def plot_err_histogram(rel_list, abs_list, title):
# plot relative error histogram.
max_err = 1.0
bin_width = 0.05 # relative err.
bins = np.arange(0. - bin_width / 2. - max_err, max_err + 2 * bin_width,
bin_width)
hist, bin_edges = np.histogram(np.array(rel_list), bins)
mpu.figure()
mpu.plot_histogram(bin_edges[:-1] + bin_width / 2.,
hist,
width=bin_width * 0.8,
xlabel='Relative Error',
plot_title=title)
# plot relative error histogram.
max_err = 20
bin_width = 2 # relative err.
bins = np.arange(0. - bin_width / 2. - max_err, max_err + 2 * bin_width,
bin_width)
hist, bin_edges = np.histogram(np.array(abs_list), bins)
mpu.figure()
mpu.plot_histogram(bin_edges[:-1] + bin_width / 2.,
hist,
width=bin_width * 0.8,
xlabel='Absolute Error',
plot_title=title)
def plot_workspace(pts, ha, z):
if pts.shape[1] == 0:
return
mpu.figure()
good_location = pts.mean(1)
mpu.plot_yx(pts[1, :].A1, pts[0, :].A1, label="ha:%.1f" % (math.degrees(ha)), axis="equal", linewidth=0)
mpu.plot_yx(good_location[1, :].A1, good_location[0, :].A1, axis="equal", linewidth=0, scatter_size=90, color="k")
mpu.savefig("z%.2f_ha%.1f.png" % (z, math.degrees(ha)))
def plot_workspace(pts, ha, z):
if pts.shape[1] == 0:
return
mpu.figure()
good_location = pts.mean(1)
mpu.plot_yx(pts[1, :].A1,
pts[0, :].A1,
label='ha:%.1f' % (math.degrees(ha)),
axis='equal',
linewidth=0)
mpu.plot_yx(good_location[1, :].A1,
good_location[0, :].A1,
axis='equal',
linewidth=0,
scatter_size=90,
color='k')
mpu.savefig('z%.2f_ha%.1f.png' % (z, math.degrees(ha)))
def plot_err_histogram(rel_list, abs_list, title):
# plot relative error histogram.
max_err = 1.0
bin_width = 0.05 # relative err.
bins = np.arange(0.-bin_width/2.-max_err, max_err+2*bin_width, bin_width)
hist, bin_edges = np.histogram(np.array(rel_list), bins)
mpu.figure()
mpu.plot_histogram(bin_edges[:-1]+bin_width/2., hist,
width=bin_width*0.8, xlabel='Relative Error',
plot_title=title)
# plot relative error histogram.
max_err = 20
bin_width = 2 # relative err.
bins = np.arange(0.-bin_width/2.-max_err, max_err+2*bin_width, bin_width)
hist, bin_edges = np.histogram(np.array(abs_list), bins)
mpu.figure()
mpu.plot_histogram(bin_edges[:-1]+bin_width/2., hist,
width=bin_width*0.8, xlabel='Absolute Error',
plot_title=title)
def plot_radial_tangential(mech_dict, savefig, fig_name=''):
radial_mech = mech_dict['force_rad_list']
tangential_mech = mech_dict['force_tan_list']
typ = mech_dict['mech_type']
mech_x = mech_dict['mechanism_x']
if typ == 'rotary':
mech_x_degrees = np.degrees(mech_x)
xlabel = 'angle (degrees)'
else:
mech_x_degrees = mech_x
xlabel = 'distance (meters)'
mpu.pl.clf()
mpu.plot_yx(radial_mech, mech_x_degrees, axis=None, label='radial force',
xlabel=xlabel, ylabel='N', color='r')
mpu.plot_yx(tangential_mech, mech_x_degrees, axis=None, label='tangential force',
xlabel=xlabel, ylabel='N', color='g')
mpu.legend()
if typ == 'rotary':
mpu.figure()
rad = mech_dict['radius']
torques_1 = rad * np.array(tangential_mech)
torques_3 = np.array(mech_dict['moment_tip_list']) + torques_1
mpu.plot_yx(torques_1, mech_x_degrees, axis=None,
label='torque from tangential',
xlabel=xlabel, ylabel='moment', color='r')
mpu.plot_yx(torques_3, mech_x_degrees, axis=None,
label='total torque',
xlabel=xlabel, ylabel='moment', color='y')
