def landing_analysis_for_crash_comparison(dataset, keys=None):
fig = plt.figure()
ax = fig.add_subplot(111)
if keys is None:
classified_keys = fa.get_classified_keys(dataset)
keys = classified_keys['straight']
keys = dataset.trajecs.keys()
for key in keys:
trajec = dataset.trajecs[key]
ftp = np.arange(trajec.frames[0]-25, trajec.frames[-1]).tolist()
ax.plot( np.log(trajec.angle_subtended_by_post[ftp]), trajec.speed[ftp], color='black', linewidth=0.5, alpha=0.05)
keys_to_highlight = ['2_29065', '2_31060', '8_10323', '6_715']
for key in keys:
color = 'gray'
dotcolor = 'blue'
if key in keys_to_highlight:
color = 'black'
dotcolor = 'purple'
trajec = dataset.trajecs[key]
ftp = np.arange(trajec.frames[0]-25, trajec.frames[-1]).tolist()
ax.plot( np.log(trajec.angle_subtended_by_post[ftp]), trajec.speed[ftp], color=color, linewidth=0.5, alpha=1)
ax.plot( np.log(trajec.angle_at_deceleration), trajec.speed_at_deceleration, '.', color=dotcolor, alpha=1)
fit, Rsq, x, y, yminus, yplus = fa.get_angle_vs_speed_curve(dataset, plot=False)
ax.plot( x, y, color='purple')
ax.fill_between(x, yplus, yminus, color='purple', linewidth=0, alpha=0.2)
fa.fix_angle_log_spine(ax, histograms=False)
fig.savefig('landing_for_crash_comparison.pdf', format='pdf')
def crash_analysis(dataset, dataset_landing, keys=None):
fig = plt.figure()
ax = fig.add_subplot(111)
keys = dataset.trajecs.keys()
for key in keys:
trajec = dataset.trajecs[key]
ftp = np.arange(trajec.frames[0], trajec.frames[-1]).tolist()
color = 'gray'
dotcolor = 'blue'
if key == '20101111_C001H001S0045':
color = 'purple'
dotcolor = 'purple'
if trajec.angle_at_deceleration*180/np.pi > 90:
print key
ax.plot( np.log(trajec.angle_subtended_by_post[ftp]), trajec.speed[ftp], color=color, linewidth=0.5, alpha=1)
ax.plot( np.log(trajec.angle_at_deceleration), trajec.speed_at_deceleration, '.', color=dotcolor, alpha=1)
fit, Rsq, x, y, yminus, yplus = fa.get_angle_vs_speed_curve(dataset_landing, plot=False)
ax.plot( x, y, color='purple')
ax.fill_between(x, yplus, yminus, color='purple', linewidth=0, alpha=0.2)
fa.fix_angle_log_spine(ax, histograms=False)
fig.savefig('crash_spagetti.pdf', format='pdf')
def landing_spagetti_plots(dataset, gain=[170000], keys=None):
fig = plt.figure()
ax = fig.add_subplot(111)
fig2 = plt.figure()
ax2 = fig2.add_subplot(111)
fig3 = plt.figure()
ax3 = fig3.add_subplot(111)
if keys is None:
classified_keys = fa.get_classified_keys(dataset)
keys = classified_keys['straight']
#keys = dataset.trajecs.keys()
#keys_to_highlight = [keys[2], keys[0], keys[4], keys[6]]
keys_to_highlight = ['2_29065', '2_31060', '8_10323', '6_715']
points_at_deceleration = True
for key in keys:
trajec = dataset.trajecs[key]
ftp = np.arange(trajec.frames[0]-25, trajec.frame_of_landing-1).tolist()
#ftp = np.arange(trajec.frame_at_deceleration-10, trajec.frame_of_landing-1).tolist()
if key in keys_to_highlight:
color = 'black'
zorder = 10
else:
color = 'blue'
zorder = 1
ax.plot( np.log(trajec.angle_subtended_by_post[ftp]), trajec.speed[ftp], color=color, linewidth=0.5, alpha=1, zorder=zorder)
if points_at_deceleration and key in keys_to_highlight:
ax.plot( np.log(trajec.angle_at_deceleration), trajec.speed_at_deceleration, '.', color='black', alpha=1, zorder=zorder)
