def generate_constraints(self, **kwargs):
r_max = kwargs['r_max']
max_cap = kwargs['max_cap']
candidate_stops = kwargs['candidate_stops']
demand_centroids = kwargs['demand_centroids']
demands = kwargs['demands']
n_buses = kwargs['n_buses']
n_candidate_stops = len(candidate_stops)
n_demands = len(demands)
costs = []
for i in range(n_demands + n_candidate_stops):
costs.append([])
for j in range(n_demands + n_candidate_stops):
org = demand_centroids[i] if i < n_demands else candidate_stops[i-n_demands]
des = demand_centroids[j] if j < n_demands else candidate_stops[j-n_demands]
if i < n_demands or j < n_demands:
costs[i].append(walktime(org, des))
else:
costs[i].append(ridetime(i - n_demands, j - n_demands))
def it(spec):
lengths = {'d': n_demands, 's': n_candidate_stops, 'n': n_demands+n_candidate_stops, 'r': r_max}
if len(spec) == 1:
return enumerate(range(lengths[spec]))
return enumerate(product(*[range(lengths[x]) for x in spec]))
vars = self.vars
cs = []
cns = []
#///////////////////////////////////////////////////////////////////////
# Origin destination shortest path problem (ODSSP)
#///////////////////////////////////////////////////////////////////////
for idx, (o, d) in it('dd'):
if o != d:
cns.append('A1(%d)' % idx)
cs.append((
linex([(vars.x[o][d][o][j], 1)
for j in range(n_demands + n_candidate_stops)
if j != o]),
'E', 1, 0))
else:
cns.append('A1(%d)' % idx)
cs.append((
linex([(vars.x[o][d][o][j], 1)
for j in range(n_demands + n_candidate_stops)]),
'E', 0, 0))
for idx, (o, d) in it('dd'):
if o != d:
cns.append('A2(%d)' % idx)
cs.append((
linex([(vars.x[o][d][i][d], 1)
for i in range(n_demands + n_candidate_stops)
if i != d]),
'E', 1, 0))
else:
cns.append('A2(%d)' % idx)
cs.append((
linex([(vars.x[o][d][i][d], 1)
for i in range(n_demands + n_candidate_stops)]),
'E', 0, 0))
for idx, (o, d, i) in it('ddn'):
if o != d:
rhs = 0
if i == o:
rhs = 1
elif i == d:
rhs = -1
cns.append('A3(%d)' % idx)
cs.append((
linex([(vars.x[o][d][i][j], 1)
for j in range(n_demands + n_candidate_stops)
if j != i] +
[(vars.x[o][d][j][i], -1)
for j in range(n_demands + n_candidate_stops)
if j != i]),
'E', rhs, 0))
for idx, (o,d,j) in it('dds'):
cns.append('A4(%d)' % idx)
cs.append((
linex([(vars.x[o][d][i][j+n_demands], 1)
for i in range(n_demands)] +
[(vars.x[o][d][j+n_demands][i], 1)
for i in range(n_demands)]),
'L', 1, 0))
for idx, (o,d) in it('dd'):
if o != d:
cns.append('A5(%d)' % idx)
cs.append((
linex([(vars.w[o][d], 1)] +
[(vars.x[o][d][i][j], -costs[i][j])
for i in range(n_demands + n_candidate_stops)
for j in range(n_demands + n_candidate_stops)
if i == o or j == d]),
'E', 0, 0))
for idx, (o,d) in it('dd'):
if o != d:
cns.append('A6(%d)' % idx)
cs.append((
linex([(vars.r[o][d], 1)] +
[(vars.x[o][d][i+n_demands][j+n_demands], -costs[i+n_demands][j+n_demands])
for i in range(n_candidate_stops)
for j in range(n_candidate_stops)]),
'E', 0, 0))
for idx, (o,d,j) in it('dds'):
if costs[o][j+n_demands] > 5:
cns.append('A7(%d)' % idx)
cs.append((linex([(vars.x[o][d][o][j+n_demands], 1)]), 'E', 0, 0))
for idx, (o,d,i) in it('dds'):
if costs[i+n_demands][d] > 5:
cns.append('A8(%d)' % idx)
cs.append((linex([(vars.x[o][d][i+n_demands][d], 1)]), 'E', 0, 0))
for idx, (o,d,i,j) in it('ddnd'):
if j != d:
cns.append('A9(%d)' % idx)
cs.append((linex([(vars.x[o][d][i][j], 1)]), 'E', 0, 0))