mpu.legend()
if savefig:
mpu.savefig(opt.dir+'/%s_force_components.png'%fig_name)
else:
mpu.show()
print 'Estimated rad, cx, cy:', rad, cx, cy
start_angle = tr.angle_within_mod180(
math.atan2(pts_2d[1, 0] - cy, pts_2d[0, 0] - cx) - math.pi / 2)
end_angle = tr.angle_within_mod180(
math.atan2(pts_2d[1, -1] - cy, pts_2d[0, -1] - cx) - math.pi / 2)
subsample_ratio = 1
pts_2d_s = pts_2d[:, ::subsample_ratio]
cep_force_clean, force_new = filter_trajectory_force(
eq_cartesian, d['force'])
cep_2d = np.matrix(cep_force_clean.p_list).T[0:2, reject_idx:]
# first draw the entire CEP and end effector trajectories
mpu.figure()
mpu.plot_yx(pts_2d_s[1, :].A1,
pts_2d_s[0, :].A1,
color='b',
label='\huge{End Effector Trajectory}',
axis='equal',
alpha=1.0,
scatter_size=7,
linewidth=0,
marker='x',
marker_edge_width=1.5)
cep_2d_s = cep_2d[:, ::subsample_ratio]
mpu.plot_yx(cep_2d_s[1, :].A1,
cep_2d_s[0, :].A1,
color='g',
z = -0.23
k = d.keys()
k_idx = np.argmin(np.abs(np.array(k)-z))
pts = d[k[k_idx]]
# visualize_boundary()
for kk in k:
pts = d[kk]
bpts = compute_boundary(pts)
cx,cy = 0.7,-0.6
rad = 0.4
# pts_in = point_contained(cx,cy,0.,rad,pts,
# start_angle=math.radians(140),
# end_angle=math.radians(190))
mpu.figure()
mpu.plot_yx(pts[1,:].A1,pts[0,:].A1,linewidth=0)
mpu.plot_yx(bpts[1,:].A1,bpts[0,:].A1,linewidth=0,color='y')
# mpu.plot_yx(pts_in[1,:].A1,pts_in[0,:].A1,linewidth=0,color='g')
# mpu.plot_yx([cy],[cx],linewidth=0,color='r')
mpu.show()
### apologies for the poor name. computes the translation and rotation
## of the torso frame that move the eq pt away from closest boundary
## and rotate such that local x axis is perp to mechanism
## returns 2x1 np matrix, angle
#def segway_motion_repulse(curr_pos_tl,cx_tl,cy_tl,cy_ts,start_pos_ts,
# eq_pt_tl,bndry):
def plot_radii(csv_data, color='#3366FF'):
keys = ['radius', 'type', 'name', 'repetition']
llists = expand(extract_keys(csv_data, keys), keys)
rad_list = np.array([float(r) for r in llists[0]]) / 100.0
types = llists[1]
names = np.array(llists[2])
all_types = set(types)
print 'Radii types', all_types
types_arr = np.array(types)
# np.where(types_arr == 'C')
cabinet_rad_list = rad_list[np.where(types_arr == 'C')[0]]
others_rad_list = rad_list[np.where(types_arr != 'C')[0]]
other_names = names[np.where(types_arr != 'C')[0]]
print '>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>'
print 'radii other names'
for i, n in enumerate(other_names):
print n, others_rad_list[i]
print '>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>'
rad_lists = [rad_list, cabinet_rad_list, others_rad_list]
titles = ['Radii of Rotary Mechanisms', 'Radii of Cabinets', 'Radii of Other Mechanisms']
bin_width = 0.05
max_radius = np.max(rad_list)
print 'MIN RADIUS', np.min(rad_list)
print 'MAX RADIUS', max_radius
mpu.set_figure_size(5.,5.)
for idx, radii in enumerate(rad_lists):
f = pb.figure()
f.set_facecolor('w')
bins = np.arange(0.-bin_width/2., max_radius+2*bin_width, bin_width)
hist, bin_edges = np.histogram(radii, bins)
h = mpu.plot_histogram(bin_edges[:-1]+bin_width/2., hist,
width=0.8*bin_width, xlabel='Radius (meters)',
ylabel='\# of mechanisms',
plot_title=titles[idx],
color=color, label='All')
pb.xlim(.1, 1.)
pb.ylim(0, 55)
mpu.legend(display_mode = 'less_space', handlelength=1.)
