def __init__(self, *inputs, label=""):
self.inputs = inputs
self.plot = Plot()
self.max_num_steps = 0
'Assuming that all models use the same dynamics and same initial set for now'
self.model = inputs[0]['model']
self.label = label
def test_lin_HarOsc():
NUM_STEPS = 5
model = HarOsc()
#trajs = generate_traj(model, 10, 200)
mod_reach = ReachSet(model)
#mod_flow = mod_reach.computeReachSet()
sir_plot = Plot()
#mod_flow = mod_reach.computeReachSet(NUM_STEPS)
SIR_LIN_ITER_STEPS = 1 #Number of steps between each recomputation of PCA Templates.
lin_strat = LinStrat(model, iter_steps=SIR_LIN_ITER_STEPS)
mod_lin_flow = mod_reach.computeReachSet(NUM_STEPS, tempstrat=lin_strat, transmode=BundleMode.AFO)
trajs = [Traj(model, point, steps=NUM_STEPS) for point in product([-5,-4],[0,1])]
'Generaste the trajectories and add them to the plot.'
sir_plot.add(mod_lin_flow, "HarOsc LINAPP")
for t in trajs:
sir_plot.add(t)
sir_plot.plot2DPhase(0,1, separate=True, plotvertices=True)
Timer.generate_stats()
def test_pca_lin_Phos():
NUM_STEPS = 30
PHOS_LIN_ITER_STEPS = 1 #Number of steps between each recomputation of LinApp Templates.
PHOS_PCA_ITER_STEPS = 1 #Number of steps between each recomputation of PCA Templates.
'PCA Strategy Parameters'
PHOS_PCA_TRAJ_STEPS = 5 #Number of steps our sample trajectories should run.
PHOS_PCA_NUM_TRAJ = 200 #Number of sample trajectories we should use for the PCA routine.
PHOS_LIFE_SPAN = 3
unit_model = Phosphorelay_UnitBox()
unit_mod_reach = ReachSet(unit_model)
#points = [[1,1,1], [1.005, 1,1], [1.01,1.01,1.01], [1.005,1.01,1.01], [1,1.005,1], [1.01,1,1.05]]
#trajs = [Traj(unit_model , point, NUM_STEPS) for point in points]
multi_strat = MultiStrategy(LinStrat(unit_model, iter_steps=PHOS_LIN_ITER_STEPS), \
DelayedPCAStrat(unit_model, traj_steps=PHOS_PCA_TRAJ_STEPS, num_trajs=PHOS_PCA_NUM_TRAJ, life_span=PHOS_LIFE_SPAN))
#PCAStrat(unit_model, traj_steps=PHOS_PCA_TRAJ_STEPS, num_trajs=PHOS_PCA_NUM_TRAJ, iter_steps=PHOS_PCA_ITER_STEPS+PHOS_PCA_DELAY))
#PCAStrat(unit_model, traj_steps=PHOS_PCA_TRAJ_STEPS, num_trajs=PHOS_PCA_NUM_TRAJ, iter_steps=PHOS_PCA_ITER_STEPS+2*PHOS_PCA_DELAY))
mod_lin_flow = unit_mod_reach.computeReachSet(NUM_STEPS,
tempstrat=multi_strat)
# points = [[0,1.97], [0.01, 1.97], [0.01,2], [0,2], [0.005,1.97], [0.005,2], [0,1.97], [0,1.985], [0.01,1.985]]
#trajs = [Traj(unit_model, point, NUM_STEPS) for point in points]
phos_plot = Plot()
phos_plot.add(mod_lin_flow)
'Add trajectories'
for traj in trajs:
phos_plot.add(traj)
phos_plot.plot2DPhase(0, 1, separate=False, plotvertices=True)
def test_sir_pca_strat():
#Compute Sapo's version.3
sir_pca = SIR_UnitBox(delta=0.5)
sir = SIR(delta=0.5)
sir_reach = ReachSet(sir)
sir_flow = sir_reach.computeReachSet(NUM_STEPS)
sir_plot = Plot()
sir_plot.add(sir_flow, "SIR SAPO")
for i in range(ITER_SPREAD, ITER_SPREAD + 1):
print(
colored("Generating PCA with Iterative Step Size: {}".format(i),
"white",
attrs=['reverse', 'blink']))
sir_pca_reach = ReachSet(sir_pca)
sir_flow_pca = sir_pca_reach.computeReachSet(
NUM_STEPS,
tempstrat=PCAStrat(sir_pca, iter_steps=i),
transmode=BundleMode.AFO)
sir_plot.add(sir_flow_pca, "SIR_PCA_{}".format(i))
sir_plot.plot2DPhase(0, 1, separate=False, plotvertices=True)
#sir_plot.plot(0,1,2)
Timer.generate_stats()
"""
def test_sir_lin_pca_strat():
NUM_STEPS = 70
SIR_PCA_ITER_STEPS = 1 #Number of steps between each recomputation of PCA Templates.