for idx, (o,d,i,j) in it('dddn'):
if i != o:
cns.append('A10(%d)' % idx)
cs.append((linex([(vars.x[o][d][i][j], 1)]), 'E', 0, 0))
for idx, (o,d) in it('dd'):
if o != d:
if costs[o][d] == 0:
f = 1
else:
f = 1.0/costs[o][d]
cns.append(('A11-1(%d)' % idx))
cs.append((
linex([(vars.s[o][d], 1),
(vars.r[o][d], 1-f),
(vars.w[o][d], 1-f)]),
'E', costs[o][d]*(1-f), 0))
cns.append(('A12(%d)' % idx))
cs.append((
linex([(vars.r[o][d], 1),
(vars.w[o][d], 1)]),
'L', costs[o][d], 0))
else:
cns.append(('A11-2(%d)' % idx))
cs.append((linex([(vars.s[o][d], 1)]),'E', 0, 0))
#///////////////////////////////////////////////////////////////////////
# Transit route network design problem (TRNDP)
#///////////////////////////////////////////////////////////////////////
for idx, (r,i,j) in it('rss'):
cns.append('B1(%d)' % idx)
cs.append((
linex([(vars.rx0[r][i][j], 1),
(vars.c[r], 1)]),
'L', 1, 0))
for idx, (i,j) in it('ss'):
cns.append('B2(%d)' % idx)
cs.append((
linex([(vars.rx[r][i][j], 1)
for r in range(r_max)] +
[(vars.rx0[r][i][j], 1)
for r in range(r_max)] +
[(vars.x[o][d][i+n_demands][j+n_demands], -1.0/n_demands/n_demands)
for o in range(n_demands)
for d in range(n_demands)]),
'G', 0, 0))
for idx, (r) in it('r'):
cns.append('B3(%d)' % idx)
cs.append((
linex([(vars.rx[r][n_candidate_stops][j], 1)
for j in range(n_candidate_stops+1)]),
'E', 1, 0))
for idx, (r) in it('r'):
cns.append('B4(%d)' % idx)
cs.append((
linex([(vars.rx[r][i][n_candidate_stops], 1)
for i in range(n_candidate_stops+1)] +
[(vars.rx[r][i][n_candidate_stops+1], 1)
for i in range(n_candidate_stops)]),
'E', 1, 0))
for idx, (r,i) in it('rs'):
cns.append('B5(%d)' % idx)
cs.append((
linex([(vars.rx[r][i][j], 1)
for j in range(n_candidate_stops+2)
if i != j] +
[(vars.rx[r][j][i], -1)
for j in range(n_candidate_stops+2)
if i != j]),
'E', 0, 0))
for idx, (r,i) in it('rs'):
cns.append('B6(%d)' % idx)
cs.append((
linex([(vars.rx[r][j][i], 1)
for j in range(n_candidate_stops+2)]),
'L', 1, 0))
for idx, (r,i) in it('rs'):
cns.append('B7(%d)' % idx)
cs.append((
linex([(vars.rx[r][i][j], 1)
for j in range(n_candidate_stops+2)]),
'L', 1, 0))
for idx, (r,i) in it('rs'):
cns.append('B8(%d)' % idx)
cs.append((linex([(vars.rx[r][i][i], 1)]), 'E', 0, 0))
for idx, (r) in it('r'):
cns.append('B9(%d)' % r)
cs.append((linex([(vars.rx[r][n_candidate_stops][n_candidate_stops+1], 1)]), 'E', 0, 0))
for idx, (r,i,j) in it('rss'):
if i != j:
cns.append('B10(%d)' % idx)
cs.append((
linex([(vars.rq[r][i], 1),
(vars.rq[r][j], -1),
(vars.rx[r][i][j], 1000)]),
'L', 999, 0))
for idx, (r,i) in it('rs'):
if i != 0:
cns.append('B11-1(%d)' % idx)
cs.append((
linex([(vars.rx[r][i][n_candidate_stops], 1)] +
[(vars.rx[r][k][n_candidate_stops], 1)
for k in range(n_candidate_stops+1)
if k < i]),
'L', 1, 0))
for idx, (r,i) in it('rs'):
if i != 0:
cns.append('B11-2(%d)' % idx)
cs.append((
linex([(vars.rx[r][i][n_candidate_stops], 1)] +
[(vars.rx[r][j][i], -1)
for j in range(n_candidate_stops+1)]),
'L', 0, 0))
for idx, (r,i) in it('rs'):
if i != 0:
cns.append('B11-3(%d)' % idx)
cs.append((
linex([(vars.rx[r][i][n_candidate_stops], 1),
(vars.c[r], 1)]),
'L', 1, 0))
for idx, (r,i) in it('rs'):
if i != 0:
cns.append('B11-4(%d)' % idx)
cs.append((
linex([(vars.rx[r][i][n_candidate_stops], 1),
(vars.c[r], 1)] +