ax3.plot( np.log(trajec.angle_subtended_by_post[ftp]), trajec.speed[ftp], color='black', linewidth=0.5, alpha=1)
if 0:
# simulate
angle, speed, expansion = sim_deceleration(trajec, gain)
if speed[0] > 0.1:
ax2.plot( np.log(angle), speed, color='black', linewidth=0.5, alpha=1)
ax3.plot( np.log(angle), speed, color='black', linewidth=0.5, alpha=1)
f = np.where( np.abs(speed-speed[0])>0 )[0][0]
#ax2.plot( np.log(angle[f]), speed[f], '.', color='blue')
fit, Rsq, x, y, yminus, yplus = fa.get_angle_vs_speed_curve(dataset, plot=False)
ax2.plot( x, y, color='purple', alpha=1)
ax2.fill_between(x, yplus, yminus, color='purple', linewidth=0, alpha=0.2)
ax.plot( x, y, color='purple', alpha=1)
ax.fill_between(x, yplus, yminus, color='purple', linewidth=0, alpha=0.2)
fa.fix_angle_log_spine(ax, histograms=False)
fig.savefig('deceleration_real.pdf', format='pdf')
fa.fix_angle_log_spine(ax2, histograms=False)
fig2.savefig('deceleration_sim.pdf', format='pdf')
fa.fix_angle_log_spine(ax3, histograms=False)
fig3.savefig('deceleration_comparison.pdf', format='pdf')
def neural_threshold_tti_vs_rsdet_models(dataset_landing, save_plot=True, movie_dataset=None, ttc=None):
distfig = plt.figure()
distax = distfig.add_subplot(111)
radius = 0.009565
a = np.linspace(0,2.5,100)
fit, Rsq, x, y, yminus, yplus = fa.get_angle_vs_speed_curve(dataset_landing, plot=False, plot_sample_trajecs=False, post_type=['checkered', 'checkered_angled', 'black', 'black_angled'], filename=None, keys=None, tti=None, color_code_posts=False)
std = np.mean(yplus - y)
# RSDET model
m = fit[0]
b = fit[1]
vel = (m*np.log(a)+b)
print std, fit
expthreshold = expansion(vel, a=a)
expthreshold_plus = expansion(vel+std, a=a)
expthreshold_minus = expansion(vel-std, a=a)
distax.plot( np.log(a), expthreshold, color='purple')
distax.fill_between(np.log(a), expthreshold_plus, expthreshold_minus, color='purple', linewidth=0, alpha=0.3)
# True time-to-contact model
if ttc is None:
ttc, ttc_threshold = match_ttc_to_rsdet(ttc0=0.13, ttc_threshold0=0)
ttc_threshold = 0
expthreshold_ttc = expansion_from_timetocontact(ttc, a) + ttc_threshold
distax.plot( np.log(a), expthreshold_ttc, ':', color='purple')
# plot a sample constant velocity trajectory
vels = [0.2, 0.4, 0.8]
for vel in vels:
fps = 5000.0
x = np.arange(.2, 0.0, -vel/fps)
d = x+radius
a = 2*np.arcsin(radius / (d))
#exp = 2/np.sqrt(1-(radius/(d))**2) * (radius/(d)**2) * vel
exp = expansion(vel, a=a)
indices = np.where(exp<12)[0].tolist()
distax.plot( np.log(a[indices]), exp[indices], color='gray', linewidth=0.5)
# plot parameters
fa.fix_angle_log_spine(distax, histograms=False, set_y=False)
ylim_max = 1000
distax.set_ylim(0,ylim_max/180.*np.pi)
rad_ticks_y = np.linspace(0,ylim_max*np.pi/180.,5,endpoint=True)
deg_tick_strings_y = [str(s) for s in np.linspace(0,ylim_max,5,endpoint=True)]
for i, s in enumerate(deg_tick_strings_y):
deg_tick_strings_y[i] = s.split('.')[0]
distax.set_yticks(rad_ticks_y)
distax.set_yticklabels(deg_tick_strings_y)
distax.set_ylabel('Expansion, deg/s')
if save_plot:
distfig.savefig('neural_threshold_distance.pdf', format='pdf')
return
def landing(dataset_landing, movie_dataset, speed=0.2):
behavior = 'landing'
fps = 1000.