'PCA Strategy Parameters'
SIR_PCA_TRAJ_STEPS = 1 #Number of steps our sample trajectories should run.
SIR_PCA_NUM_TRAJ = 100 #Number of sample trajectories we should use for the PCA routine.
SIR_LIN_ITER_STEPS = 1
#
SIR_PCA_LIFE_SPAN = 3
sir_pca = SIR_UnitBox(delta=0.5)
sir_plot = Plot()
points = [[0.79,0.19,0], [0.79, 0.2,0], [0.8,0.19,0], [0.8,0.2,0], [0.79,0.195,0], [0.8,0.195,0], [0.795,0.19,0], [0.795,0.2,0]]
trajs = [Traj(sir_pca, point, NUM_STEPS) for point in points]
pca_strat = MultiStrategy(LinStrat(sir_pca, iter_steps=SIR_LIN_ITER_STEPS), \
DelayedPCAStrat(sir_pca, traj_steps=SIR_PCA_TRAJ_STEPS, num_trajs=SIR_PCA_NUM_TRAJ, life_span=SIR_PCA_LIFE_SPAN))
sir_pca_reach = ReachSet(sir_pca)
sir_flow_pca = sir_pca_reach.computeReachSet(NUM_STEPS, tempstrat=pca_strat)
sir_plot.add(sir_flow_pca, "SIR_LinApp&PCA")
'Add trajectories'
for traj in trajs:
sir_plot.add(traj)
# sir_plot.plot2DPhase(0,1,separate=False, plotvertices=True)
sir_plot.plot2DPhase(1,2,separate=False, plotvertices=True)
sir_plot.plot2DPhase(0,2,separate=False, plotvertices=True)
Timer.generate_stats()
def test_pca_lin_Rossler():
NUM_STEPS = 5
ROSS_PCA_ITER_STEPS = 1 #Number of steps between each recomputation of PCA Templates.
'PCA Strategy Parameters'
ROSS_PCA_TRAJ_STEPS = 1 #Number of steps our sample trajectories should run.
ROSS_PCA_NUM_TRAJ = 200 #Number of sample trajectories we should use for the PCA routine.
rossler_pca = Rossler_UnitBox(delta=0.5)
rossler_plot = Plot()
points = [[0.05,4.95,0.05], [0.1,4.95,0.05], [0.05,5,0.05], [0.1,5,0.05], [0.05,4.95,0.05], [0.05,4.95,0.1], [0.1,4.95,0.1], [0.1,5,0.1]]
trajs = [Traj(rossler_pca, point, NUM_STEPS) for point in points]
pca_strat = PCALinStrat(rossler_pca, traj_steps=ROSS_PCA_TRAJ_STEPS, num_trajs=ROSS_PCA_NUM_TRAJ, iter_steps=ROSS_PCA_ITER_STEPS)
ross_pca_reach = ReachSet(rossler_pca)
ross_flow_pca = ross_pca_reach.computeReachSet(NUM_STEPS, tempstrat=pca_strat)
rossler_plot.add(ross_flow_pca, "SIR_LinApp&PCA")
'Add trajectories'
for traj in trajs:
rossler_plot.add(traj)
rossler_plot.plot2DPhase(0,1,separate=True, plotvertices=True)
Timer.generate_stats()
def test_sir_cut_strat():
#Compute Sapo's version.