[(vars.rx[r][k][n_candidate_stops], 1)
for k in range(n_candidate_stops+1)
if k < i] +
[(vars.rx[r][j][i], -1)
for j in range(n_candidate_stops+1)]),
'G', 0, 0))
for idx, (r,i,j) in it('rss'):
cns.append('B12-1(%d)' % idx)
cs.append((
linex([(vars.rx0[r][i][j], 1),
(vars.rx[r][i][n_candidate_stops], -1),
(vars.rx[r][n_candidate_stops][j], -1)]),
'G', -1, 0))
for idx, (r,i,j) in it('rss'):
cns.append('B12-2(%d)' % idx)
cs.append((
linex([(vars.rx0[r][i][j], 1),
(vars.rx[r][i][n_candidate_stops], -1)]),
'L', 0, 0))
for idx, (r,i,j) in it('rss'):
cns.append('B12-3(%d)' % idx)
cs.append((
linex([(vars.rx0[r][i][j], 1),
(vars.rx[r][n_candidate_stops][j], -1)]),
'L', 0, 0))
for idx, (r,i,j) in it('rss'):
cns.append('B13(%d)' % idx)
cs.append((
linex([(vars.rx[r][i][j], 1),
(vars.rx0[r][i][j], 1)]),
'L', 1, 0))
#///////////////////////////////////////////////////////////////////////
# Frequency determination problem (FDP)
#///////////////////////////////////////////////////////////////////////
for idx, (r) in it('r'):
cns.append('C1(%d)' % r)
cs.append((
linex([(vars.c[r], 1)] +
[(vars.rx[r][i][n_candidate_stops+1], -1)
for i in range(n_candidate_stops)]),
'E', 0, 0))
for idx, (i,j) in it('ss'):
if i != j:
cns.append('C3-1(%d)' % idx)
cs.append((
linex([(vars.x[o][d][i+n_demands][j+n_demands], demands[o][d])
for o in range(n_demands)
for d in range(n_demands)] +
[(vars.z1[r][i][j], -max_cap)
for r in range(r_max)] +
[(vars.z3[r][i][j], -max_cap)
for r in range(r_max)]),
'L', 0, 0))
for idx, (r,i,j) in it('rss'):
if i != j:
cns.append('C3-2(%d)' % idx)
cs.append((
linex([(vars.z1[r][i][j], 1),
(vars.rx[r][i][j], -60)]),
'L', 0, 0))
for idx, (r,i,j) in it('rss'):
if i != j:
cns.append('C3-3(%d)' % idx)
cs.append((
linex([(vars.z1[r][i][j], 1),
(vars.f[r], -1)]),
'L', 0, 0))
for idx, (r,i,j) in it('rss'):
if i != j:
cns.append('C3-4(%d)' % idx)
cs.append((
linex([(vars.z1[r][i][j], 1),
(vars.f[r], -1),
(vars.rx[r][i][j], -60)]),
'G', -60, 0))
for idx, (r,i,j) in it('rss'):
if i != j:
cns.append('C3-5(%d)' % idx)
cs.append((
linex([(vars.z3[r][i][j], 1),
(vars.rx0[r][i][j], -60)]),
'L', 0, 0))
for idx, (r,i,j) in it('rss'):
if i != j:
cns.append('C3-6(%d)' % idx)
cs.append((
linex([(vars.z3[r][i][j], 1),
(vars.f[r], -1)]),
'L', 0, 0))
for idx, (r,i,j) in it('rss'):
if i != j:
cns.append('C3-7(%d)' % idx)
cs.append((
linex([(vars.z3[r][i][j], 1),
(vars.f[r], -1),
(vars.rx0[r][i][j], -60)]),
'G', -60, 0))
for idx, (r,i,j) in it('rss'):
if i != j:
cns.append('C3-8(%d)' % idx)
cs.append((
linex([(vars.z3[r][i][j], 1),
(vars.c[r], 1)]),
'L', 1, 0))
cns.append('C4-1')
cs.append((
linex([(vars.z1[r][i][j], 1)
for r in range(r_max)
for i in range(n_candidate_stops)
for j in range(n_candidate_stops)
if i != j] +
[(vars.z2[r][i][j], 1)
for r in range(r_max)
for i in range(n_candidate_stops)
for j in range(n_candidate_stops)
if i != j]),
'L', n_buses, 0))
for idx, (r,i,j) in it('rss'):
if i != j:
cns.append('C4-2(%d)' % idx)
cs.append((
linex([(vars.z2[r][i][j], 1),
(vars.rx[r][i][j], -60)]),
'L', 0, 0))
for idx, (r,i,j) in it('rss'):
if i != j:
cns.append('C4-3(%d)' % idx)
cs.append((
linex([(vars.z2[r][i][j], 1),
(vars.c[r], -60)]),
'L', 0, 0))
for idx, (r,i,j) in it('rss'):
if i != j:
cns.append('C4-4(%d)' % idx)
cs.append((
linex([(vars.z2[r][i][j], 1),