#-- different classes in different colors in the same histogram.
f = pb.figure()
f.set_facecolor('w')
bins = np.arange(0.-bin_width/2., max_radius+2*bin_width, bin_width)
hist, bin_edges = np.histogram(rad_lists[0], bins)
h = mpu.plot_histogram(bin_edges[:-1]+bin_width/2., hist,
width=0.8*bin_width, xlabel='Radius (meters)',
ylabel='\# of mechanisms',
plot_title=titles[1],
color='g', label='Cabinets')
hist, bin_edges = np.histogram(rad_lists[2], bins)
h = mpu.plot_histogram(bin_edges[:-1]+bin_width/2., hist,
width=0.8*bin_width, xlabel='Radius (meters)',
ylabel='\# of mechanisms',
plot_title='Cabinets and Other Mechanisms',
color='y', label='Other')
pb.xlim(.1, 1.)
pb.ylim(0, 55)
mpu.legend(display_mode = 'less_space', handlelength=1.)
color_list = ['g', 'b', 'r']
marker_list = ['s', '^', 'v']
label_list = ['All', 'Cabinets', 'Other']
scatter_size_list = [8, 5, 5]
mpu.set_figure_size(5.,5.)
mpu.figure()
for idx, radii in enumerate(rad_lists):
bins = np.arange(0.-bin_width/2., max_radius+2*bin_width, bin_width)
hist, bin_edges = np.histogram(radii, bins)
bin_midpoints = np.arange(0., max_radius+bin_width, bin_width)
mpu.plot_yx(hist, bin_midpoints, color = color_list[idx],
alpha = 0.6, marker = marker_list[idx],
scatter_size = scatter_size_list[idx], xlabel='Radius (meters)',
ylabel='\# of mechanisms', label = label_list[idx])
mpu.legend(display_mode = 'less_space')
def plot_radii(csv_data, color='#3366FF'):
keys = ['radius', 'type', 'name', 'repetition']
llists = expand(extract_keys(csv_data, keys), keys)
rad_list = np.array([float(r) for r in llists[0]]) / 100.0
types = llists[1]
names = np.array(llists[2])
all_types = set(types)
print 'Radii types', all_types
types_arr = np.array(types)
# np.where(types_arr == 'C')
cabinet_rad_list = rad_list[np.where(types_arr == 'C')[0]]
others_rad_list = rad_list[np.where(types_arr != 'C')[0]]
other_names = names[np.where(types_arr != 'C')[0]]
print '>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>'
print 'radii other names'
for i, n in enumerate(other_names):
print n, others_rad_list[i]
print '>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>'
rad_lists = [rad_list, cabinet_rad_list, others_rad_list]
titles = [
'Radii of Rotary Mechanisms', 'Radii of Cabinets',
'Radii of Other Mechanisms'
]
bin_width = 0.05
max_radius = np.max(rad_list)
print 'MIN RADIUS', np.min(rad_list)
print 'MAX RADIUS', max_radius
mpu.set_figure_size(5., 5.)
for idx, radii in enumerate(rad_lists):
f = pb.figure()
f.set_facecolor('w')
bins = np.arange(0. - bin_width / 2., max_radius + 2 * bin_width,
bin_width)
hist, bin_edges = np.histogram(radii, bins)
h = mpu.plot_histogram(bin_edges[:-1] + bin_width / 2.,
hist,
width=0.8 * bin_width,
xlabel='Radius (meters)',
ylabel='\# of mechanisms',
plot_title=titles[idx],
color=color,
label='All')
pb.xlim(.1, 1.)
pb.ylim(0, 55)
mpu.legend(display_mode='less_space', handlelength=1.)