sir_cut = SIR()
sir = SIR_UnitBox()
sir_reach = ReachSet(sir)
sir_flow = sir_reach.computeReachSet(NUM_STEPS)
sir_plot = Plot()
sir_plot.add(sir_flow)
sir_cut_reach = ReachSet(sir_cut)
sir_cut_flow = sir_cut_reach.computeReachSet(NUM_STEPS, CutStrat(sir_cut))
sir_sapo_flow = sir_cut_reach.computeReachSet(NUM_STEPS)
sir_plot.add(sir_cut_flow, "SIR_CUT_150")
sir_plot.add(sir_sapo_flow, "SIR_SAPO")
sir_plot.plot(0, 1, 2)
Timer.generate_stats()
def test_pca_Quad():
NUM_STEPS = 5
QUAD_PCA_ITER_STEPS = 1 #Number of steps between each recomputation of PCA Templates.
'PCA Strategy Parameters'
QUAD_PCA_TRAJ_STEPS = 1 #Number of steps our sample trajectories should run.
QUAD_PCA_NUM_TRAJ = 200 #Number of sample trajectories we should use for the PCA routine.
quad_pca = Quadcopter_UnitBox()
quad_plot = Plot()
#trajs = [Traj(rossler_pca, point, NUM_STEPS) for point in points]
pca_strat = PCAStrat(quad_pca,
traj_steps=QUAD_PCA_TRAJ_STEPS,
num_trajs=QUAD_PCA_NUM_TRAJ,
iter_steps=QUAD_PCA_ITER_STEPS)
quad_pca_reach = ReachSet(quad_pca)
quad_flow_pca = quad_pca_reach.computeReachSet(NUM_STEPS,
tempstrat=pca_strat)
'Add trajectories'
#for traj in trajs:
# rossler_plot.add(traj)
quad_plot.add(quad_flow_pca, "Quad_PCA")
quad_plot.plot2DPhase(2, 5, separate=True, plotvertices=True)
def test_sir_lin_strat():
#Compute Sapo's version.
sir_lin = SIR_UnitBox(delta=0.5)
sir = SIR()
#sir_reach = ReachSet(sir)
#sir_flow = sir_reach.computeReachSet(NUM_STEPS)
sir_plot = Plot()
#sir_plot.add(sir_flow)
for i in range(10, 11):
print(
colored(
"Generating Lin_Approx with Iterative Step Size: {}".format(i),
"white",
attrs=['reverse', 'blink']))
sir_lin_reach = ReachSet(sir_lin)
sir_flow_lin = sir_lin_reach.computeReachSet(
NUM_STEPS, LinStrat(sir_lin, iter_steps=i))
sir_plot.add(sir_flow_lin, "SIR_LIN_{}".format(i))
#sir_plot.plot(0,1,2)
sir_plot.plot2DPhase(0, 1, separate=True)
Timer.generate_stats()
"""
def test_pca_HarOsc():
NUM_STEPS = 4
model = HarOsc()
#trajs = generate_traj(model, 10, 200)
mod_reach = ReachSet(model)
#mod_flow = mod_reach.computeReachSet()
sir_plot = Plot()
SIR_PCA_ITER_STEPS = 1 #Number of steps between each recomputation of PCA Templates.
'PCA Strategy Parameters'
SIR_PCA_TRAJ_STEPS = 2 #Number of steps our sample trajectories should run.
SIR_PCA_NUM_TRAJ = 100 #Number of sample trajectories we should use for the PCA routine.
pca_strat = PCAStrat(model, traj_steps=SIR_PCA_TRAJ_STEPS, num_trajs=SIR_PCA_NUM_TRAJ, iter_steps=SIR_PCA_ITER_STEPS)
mod_pca_flow = mod_reach.computeReachSet(NUM_STEPS, tempstrat=[pca_strat], transmode=BundleMode.AFO)
#trajs = generate_traj(model, 10, 200)
'Generaste the trajectories and add them to the plot.'
