You can not select more than 25 topics
Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
71 lines
3.3 KiB
71 lines
3.3 KiB
import stormpy
|
|
import stormpy.simulator
|
|
|
|
|
|
def mdp():
|
|
options = stormpy.BuilderOptions([])
|
|
options.set_build_state_valuations(True)
|
|
options.set_build_choice_labels(True)
|
|
options.set_build_all_labels()
|
|
options.set_build_all_reward_models()
|
|
|
|
program = stormpy.parse_prism_program("/media/task_graph_scheduling.prism")
|
|
formula_str = "R{\"time\"}min=? [ F \"tasks_complete\" ];"
|
|
formulas = stormpy.parse_properties_for_prism_program(formula_str, program)
|
|
model = stormpy.build_sparse_model_with_options(program, options)
|
|
print(model)
|
|
|
|
initial_state = model.initial_states[0]
|
|
assert initial_state == 0
|
|
result = stormpy.model_checking(model, formulas[0], extract_scheduler=True)
|
|
assert result.has_scheduler
|
|
scheduler = result.scheduler
|
|
assert scheduler.memoryless
|
|
assert scheduler.deterministic
|
|
|
|
dtmc = model.apply_scheduler(scheduler)
|
|
print(dtmc)
|
|
simulator = stormpy.simulator.create_simulator(dtmc, seed=42)
|
|
simulator.set_observation_mode(stormpy.simulator.SimulatorObservationMode.PROGRAM_LEVEL)
|
|
|
|
step_counter = 0
|
|
busy1_counter = 0
|
|
busy2_counter = 0
|
|
print("""
|
|
┌t1─┐ ┌t3─────┐ ┌t5───────┐
|
|
│A+B│-->│Cx(A+B)│-->│DxCx(A+B)│--. ┌t6─────────────────────┐
|
|
└───┘\ └───────┘ └─────────┘ \___\│DxCx(A+B)+((A+B)+(CxD))│
|
|
┌t2─┐ \ ┌t4─────────┐ / /└───────────────────────┘
|
|
│CxD│-->│(A+B)+(CxD)│------------°
|
|
└───┘ └───────────┘
|
|
""")
|
|
print(" " * 32 + "[+,x,x,+,x,+]")
|
|
while True:
|
|
if simulator.is_done():
|
|
print(f"Tasks complete! Ratio: (approx.): {busy1_counter/(busy1_counter+busy2_counter):1.4f}/{busy2_counter/(busy1_counter+busy2_counter):1.4f}, Steps needed (approx.): {step_counter}")
|
|
break
|
|
observation, reward, labels = simulator.step()
|
|
#input("")
|
|
busy1 = str(observation["p1_idle"]) != ("true")
|
|
busy2 = str(observation["p2_idle"]) != ("true")
|
|
if busy1: busy1_counter += 1
|
|
if busy2: busy2_counter += 1
|
|
tasks = [str(observation[f"task{i}"]) for i in range(1,7)]
|
|
print(f"Step {step_counter:3}: P1/2:{'busy' if busy1 else 'idle'}/{'busy' if busy2 else 'idle'} tasks: [{','.join(tasks)}], {str(observation['add_tick_1'])},{str(observation['mult_tick_1'])},{str(observation['add_tick_2'])},{str(observation['mult_tick_2'])},labels: {labels if len(labels) != 0 else ''}, ")
|
|
step_counter += 1
|
|
try:
|
|
with open("/media/induced_scheduling_dtmc.dot", "w") as text_file:
|
|
text_file.write(dtmc.to_dot())
|
|
except Exception as e:
|
|
print(e)
|
|
print("Could not write to file.")
|
|
input("Hit enter to evaluate: 'P=? [ F<={i} \"tasks_complete\" ]' on the DTMC:")
|
|
print(dtmc)
|
|
|
|
for i in range(30,61):
|
|
formula_str = f"P=? [ F<={i} \"tasks_complete\" ]"
|
|
formulas = stormpy.parse_properties_for_prism_program(formula_str, program)
|
|
result = stormpy.model_checking(dtmc, formulas[0])
|
|
print(f"Probability to finish within {i} timesteps : {result.at(dtmc.initial_states[0]):>1.12f}")
|
|
if __name__ == '__main__':
|
|
mdp()
|