#-- different classes in different colors in the same histogram.
f = pb.figure()
f.set_facecolor('w')
bins = np.arange(0. - bin_width / 2., max_radius + 2 * bin_width,
bin_width)
hist, bin_edges = np.histogram(rad_lists[0], bins)
h = mpu.plot_histogram(bin_edges[:-1] + bin_width / 2.,
hist,
width=0.8 * bin_width,
xlabel='Radius (meters)',
ylabel='\# of mechanisms',
plot_title=titles[1],
color='g',
label='Cabinets')
hist, bin_edges = np.histogram(rad_lists[2], bins)
h = mpu.plot_histogram(bin_edges[:-1] + bin_width / 2.,
hist,
width=0.8 * bin_width,
xlabel='Radius (meters)',
ylabel='\# of mechanisms',
plot_title='Cabinets and Other Mechanisms',
color='y',
label='Other')
pb.xlim(.1, 1.)
pb.ylim(0, 55)
mpu.legend(display_mode='less_space', handlelength=1.)
color_list = ['g', 'b', 'r']
marker_list = ['s', '^', 'v']
label_list = ['All', 'Cabinets', 'Other']
scatter_size_list = [8, 5, 5]
mpu.set_figure_size(5., 5.)
mpu.figure()
for idx, radii in enumerate(rad_lists):
bins = np.arange(0. - bin_width / 2., max_radius + 2 * bin_width,
bin_width)
hist, bin_edges = np.histogram(radii, bins)
bin_midpoints = np.arange(0., max_radius + bin_width, bin_width)
mpu.plot_yx(hist,
bin_midpoints,
color=color_list[idx],
alpha=0.6,
marker=marker_list[idx],
scatter_size=scatter_size_list[idx],
xlabel='Radius (meters)',
ylabel='\# of mechanisms',
label=label_list[idx])
mpu.legend(display_mode='less_space')
mpu.plot_yx(good_location[1, :].A1,
good_location[0, :].A1,
axis='equal',
linewidth=0,
scatter_size=90,
color='k')
mpu.savefig('z%.2f_ha%.1f.png' % (z, math.degrees(ha)))
argv = sys.argv
fname = sys.argv[1]
dd = ut.load_pickle(fname)
color_list = ['b', 'y', 'g']
i = 0
mpu.figure(dpi=100)
for ha in dd.keys():
d = dd[ha]
l = []
key_list = d['pts'].keys()
for k in key_list:
pts = d['pts'][k]
l.append(pts.shape[1])
#plot_workspace(pts,ha,k)
ll = zip(key_list, l)
ll.sort()
key_list, l = zip(*ll)
if ha == 0:
label = 'Hook Left'
def split_forces_hooktip_test(hand_mat):
kin_dict = ke.fit(hand_mat, tup)
center_hand = np.matrix((kin_dict['cx'], kin_dict['cy'],
kin_dict['cz'])).T
print 'kin_dict:', kin_dict
# radial vectors.
radial_mat = hand_mat - center_hand
radial_mat = radial_mat / ut.norm(radial_mat)