(vars.f[r], -1)]),
'L', 0, 0))
for idx, (r,i,j) in it('rss'):
if i != j:
cns.append('C4-5(%d)' % idx)
cs.append((
linex([(vars.z2[r][i][j], 1),
(vars.f[r], -1),
(vars.rx[r][i][j], -60),
(vars.c[r], -60)]),
'G', -120, 0))
for idx, (r) in it('r'):
cns.append('C5(%d)' % idx)
cs.append((linex([(vars.f[r], 1)]), 'G', 0.2, 0))
cs = zip(*cs)
self._cp.linear_constraints.add(
lin_expr=cs[0],
senses=cs[1],
rhs=cs[2],
range_values=cs[3])
self._cp.linear_constraints.set_names(enumerate(cns))
def generate_constraints(self):
r_max = 1
candidate_stops = [ #[55.785470, 12.523041],
[55.786239, 12.522234], [55.786541, 12.526961],
[55.789277, 12.523946], [55.786431, 12.521428],
[55.785898, 12.520618], [55.785138, 12.520728],
[55.784637, 12.520459], [55.782099, 12.514132],
[55.782460, 12.519244], [55.782226, 12.520035]
]
demand_centroids = [[55.786344, 12.524316], [55.786344, 12.524316],
[55.786872, 12.5259], [55.78944, 12.525146],
[55.786745, 12.520469], [55.787486, 12.518472],
[55.787911, 12.518621], [55.785438, 12.519377],
[55.785154, 12.520434], [55.785241, 12.518699],
[55.783, 12.512931], [55.781929, 12.521461]]
demands = [[0, 100, 3, 36, 0, 14, 13, 19, 42, 55, 5, 20],
[190, 0, 2, 25, 0, 18, 6, 8, 40, 37, 5, 4],
[13, 10, 0, 3, 0, 3, 0, 1, 5, 3, 0, 0],
[137, 1, 0, 0, 0, 10, 4, 1, 17, 34, 1, 1],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
[47, 9, 2, 4, 0, 0, 63, 0, 0, 6, 2, 0],
[46, 4, 0, 0, 0, 173, 0, 0, 0, 2, 2, 1],
[61, 2, 0, 1, 0, 1, 0, 0, 78, 116, 0, 0],
[116, 13, 1, 5, 0, 8, 0, 52, 0, 84, 16, 8],
[167, 6, 0, 6, 0, 33, 13, 52, 230, 0, 23, 6],
[1, 1, 0, 0, 0, 1, 0, 0, 7, 1, 0, 1],
[97, 55, 0, 6, 0, 3, 2, 2, 47, 39, 9, 0]]
n_candidate_stops = len(candidate_stops)
n_buses = 20
costs = []
for i1, (stop_x1, stop_y1) in enumerate([[0, 0]] + candidate_stops):
costs.append([])
for i2, (stop_x2,
stop_y2) in enumerate([[0, 0]] + candidate_stops):
costs[i1].append(
275 * (abs(stop_x1 - stop_x2) + abs(stop_y1 - stop_y2)))
vars = self.vars
cs = []
cns = []
idx = 0
for r in range(r_max):
for i in range(n_candidate_stops + 1):
for j in range(n_candidate_stops + 1):
cns.append('XF1(%d)' % idx)
cs.append((linex([(vars.xf[r][i][j], 1),
(vars.f[r], -1)]), 'L', 0, 0))
cns.append('XF2(%d)' % idx)
cs.append((linex([(vars.xf[r][i][j], 1),
(vars.x[r][i][j], -60)]), 'L', 0, 0))
cns.append('XF3(%d)' % idx)
cs.append((linex([(vars.xf[r][i][j], 1), (vars.f[r], -1),
(vars.x[r][i][j], -60)]), 'G', -60, 0))
idx += 1
for r in range(r_max):
cns.append('C1(%d)' % r)
cs.append(
(linex([(vars.x[r][0][j], 1)
for j in range(n_candidate_stops + 1)]), 'E', 1, 0))
for r in range(r_max):
cns.append('C2(%d)' % r)
cs.append(
(linex([(vars.x[r][i][0], 1)
for i in range(n_candidate_stops + 1)]), 'L', 1, 0))
idx = 0
for r in range(r_max):
for j in range(n_candidate_stops):
cns.append('C3(%d)' % idx)
cs.append((linex(
[(vars.x[r][i][j + 1], 1)
for i in range(n_candidate_stops + 1) if i != (j + 1)] +
[(vars.x[r][j + 1][i], -1)
for i in range(n_candidate_stops + 2) if i != (j + 1)]),
'E', 0, 0))
idx += 1
idx = 0
for r in range(r_max):
for j in range(n_candidate_stops + 1):
cns.append('C4(%d)' % idx)
cs.append((linex([(vars.x[r][i][j + 1], 1)
for i in range(n_candidate_stops + 1)
if i != (j + 1)]), 'L', 1, 0))
idx = 0
for r in range(r_max):
for i in range(n_candidate_stops):
