def evaluate(self, p, sim, plt):
start_time = time.time()
while time.time() - start_time < p.T:
sim.run(p.dt, progress_bar=False)
#sim.run(p.T)
self.link.write(self.path + 'duty_cycle_sp', '0')
if plt is not None:
plt.plot(sim.data[self.p_t], sim.data[self.p_desired])
plt.plot(sim.data[self.p_t], sim.data[self.p_q])
#plt.plot(sim.data[self.p_t], sim.data[self.p_u])
rate = len(sim.data[self.p_t]) / float(p.T)
q = sim.data[self.p_q][:,0]
d = sim.data[self.p_desired][:,0]
t = sim.data[self.p_t][:,0]
offset = benchmark.find_offset(q, d)
#offset = 1
delay = np.mean(t[offset:]-t[:-offset])
diff = d[:-offset] - q[offset:]
N = len(q) / 2
diff = diff[N:]
rmse = np.sqrt(np.mean(diff**2))
return dict(rate=rate, delay=delay, rmse=rmse)
def evaluate(self, p, sim, plt):
#start_time = time.time()
#while time.time() - start_time < p.T:
# sim.run(p.dt, progress_bar=False)
sim.run(p.T)
#q = sim.data[self.p_q][:,0]
q = np.array(self.system_state)[:,0]
d = np.array(self.system_desired)[:,0]
t = np.array(self.system_times)
N = len(q) / 2
# find an offset that lines up the data best (this is the delay)
offsets = []
for i in range(p.D):
#q = sim.data[self.p_q][:,i]
q = np.array(self.system_state)[:,i]
d = np.array(self.system_desired)[:,i]
offset = benchmark.find_offset(q[N:], d[N:])
if offset == 0:
offset = 1
offsets.append(offset)
offset = int(np.mean(offsets))
delay = np.mean(t[offset:]-t[:-offset])
spinn_dt = np.mean(t[1:]-t[:-1])
if plt is not None:
plt.plot(t[offset:], d[:-offset], label='$q_d$')
#plt.plot(t[offset:], d[offset:])
t2 = self.system_times
plt.plot(t2[offset:], q[offset:], label='$q$')
plt.legend(loc='upper left')
#plt.plot(np.correlate(d, q, 'full')[len(q):])
diff = np.array(self.system_desired)[:-offset] - np.array(self.system_state)[offset:]
diff = diff[N:]
rmse = np.sqrt(np.mean(diff.flatten()**2))
return dict(delay=delay, rmse=rmse, spinn_dt=spinn_dt)