# cannot use hook tip to compute mechanism angle because I
# don't have a way of knowing when the hook starts opening the
# mechanism. (Think hook makes contact with door, moves in
# freespace and then makes contact with the handle.)
#start_rad = radial_mat[:,0]
#ct = (start_rad.T * radial_mat).A1
#st = ut.norm(np.matrix(np.cross(start_rad.A1, radial_mat.T.A)).T).A1
#mech_angle_l = np.arctan2(st, ct).tolist()
_, nrm_hand = project_points_plane(hand_mat)
print 'nrm_hand:', nrm_hand.A1
f_cam_l = []
m_cam_l = []
m_base_l = []
frad_l = []
ftan_l = []
hook_num = cd['hook_checker_number']
print 'hook_num:', hook_num
for i, f in enumerate(force_mat.T):
f = f.T
m = moment_mat[:,i]
f_cam, m_cam, m_base = ft_to_camera_3(f, m, hook_rot_l[i], hook_num,
return_moment_cam = True)
f_cam_l.append(f_cam)
m_cam_l.append(abs((m_cam.T*nrm_hand)[0,0]))
m_base_l.append(abs((m_base.T*nrm_hand)[0,0]))
#m_base_l.append(np.linalg.norm(f))
tangential_vec = np.matrix(np.cross(radial_mat[:,i].A1, nrm_hand.A1)).T
ftan = (f_cam.T * tangential_vec)[0,0]
ftan_l.append(ftan)
#frad = np.linalg.norm(f_cam - tangential_vec * ftan)
frad = (f_cam.T*radial_mat[:,i])[0,0]
frad_l.append(abs(frad))
fig1 = mpu.figure()
mech_ang_deg = np.degrees(mech_angle_l)
mpu.plot_yx(ftan_l, mech_ang_deg, label='Tangential Force (hook tip)', color='b')
mpu.plot_yx(frad_l, mech_ang_deg, label='Radial Force (hook tip)', color='y')
mech_pkl_name = glob.glob(opt.dir + '/open_mechanism_trajectories_*.pkl')[0]
md = ut.load_pickle(mech_pkl_name)
radial_mech = md['force_rad_list']
tangential_mech = md['force_tan_list']
mech_x = np.degrees(md['mechanism_x'])
mpu.plot_yx(tangential_mech, mech_x, label='Tangential Force (mechanism checker)', color='g')
mpu.plot_yx(radial_mech, mech_x, label='Radial Force (mechanism checker)', color='r')
mpu.legend()
fig2 = mpu.figure()
mpu.plot_yx(m_cam_l, mech_ang_deg, label='\huge{$m_{axis}$}')
mpu.plot_yx(m_base_l, mech_ang_deg, label='\huge{$m^s$}',
color = 'r')
mpu.legend()
mpu.show()
cd = ut.load_pickle(glob.glob(opt.dir + '/combined_log*.pkl')[0])
tup = ke.init_ros_node()
ma2 = compute_mech_angle_2(cd, tup, project_plane=False)
ma1 = compute_mech_angle_1(cd)
ma3 = compute_mech_angle_2(cd, tup, project_plane=True)
# ma4 = compute_mech_angle_1(cd, min_evec)
lab1 = 'orientation only'
lab2 = 'checker origin position + circle fit'
lab3 = 'checker origin position + PCA projection + circle fit'
# lab4 = 'PCA projection + orientation'
mpu.figure()
mpu.plot_yx(np.degrees(ma3), color='r', label=lab3,
linewidth = 1, scatter_size = 5)
mpu.plot_yx(np.degrees(ma2), color='b', label=lab2,
linewidth = 1, scatter_size = 5)
mpu.plot_yx(np.degrees(ma1), color='y', label=lab1,
linewidth = 1, scatter_size = 5)
mpu.legend()
vel3 = ma.compute_velocity(ma3, cd['time_list'], 1)
vel2 = ma.compute_velocity(ma2, cd['time_list'], 1)
vel1 = ma.compute_velocity(ma1, cd['time_list'], 1)
mpu.figure()
mpu.plot_yx(np.degrees(vel3), np.degrees(ma3), color='r',
label=lab3, linewidth = 1, scatter_size = 5)
mpu.plot_yx(np.degrees(vel2), np.degrees(ma2), color='b',
if pts.shape[1] == 0:
return
mpu.figure()
good_location = pts.mean(1)
mpu.plot_yx(pts[1, :].A1, pts[0, :].A1, label="ha:%.1f" % (math.degrees(ha)), axis="equal", linewidth=0)
mpu.plot_yx(good_location[1, :].A1, good_location[0, :].A1, axis="equal", linewidth=0, scatter_size=90, color="k")
mpu.savefig("z%.2f_ha%.1f.png" % (z, math.degrees(ha)))
argv = sys.argv
fname = sys.argv[1]
dd = ut.load_pickle(fname)
color_list = ["b", "y", "g"]
i = 0
mpu.figure(dpi=100)
for ha in dd.keys():
d = dd[ha]
l = []
key_list = d["pts"].keys()
for k in key_list:
pts = d["pts"][k]
l.append(pts.shape[1])
# plot_workspace(pts,ha,k)
ll = zip(key_list, l)
ll.sort()
key_list, l = zip(*ll)
if ha == 0:
label = "Hook Left"