sir_plot.add(mod_pca_flow, "HarOsc PCA")
sir_plot.plot2DPhase(0,1, separate=True, plotvertices=True)
Timer.generate_stats()
def test_all_pca():
num_steps = 100
model_var_dict = {
Basic: [0],
Rossler_PCA: [0, 1, 2],
SIR_PCA: [0, 1, 2],
LotkaVolterra_PCA: [0, 1, 2],
Quadcopter_PCA: [2, 5, 13],
Phosphorelay_PCA: [0, 1]
}
model_path_dict = {
Basic: "/Users/edwardkim/Work/kaa-optimize/figures/Basic",
Rossler_PCA: "/Users/edwardkim/Work/kaa-optimize/figures/Rossler",
SIR_PCA: "/Users/edwardkim/Work/kaa-optimize/figures/SIR",
LotkaVolterra_PCA:
"/Users/edwardkim/Work/kaa-optimize/figures/LotVolt",
Quadcopter_PCA: "/Users/edwardkim/Work/kaa-optimize/figures/Quad",
Phosphorelay_PCA: "/Users/edwardkim/Work/kaa-optimize/figures/Phos"
}
for model, var_ind_list in model_var_dict.items():
print("\n Generating Model: {}\n".format(model))
flowpipe = ReachSet(model()).computeReachSet(num_steps, PCAStrat)
plot = Plot()
plot.add(flowpipe)
plot.plot(*var_ind_list, path=model_path_dict[model])
def test_phase_sir():
sir_mod = SIR_UnitBox(delta=0.5)
sir_reach = ReachSet(sir_mod)
flowpipe = sir_reach.computeReachSet(50)
plot = Plot()
plot.add(flowpipe)
plot.plot2DPhase(1, 2)
def test_basic():
basic_mod = Basic()
basic_reach = ReachSet(basic_mod)
flowpipe = basic_reach.computeReachSet(300)
basic_plot = Plot()
basic_plot.add_flowpipe(flowpipe)
basic_plot.plot(0)
def pca_lin_comp():
sir = SIR_UnitBox()
sir_reach = ReachSet(sir)
sir_flow = sir_reach.computeReachSet(NUM_STEPS)
sir_plot = Plot()
sir_plot.add(sir_flow)
sir_lin_flow = sir_reach.computeReachSet(NUM_STEPS,
LinStrat(sir, iter_steps=150))
#sir_pca_flow = sir_reach.computeReachSet(NUM_STEPS, PCAStrat(sir, iter_steps=50))
sir_plot.add(sir_lin_flow, "SIR_LIN")
#sir_plot.add(sir_pca_flow, "SIR_PCA")
sir_plot.plot(0, 1, 2)
Timer.generate_stats()
def test_LL():
model = LL()
mod_reach = ReachSet(model)
mod_flow = mod_reach.computeReachSet(150)
ll_plot = Plot()
ll_plot.add(mod_flow)
ll_plot.plot(0)
Timer.generate_stats()
def test_rossler_phase():
model = Rossler()
mod_reach = ReachSet(model)
mod_flow = mod_reach.computeReachSet(200)
rossler_plot = Plot()
rossler_plot.add(mod_flow)
rossler_plot.plot2DPhase(0,1)
Timer.generate_stats()
def test_Quad():
model = Quadcopter()
mod_reach = ReachSet(model)
mod_flow = mod_reach.computeReachSet(3)
quad_plot = Plot()
quad_plot.add(mod_flow)
quad_plot.plot(2, 5, 13)
Timer.generate_stats()
def test_Rossler():
model = Rossler()
mod_reach = ReachSet(model)
mod_flow = mod_reach.computeReachSet(300)
rossler_plot = Plot()
rossler_plot.add(mod_flow)
rossler_plot.plot(0,1,2)
Timer.generate_stats()
def test_basic2():
basic_mod = Basic2()
basic_reach = ReachSet(basic_mod)
flowpipe = basic_reach.computeReachSet(300)
basic_plot = Plot()
basic_plot.add(flowpipe)
basic_plot.plot(0)
Timer.generate_stats()
def test_LV():
model = LotkaVolterra()
mod_reach = ReachSet(model)
mod_flow = mod_reach.computeReachSet(100)
plot = Plot()
plot.add(mod_flow)
plot.plot(0,1,2)
Timer.generate_stats()
def test_Phos():
model = Phosphorelay()
#unit_model = Phosphorelay_UnitBox()
mod_reach = ReachSet(model)
#mod_unit_reach = ReachSet(unit_model)
#unit_flow = mod_unit_reach.computeReachSet(200)
mod_flow = mod_reach.computeReachSet(30)
phos_plot = Plot()
phos_plot.add(mod_flow)
phos_plot.plot2DPhase(0, 1, separate=False, plotvertices=True)
Timer.generate_stats()
def test_OscPart():
model = OscPart()
#trajs = generate_traj(model, 10, 200)
mod_reach = ReachSet(model)
mod_flow = mod_reach.computeReachSet(20)
sir_plot = Plot()
#trajs = generate_traj(model, 10, 200)
'Generaste the trajectories and add them to the plot.'