dt = 1/fps
r = 0.009565
radius = r
pos0 = [-0.2, 0]
vel = [speed, 0]
dvda = -0.2
nf = 5000
positions = np.zeros([nf, 2])
positions[0] = pos0
velocities = np.zeros([nf, 2])
velocities[0] = vel
speed = np.zeros([nf])
speed[0] = np.linalg.norm(velocities[0])
distance = np.zeros([nf])
angle_subtended_by_post = np.zeros([nf])
leg_ext = np.zeros([nf])
frames = [0]
frame_at_deceleration = None
deceleration_initiated = False
for f in range(1,nf):
if np.linalg.norm(positions[f-1])-radius <= 0.0001:
landed = True
else:
landed = False
if not landed:
frames.append(f)
positions[f] = positions[f-1] + velocities[f-1]*dt
distance[f] = np.linalg.norm(positions[f]) - radius
angle_subtended_by_post[f] = 2*np.arcsin( radius / (distance[f]+radius) )
if f>5:
#velocities[f,0] = -.21*np.log(angle_subtended_by_post[f])+.2
da = np.log(angle_subtended_by_post[f])-np.log(angle_subtended_by_post[f-1])
a = angle_subtended_by_post
af = np.min([a[f], 3])
exp0 = (a[f]-a[f-1])/dt #/ (-2.*np.tan(a[f]/2.))
exp1 = (a[f-1]-a[f-2])/dt #/ (-2.*np.tan(a[f-1]/2.))
m = -0.21/radius
b = 0.159/radius
expthreshold = (m*np.log(af)+b)*(2*np.tan(af/2.)*np.sin(af/2.))
exp0 -= expthreshold
exp1 -= expthreshold
exp0 = np.max([exp0, 0])
exp1 = np.max([exp1, 0])
#c = -1*exp0 / 3500.
dda = (exp1-exp0)/dt
c = dda / 150000.
print dda, velocities[f-1,0]
c = np.min([c,0])
v = np.max([speed[f-1] + c, 0.0])
velocities[f,0] = v
else:
velocities[f] = velocities[f-1]
speed[f] = np.linalg.norm(velocities[f])
if speed[f] > -0.21*np.log(angle_subtended_by_post[f])+0.159:
deceleration_initiated = True
if frame_at_deceleration is None:
frame_at_deceleration = f
else:
deceleration_initiated = False
if angle_subtended_by_post[f]*180/np.pi > 70 or np.isnan(angle_subtended_by_post[f]):
leg_ext[f] = 1
fig2 = plt.figure()
ax2 = fig2.add_subplot(111)
fit, Rsq, x, y, yminus, yplus = fa.get_angle_vs_speed_curve(dataset_landing, plot=False)
ax2.plot( x, y, color='blue', alpha=0.3)
ax2.fill_between(x, yplus, yminus, color='blue', linewidth=0, alpha=0.2)
angle_at_leg_extension, bins, data_filtered, xvals = fa.leg_extension_angle_histogram(movie_dataset, plot=False)
ax2.plot(xvals, data_filtered, color='red', alpha=0.3)
ax2.fill_between(xvals, data_filtered, np.zeros_like(xvals), color='red', linewidth=0, alpha=0.2)
ax2.plot(np.log(angle_subtended_by_post), speed, color='black')
fa.fix_angle_log_spine(ax2, histograms=False)
fig2.subplots_adjust(bottom=0.3, top=0.8, right=0.9, left=0.25)
filename = 'landing_cartoon_plot.pdf'
fig2.savefig(filename, format='pdf')
