")
+
+  # Now we can look at a more complicated trajectory
+  flybys = [Earth, Venus, Mars]
+  launch_window = [DateTime(2021,10,1), DateTime(2021,12,1)]
+  latest_arrival = DateTime(2023,1,1)
+  dates = [DateTime(2021,11,1), DateTime(2022,3,27), DateTime(2022,8,28)]
+  phases = Vector{Phase}()
+  launch_v∞, _, tof1 = Thesis.lamberts(flybys[1], flybys[2], dates[1], dates[2])
+  println(launch_v∞)
+  for i in 1:length(dates)-2
+    v∞_out1, v∞_in1, tof1 = Thesis.lamberts(flybys[i], flybys[i+1], dates[i], dates[i+1])
+    v∞_out2, v∞_in2, tof2 = Thesis.lamberts(flybys[i+1], flybys[i+2], dates[i+1], dates[i+2])
+    push!(phases, Phase(flybys[i+1], v∞_in1, v∞_out2, tof1, 0.01*ones(20,3)))
+  end
+  v∞_out, v∞_in, tof = Thesis.lamberts(flybys[end-1], flybys[end], dates[end-1], dates[end])
+  push!(phases, Phase(flybys[end], v∞_in, v∞_in, tof, 0.01*ones(20,3)))
+  # This isn't quite right. V∞ discrepancy is too big
+  guess = Mission_Guess(bepi, 12_000., dates[1], launch_v∞, phases)
+  m = solve_mission(guess, launch_window, latest_arrival, verbose=true)
+  p = plot(m, title="NLP Test Solution (2 Phases)")
+  savefig(p,"../plots/nlp_test_2_phase.html")
 
 end