sir_plot.add(mod_flow)
sir_plot.plot(0, 1, 2)
Timer.generate_stats()
def test_rossler_lin_strat():
#Compute Sapo's version.
rossler_lin = Rossler_UnitBox()
rossler = Rossler()
rossler_reach = ReachSet(rossler)
rossler_flow = rossler_reach.computeReachSet(NUM_STEPS)
rossler_plot = Plot()
rossler_plot.add(rossler_flow)
for i in range(10, 11):
print(
colored("Generating LIN with Iterative Step Size: {}".format(i),
"white",
attrs=['reverse', 'blink']))
rossler_lin_reach = ReachSet(rossler_lin)
rossler_flow_lin = rossler_lin_reach.computeReachSet(
NUM_STEPS, LinStrat(rossler_lin, iter_steps=i))
rossler_plot.add(rossler_flow_lin, "Rossler_LIN_{}".format(i))
rossler_plot.plot(0, 1, 2)
Timer.generate_stats()
def test_lv_pca_strat():
#Compute Sapo's version.
lv_pca = LotkaVolterra_UnitBox()
lv = LotkaVolterra()
lv_reach = ReachSet(lv)
lv_flow = lv_reach.computeReachSet(NUM_STEPS)
lv_plot = Plot()
lv_plot.add(lv_flow)
for i in range(1, ITER_SPREAD):
print(
colored("Generating PCA with Iterative Step Size: {}".format(2 *
i),
"white",
attrs=['reverse', 'blink']))
lv_pca_reach = ReachSet(lv_pca)
lv_flow_pca = lv_pca_reach.computeReachSet(
NUM_STEPS, PCAStrat(lv_pca, iter_steps=2 * i))
lv_plot.add(lv_flow_pca, "LV_PCA_{}".format(2 * i))
lv_plot.plot(0, 1, 2)
Timer.generate_stats()
def test_quad_pca_strat():
#Compute Sapo's version.
quad_pca = Quadcopter_UnitBox()
quad = Quadcopter()
quad_reach = ReachSet(quad)
quad_flow = quad_reach.computeReachSet(NUM_STEPS)
quad_plot = Plot()
quad_plot.add(quad_flow)
for i in range(1, ITER_SPREAD):
print(
colored("Generating PCA with Iterative Step Size: {}".format(2 *
i),
"white",
attrs=['reverse', 'blink']))
quad_pca_reach = ReachSet(quad_pca)
quad_flow_pca = quad_pca_reach.computeReachSet(
NUM_STEPS, PCAStrat(quad_pca, iter_steps=2 * i))
quad_plot.add(quad_flow_pca, "QUAD_PCA_{}".format(2 * i))
quad_plot.plot(2, 5, 13)
Timer.generate_stats()
def test_SIR():
model = SIR(delta=0.5)
model_unit = SIR_UnitBox()
#trajs = generate_traj(model, 10, 200)
mod_reach = ReachSet(model)
mod_unit_reach = ReachSet(model_unit)
mod_flow = mod_reach.computeReachSet(70)
#mod_unit_flow = mod_unit_reach.computeReachSet(300)
sir_plot = Plot()
#trajs = generate_traj(model, 10, 200)
'Generaste the trajectories and add them to the plot.'