cns.append('C5(%d)' % idx)
cs.append((linex([(vars.x[r][i + 1][j], 1)
for j in range(n_candidate_stops + 1)
if (i + 1) != j]), 'L', 1, 0))
idx = 0
for r in range(r_max):
cns.append('C7a(%d)' % idx)
cs.append((linex([(vars.x[r][i + 1][n_candidate_stops + 1], 1)
for i in range(n_candidate_stops)]), 'L', 1, 0))
cns.append('C7b(%d)' % idx)
cs.append(
(linex([(vars.x[r][i + 1][n_candidate_stops + 1], 1)
for i in range(n_candidate_stops)] +
[(vars.x[r][i][0], 1)
for i in range(n_candidate_stops + 1)]), 'E', 1, 0))
idx = 0
for r in range(r_max):
for i in range(n_candidate_stops):
cns.append('C6(%d)' % idx)
cs.append((linex([(vars.x[r][i + 1][i + 1], 1)]), 'E', 0, 0))
idx = 0
for r in range(r_max):
for i in range(n_candidate_stops + 1):
for j in range(n_candidate_stops + 1):
cns.append('C7-1(%d)' % idx)
cs.append((linex([(vars.zxf[r][i][j], 1),
(vars.xf[r][i][j], -1)]), 'L', 0, 0))
cns.append('C7-2(%d)' % idx)
cs.append((linex([(vars.zxf[r][i][j], 1),
(vars.x[r][0][0], 60)]), 'L', 60, 0))
cns.append('C7-3(%d)' % idx)
cs.append((linex([(vars.zxf[r][i][j], 1),
(vars.xf[r][i][j], -1),
(vars.x[r][0][0], 60)]), 'G', 0, 0))
idx += 1
idx = 0
for r in range(r_max):
for i in range(n_candidate_stops + 1):
for j in range(n_candidate_stops + 1):
cns.append('zzxf1(%d)' % idx)
cs.append((linex([(vars.zzxf[r][i][j], 1),
(vars.zxf[r][i][j], -1)]), 'L', 0, 0))
cns.append('zzxf2(%d)' % idx)
cs.append(
(linex([(vars.zzxf[r][i][j], 1)] +
[(vars.x[r][i + 1][n_candidate_stops + 1], -60)
for i in range(n_candidate_stops)]), 'L', 0,
0))
cns.append('zzxf3(%d)' % idx)
cs.append(
(linex([(vars.zzxf[r][i][j], 1), (vars.zxf[r][i][j],
-1)] +
[(vars.x[r][i + 1][n_candidate_stops + 1], -60)
for i in range(n_candidate_stops)]), 'G', -60,
0))
idx += 1
cns.append('C8')
cs.append(
(linex([(vars.zxf[r][i + 1][j + 1], costs[i + 1][j + 1])
for r in range(r_max) for i in range(n_candidate_stops)
for j in range(n_candidate_stops)] +
[(vars.zzxf[r][i + 1][j + 1], costs[i + 1][j + 1])
for r in range(r_max) for i in range(n_candidate_stops)
for j in range(n_candidate_stops)]), 'L', n_buses, 0, 0))
for r in range(r_max):
cns.append('C9(%d)' % r)
cs.append((
linex(
[(vars.x[r][i + 1][j + 1], costs[i + 1][j + 1])
for i in range(n_candidate_stops)
for j in range(n_candidate_stops) if i != j] +
# [(vars.cxx[r][i][j], costs[i+1][j+1])
# for i in range(n_candidate_stops)
# for j in range(n_candidate_stops)
# if i != j] +
[(vars.T[r], -1)]),
'E',
0,
0))
idx = 0
for p in range(len(demands)):
for o in range(len(candidate_stops)):
if walktime(candidate_stops[o], demand_centroids[p]) >= 5:
for q in range(len(demands)):
for d in range(len(candidate_stops)):
cns.append('FILTER(%d)' % idx)
cs.append(
(linex([(vars.s[p][q][o][d], 1)]), 'E', 0, 0))
idx += 1
for q in range(len(demands)):
for d in range(len(candidate_stops)):
if walktime(candidate_stops[d], demand_centroids[q]) >= 5:
for p in range(len(demands)):
for o in range(len(candidate_stops)):
cns.append('FILTER(%d)' % idx)
cs.append(
(linex([(vars.s[p][q][o][d], 1)]), 'E', 0, 0))
idx += 1
idx = 0
for p in range(len(demands)):
for q in range(len(demands)):
if demands[p][q] != 0:
cns.append('S2(%d)' % idx)
cs.append((linex([(vars.s[p][q][c1][c2], 1)
for c1 in range(n_candidate_stops)
for c2 in range(n_candidate_stops)]),
'E', 1, 0))
idx += 1
else:
