def validate_MDP_policy(root_path, flag_with_obs=True, flag_plan=True):
s_g = np.array([4.0, 3.0, 0.0])
modality = "haptic"
obs_list = [(1.5, 2.00, 1.0, 1.25),
(2.0, 1.0, 1.25, 0.5)]
if flag_with_obs:
file_name = root_path + "/mdp_planenr_obs_" + modality + ".pkl"
else:
file_name = root_path + "/mdp_planner_" + modality + ".pkl"
if flag_plan:
human_model = MovementModel()
human_model.load_model(root_path)
human_model.set_default_param()
planner = MDPFixedTimePolicy(tmodel=human_model)
if flag_with_obs:
planner.gen_env(obs_list)
planner.compute_policy(s_g, modality, max_iter=30)
else:
with open(file_name) as f:
planner = pickle.load(f)
fig, axes = plt.subplots()
planner.visualize_policy(axes)
plt.show()
if flag_plan:
with open(file_name, "w") as f:
pickle.dump(planner, f)
sim = Simulator(planner)
n_trials = 30
traj_list = []
start_time = time.time()
for i in range(n_trials):
traj_list.append(sim.run_trial((0.5, 0.5, 0.0), s_g, modality, 30.0, tol=0.5))
print "--- %s seconds ---" % (time.time() - start_time)
fig, axes = plt.subplots()
for i in range(n_trials):
t, traj = traj_list[i]
axes.plot(traj[:, 0], traj[:, 1])
axes.axis("equal")
axes.scatter(s_g[0], s_g[1])
if flag_with_obs:
for x, y, w, h in obs_list:
rect = Rectangle((x, y), w, h)
axes.add_patch(rect)
plt.show()
def validate_MDP_policy(root_path, flag_with_obs=True, flag_plan=True):
s_g = np.array([4.0, 3.0, 0.0])
modality = "haptic"
obs_list = [(1.5, 2.00, 1.0, 1.25), (2.0, 1.0, 1.25, 0.5)]
if flag_with_obs:
file_name = root_path + "/mdp_planenr_obs_" + modality + ".pkl"
else:
file_name = root_path + "/mdp_planner_" + modality + ".pkl"
if flag_plan:
human_model = MovementModel()
human_model.load_model(root_path)
human_model.set_default_param()
planner = MDPFixedTimePolicy(tmodel=human_model)
if flag_with_obs:
planner.gen_env(obs_list)
planner.compute_policy(s_g, modality, max_iter=30)
else:
with open(file_name) as f:
planner = pickle.load(f)
fig, axes = plt.subplots()
planner.visualize_policy(axes)
plt.show()
if flag_plan:
with open(file_name, "w") as f:
pickle.dump(planner, f)
sim = Simulator(planner)
n_trials = 30
traj_list = []
start_time = time.time()
for i in range(n_trials):
traj_list.append(
sim.run_trial((0.5, 0.5, 0.0), s_g, modality, 30.0, tol=0.5))
print "--- %s seconds ---" % (time.time() - start_time)
fig, axes = plt.subplots()
for i in range(n_trials):
t, traj = traj_list[i]
axes.plot(traj[:, 0], traj[:, 1])
axes.axis("equal")
axes.scatter(s_g[0], s_g[1])
if flag_with_obs:
for x, y, w, h in obs_list:
rect = Rectangle((x, y), w, h)
axes.add_patch(rect)
plt.show()
def simulate_naive_policy(n_trials, s_g, modality, usr):
planner = NaivePolicy()
model_path = "/home/yuhang/Documents/proactive_guidance/training_data/user" + str(usr)
sim = Simulator(planner, model_path)
traj_list = []
for i in range(n_trials):
traj_list.append(sim.run_trial((-1.0, 2.0, 0.0), s_g, modality, 20.0, tol=0.5))
fig, axes = plt.subplots()
for i in range(n_trials):
t, traj = traj_list[i]
axes.plot(traj[:, 0], traj[:, 1])
axes.axis("equal")
axes.scatter(s_g[0], s_g[1])
plt.show()
def validate_free_space_policy(planner, s_g, modality, path, model_path):
fig, axes = plt.subplots()
planner.visualize_policy(axes)
fig.savefig(path + "/value_func.png")
sim = Simulator(planner, model_path)
n_trials = 30
traj_list = []
for i in range(n_trials):
traj_list.append(sim.run_trial((-1.0, 2.0, 0.0), s_g, modality, 30.0, tol=0.5))
fig, axes = plt.subplots()
for i in range(n_trials):
t, traj = traj_list[i]
axes.plot(traj[:, 0], traj[:, 1])
axes.axis("equal")
axes.scatter(s_g[0], s_g[1])
fig.savefig(path + "/simulation.png")
def validate_free_space_policy(planner, s_g, modality, path, model_path):
fig, axes = plt.subplots()
planner.visualize_policy(axes)
fig.savefig(path + "/value_func.png")
sim = Simulator(planner, model_path)
n_trials = 30
traj_list = []
for i in range(n_trials):
traj_list.append(
sim.run_trial((-1.0, 2.0, 0.0), s_g, modality, 30.0, tol=0.5))
fig, axes = plt.subplots()
for i in range(n_trials):
t, traj = traj_list[i]
axes.plot(traj[:, 0], traj[:, 1])
axes.axis("equal")
axes.scatter(s_g[0], s_g[1])
fig.savefig(path + "/simulation.png")
def simulate_naive_policy(n_trials, s_g, modality, usr):
planner = NaivePolicy()
model_path = "/home/yuhang/Documents/proactive_guidance/training_data/user" + str(
usr)
sim = Simulator(planner, model_path)
traj_list = []
for i in range(n_trials):
traj_list.append(
sim.run_trial((-1.0, 2.0, 0.0), s_g, modality, 20.0, tol=0.5))
fig, axes = plt.subplots()
for i in range(n_trials):
t, traj = traj_list[i]
axes.plot(traj[:, 0], traj[:, 1])
axes.axis("equal")
axes.scatter(s_g[0], s_g[1])
plt.show()