#for traj in trajs:
# sir_plot.add(traj)
sir_plot.add(mod_flow)
#sir_plot.add(mod_unit_flow)
sir_plot.plot2DPhase(0,1)
Timer.generate_stats()
def test_rossler_pca_strat():
#Compute Sapo's version.
rossler_pca = Rossler_UnitBox()
rossler = Rossler()
rossler_reach = ReachSet(rossler)
#rossler_flow = rossler_reach.computeReachSet(NUM_STEPS)
rossler_plot = Plot()
#rossler_plot.add(rossler_flow)
for i in range(3, 4):
print(
colored("Generating PCA with Iterative Step Size: {}".format(i),
"white",
attrs=['reverse', 'blink']))
rossler_pca_reach = ReachSet(rossler_pca)
rossler_flow_pca = rossler_pca_reach.computeReachSet(
NUM_STEPS, PCAStrat(rossler_pca, iter_steps=i))
rossler_plot.add(rossler_flow_pca, "Rossler_PCA_{}".format(i))
rossler_plot.plot(0, 1, 2)
Timer.generate_stats()
def test_haroscrotate():
NUM_STEPS = 4
model = HarOscRotate()
mod_reach = ReachSet(model)
mod_flow = mod_reach.computeReachSet(NUM_STEPS)
SIR_PCA_ITER_STEPS = 1 #Number of steps between each recomputation of PCA Templates.
'PCA Strategy Parameters'
SIR_PCA_TRAJ_STEPS = 1 #Number of steps our sample trajectories should run.
SIR_PCA_NUM_TRAJ = 100 #Number of sample trajectories we should use for the PCA routine.
#pca_strat = PCAStrat(model, traj_steps=SIR_PCA_TRAJ_STEPS, num_trajs=SIR_PCA_NUM_TRAJ, iter_steps=SIR_PCA_ITER_STEPS)
#mod_pca_flow = mod_reach.computeReachSet(NUM_STEPS, tempstrat=pca_strat)
vdp_plot = Plot()
vdp_plot.add(mod_flow, "HarOsc")
#vdp_plot.add(mod_pca_flow, "HarOsc PCA")
vdp_plot.plot2DPhase(0, 1)
Timer.generate_stats()
class Experiment:
def __init__(self, *inputs, label=""):
self.inputs = inputs
self.plot = Plot()
self.max_num_steps = 0
'Assuming that all models use the same dynamics and same initial set for now'
self.model = inputs[0]['model']
self.label = label
"""
Execute the reachable set simulations and add the flowpipes to the Plot.
"""
def execute(self):
self.output_flowpipes = []
for experi_input in self.inputs:
model = experi_input['model']
strat = experi_input['strat']
label = experi_input['label']
num_steps = experi_input['num_steps']
mod_reach = ReachSet(model)
mod_flow = mod_reach.computeReachSet(num_steps, tempstrat=strat)
self.plot.add(mod_flow, label=label)
self.output_flowpipes.append(mod_flow)
self.max_num_steps = max(self.max_num_steps, num_steps)
"""
Plot the results fed into the Plot object
"""
def plot_results(self, *var_tup, plottrajs=True):
border_sim_trajs = self.__simulate_border_points(self.max_num_steps)
if plottrajs:
self.plot.add(border_sim_trajs)
self.plot.plot(*var_tup)
def get_total_vol_results(self):
assert self.output_flowpipes is not None, "Execute Experiment with ExperimentInputs before retrieving volume data."
return [flowpipe.total_volume for flowpipe in self.output_flowpipes]
"""
Extract the initial box intervals from the model
"""
def __get_init_box(self):
init_offu = self.model.bund.offu[:self.model.
dim] #Assume first dim # of offsets are associated to initial box
init_offl = self.model.bund.offl[:self.model.dim]
return [[-lower_off, upper_off]
for lower_off, upper_off in zip(init_offl, init_offu)]
"""
Sample points from the edges of the box and propagate them for a number of steps.
"""
def __simulate_border_points(self, num_steps):
init_box_inter = self.__get_init_box()
border_points = get_init_box_borders(init_box_inter)
trajs = [Traj(self.model, point, num_steps) for point in border_points]
return TrajCollection(trajs)