for c1 in range(n_candidate_stops):
for c2 in range(n_candidate_stops):
cns.append('S2N(%d)' % idx)
cs.append((linex([(vars.s[p][q][c1][c2], 1)]), 'E',
0, 0))
idx += 1
for c in range(n_candidate_stops):
cns.append('S2N2(%d)' % idx)
cs.append((linex([(vars.s[p][q][c][c], 1)]), 'E', 0, 0))
idx += 1
# Sub tour elimination
idx = 0
for r in range(r_max):
for i in range(n_candidate_stops):
for j in range(n_candidate_stops):
if i != j:
cns.append('SUB(%d)' % idx)
cs.append((linex([
(vars.q[r][i], 1), (vars.q[r][j], -1),
(vars.x[r][i + 1][j + 1], 1000)
]), 'L', 999, 0))
idx += 1
idx = 0
for r in range(r_max):
for c2 in range(n_candidate_stops):
cns.append("DEL(%d)" % idx)
cs.append(
(linex([(vars.delta[r][c2], 1)] +
[(vars.x[r][i + 1][c2 + 1], -1)
for i in range(n_candidate_stops)]), 'E', 0, 0))
idx += 1
idx = 0
for r in range(r_max):
for c2 in range(n_candidate_stops):
cns.append("DELF1(%d)" % idx)
cs.append((linex([(vars.deltaf[r][c2], 1),
(vars.f[r], -1)]), 'L', 0, 0))
cns.append("DELF2(%d)" % idx)
cs.append((linex([(vars.deltaf[r][c2], 1),
(vars.delta[r][c2], -60)]), 'L', 0, 0))
cns.append("DELF3(%d)" % idx)
cs.append((linex([(vars.deltaf[r][c2], 1), (vars.f[r], -1),
(vars.delta[r][c2], -60)]), 'G', -60, 0))
idx += 1
idx = 0
for c in range(n_candidate_stops):
cns.append("CAP(%d)" % idx)
cs.append((linex([(vars.deltaf[r][c], 40) for r in range(r_max)] +
[(vars.s[p][q][c1][c], -demands[p][q] / 15.0)
for c1 in range(n_candidate_stops)
for p in range(len(demands))
for q in range(len(demands[0]))]), 'G', 0, 0))
idx += 1
############################################################
# Shortest path constraints
############################################################
idx = 0
for o in range(n_candidate_stops):
for d in range(n_candidate_stops):
if o != d:
cns.append('SP1(%d)' % idx)
cs.append((linex([(vars.t[o][d], 1)] +
[(vars.xs[o][d][r][i][j], -ridecost(i, j))
for r in range(r_max)
for i in range(n_candidate_stops + 1)
for j in range(n_candidate_stops + 2)]),
'E', 0, 0))
else:
cns.append('SP1(%d)' % idx)
cs.append((linex([(vars.t[o][d], 1)]), 'E', 0, 0))
idx += 1
idx = 0
for o in range(n_candidate_stops):
for d in range(n_candidate_stops):
for r in range(r_max):
for i in range(n_candidate_stops + 1):
for j in range(n_candidate_stops + 2):
if o != d:
cns.append('SP2(%d)' % idx)
cs.append((linex([(vars.xs[o][d][r][i][j], 1),
(vars.x[r][i][j], -1)]), 'L',
0, 0))
else:
cns.append('SP2(%d)' % idx)
cs.append((linex([(vars.xs[o][d][r][i][j], 1)
]), 'E', 0, 0))
idx += 1
idx = 0
for o in range(n_candidate_stops):
for d in range(n_candidate_stops):
if o != d:
for i in range(n_candidate_stops):
rhs = 0
if i == o:
rhs = 1
elif i == d:
rhs = -1
cns.append('SP3(%d)' % idx)
cs.append((linex([(vars.xs[o][d][r][i + 1][j], 1)
for r in range(r_max)
for j in range(n_candidate_stops + 2)
if j != (i + 1)] +
[(vars.xs[o][d][r][j][i + 1], -1)
for r in range(r_max)
for j in range(n_candidate_stops + 1)
if j != (i + 1)]), 'E', rhs, 0))
idx += 1
idx = 0
for p in range(len(demands)):
for q in range(len(demands)):
if demands[p][q] != 0:
for o in range(n_candidate_stops):
for d in range(n_candidate_stops):
cns.append('ST1(%d)' % idx)
cs.append((linex([(vars.st[p][q][o][d], 1),
(vars.s[p][q][o][d], -3 * 60)]),
'L', 0, 0))
cns.append('ST2(%d)' % idx)
cs.append((linex([(vars.st[p][q][o][d], 1),
(vars.t[o][d], -1)]), 'L', 0, 0))
cns.append('ST3(%d)' % idx)
cs.append((linex([(vars.st[p][q][o][d], 1),
(vars.t[o][d], -1),
(vars.s[p][q][o][d], -180)]),
'G', -180, 0))
idx += 1
else:
for o in range(n_candidate_stops):
for d in range(n_candidate_stops):
cns.append('STN(%d)' % idx)
cs.append(
(linex([(vars.st[p][q][o][d], 1)]), 'E', 0, 0))
idx += 1
idx = 0
for p in range(len(demands)):
for q in range(len(demands)):
for o in range(len(candidate_stops)):
for d in range(len(candidate_stops)):
cns.append('FILT2-1(%d)' % idx)
cs.append((linex([(vars.s[p][q][o][d], 1)] +
[(vars.delta[r][o], -1)
for r in range(r_max)]), 'L', 0, 0))
idx += 1
cns.append('FILT2-2(%d)' % idx)
cs.append((linex([(vars.s[p][q][o][d], 1)] +
[(vars.delta[r][d], -1)
for r in range(r_max)]), 'L', 0, 0))
idx += 1
cs = zip(*cs)
self._cp.linear_constraints.add(lin_expr=cs[0],
senses=cs[1],
rhs=cs[2],
range_values=cs[3])
self._cp.linear_constraints.set_names(enumerate(cns))
def varyMaxTT(self, maxtt=False, step=5):
self.vars.solutionFound = False
vars = self.vars
cs = []
cns = []
directory_maxtt = os.path.join(self.base_directory, "maxtt")
if not os.path.exists(directory_maxtt):
os.makedirs(directory_maxtt)
nm = self._cp.variables.get_names()
solution_exists = True
if not maxtt:
maxtt = self.sc.plan_horizon / 2
results = []
cons_idx_first = self._cp.linear_constraints.get_num()
# C17
vars.solutionFound = False
for x, g in enumerate(self.sc.groups):
cns.append('C17(%d)' % x)
cs.append((linex([(vars.tg[g.id], 1)] + (
[(vars.to[g.id][k], 1)
for k in range(g.routes[0].n_buses)] if self.
options['max_transfer_waiting_time_constraint'] else [])), 'L',
maxtt))
cs = zip(*cs)
self._cp.linear_constraints.add(lin_expr=cs[0],
senses=cs[1],
rhs=cs[2])
self._cp.linear_constraints.set_names([(i + cons_idx_first, cn)
for i, cn in enumerate(cns)])
while solution_exists and maxtt > 0:
try:
directory = os.path.join(self.base_directory, "maxtt",
"%d" % (maxtt, ))
if not os.path.exists(directory):
os.makedirs(directory)
self.solve(name=os.path.join(
self.base_directory, "maxtt", "%d" %
(maxtt, ), ('%s_%s_%s_maxtt%d' %
(self.timestamp, self.__class__.__name__,
self.sc.name, maxtt))))
results += [
(maxtt, g, self.sc.groups[g].size, tg, sum(to))
for g, (to,
tg) in enumerate(zip(self.vars.to, self.vars.tg))
]
fig_schedule = viz.schedule(self.vars, self.sc)
plot(fig_schedule,
image_filename="schedule",
image="png",
output_type="file",
image_width=888,
image_height=fig_schedule.layout.height)
time.sleep(2)
os.rename(
"/Users/kelvinlee/Downloads/schedule.png",
os.path.join(self.base_directory, "maxtt", str(maxtt),
"schedule.png"))
fig_trip = viz.trip(self.vars, self.sc)
plot(fig_trip,
image_filename="trip",
image="png",
output_type="file",
image_width=888,
image_height=fig_trip.layout.height)
time.sleep(2)
os.rename(
"/Users/kelvinlee/Downloads/trip.png",
os.path.join(self.base_directory, "maxtt", str(maxtt),
"trip.png"))
# fig = viz.comprehensive(self.vars, self.sc)
# plot(fig,
# filename=os.path.join(
# self.base_directory,
# "maxtt",
# ('%s_%s_%s_maxtt%d_viz.html' % (
# self.timestamp, self.__class__.__name__, self.sc.name, maxtt))),
# auto_open=False)
maxtt -= step
self._cp.linear_constraints.set_rhs(
zip(cns, [maxtt] * len(cns)))
except Exception as e:
print e
solution_exists = False
f = open(os.path.join(self.base_directory, "maxtt", "results.csv"),
'w+')
f.write(('Max transfer time, Group, Demand, '
'Transfer time, Waiting time\n'))
for row in results:
f.write(','.join(map(str, list(row))) + '\n')
f.close()
def generate_constraints(self):
routes, groups = (self.sc.routes, self.sc.groups)
omega, theta = (self.sc.omega, self.sc.theta)
plan_horizon, M = (self.sc.plan_horizon, self.sc.M)
vars = self.vars
r = routes[0]
cs = []
cns = []
# C1: First bus
if r.first_bus is not None:
cns.append('C1')
cs.append((linex([(vars.d[0][0], 1)]), 'E', r.first_bus, 0))
# C2: Last bus
if r.last_bus is not None:
cns.append('C2')
cs.append(
(linex([(vars.d[0][r.n_buses - 1], 1)]), 'E', r.last_bus, 0))
# C3: Headway range
for x, (r, s, k) in enumerate(routes.rsk(1)):
if r.hw_max != r.hw_min or (r.hw_max == r.hw_min and s == 0):
cns.append('C3(%d)' % x)
cs.append((linex([(vars.d[s][k + 1], 1), (vars.d[s][k], -1)]),
'R', r.hw_max, r.hw_min - r.hw_max))
# C?: Bus order
for x, (r, s, k) in enumerate(routes.rsk(1)):
cns.append('C?(%d)' % x)
cs.append((linex([(vars.a[s][k], 1),
(vars.a[s][k + 1], -1)]), 'L', 0, 0))
for x, (r, s, k) in enumerate(routes.rsk(1)):
cns.append('C??(%d)' % x)
cs.append((linex([(vars.d[s][k], 1),
(vars.d[s][k + 1], -1)]), 'L', 0, 0))
# C4: Departure-arrival relationship
for x, (r, s, k) in enumerate(routes.rsk()):
cns.append('C4(%d)' % x)
cs.append((linex([(vars.d[s][k], 1), (vars.a[s][k], -1),
(vars.h[s][k], -1)]), 'E', 0, 0))
for x, (r, s, k) in enumerate(routes.rsk()):
cns.append('Dwell(%d)' % x)
cs.append((
# linex([(vars.h[s][k], 1)]), 'R', r.dw_max, r.dw_min-r.dw_max))
linex([(vars.h[s][k], 1)]),
'E',
1,
0))
# C5:
for x, (r, s, k) in enumerate(routes.rsk(0, 1)):
cns.append('C5(%d)' % x)
cs.append((linex([(vars.a[s + 1][k], 1),
(vars.d[s][k], -1)]), 'E', theta[s], 0))
# C6:
for x, (g, r, _, o, k) in enumerate(groups.gp(1, True, False)):
cns.append('C6(%d)' % x)
cs.append((linex([(vars.d[r.stops.index(o)][k], 1),
(vars.p[g.id][k], -M)]), 'R', g.alpha, -M))
# C72:
for x, (g, r, k) in enumerate(groups.gp(0, True, True)):
cns.append('C72(%d' % x)
cs.append((linex([(vars.pt[g.id][k], 1),
(vars.p[g.id][k], -1)]), 'L', 0, 0))
# At least one pt = 1
for x, (g, r) in enumerate(groups.gp(0, False)):
cns.append('CTest(%d' % x)
cs.append((linex([(pt, 1) for pt in vars.pt[g.id]]), 'E', 1, 0))
# # C7:
# for x, (g, r, k) in enumerate(groups.gp(0, True)):
# cns.append('C7(%d)' % x)
# cs.append((
# linex([(vars.p[g.id][k], 1),
# (vars.pt[g.id][k], -M)] +
# ([(vars.p[g.id][k-1], -1)]
# if k != 0 else [])),
# 'R', 0, M))
# C8:
for x, (r, s, k) in enumerate(routes.rsk()):
cns.append('C8(%d)' % x)
cs.append(
(linex([(vars.c[s][k], 1)] +
([(vars.c[s - 1][k], -1)] if s != 0 else []) +
([(vars.pt[g.id][k], -g.size)
for g in groups if g.origin == r.stops[s]]) +
([(vars.pt[g.id][k], g.size)
for g in groups if g.destination == r.stops[s]])),
'E', 0, 0))
# C9:
for x, (r, s, k) in enumerate(routes.rsk(0)):
cns.append('C9(%d)' % x)
cs.append((linex([(vars.c[s][k], 1)]), 'L', 50, 0))
for x, (g, r, k) in enumerate(groups.gp(0, True, True)):
cns.append('C10_2(%d' % x)
cs.append((linex([(vars.pta[g.id][k], 1),
(vars.pt[g.id][k], -M)]), 'L', 0, 0))
for x, (g, r, _, d, k) in enumerate(groups.gp(0, True, False)):
cns.append('C10_3(%d' % x)
cs.append((linex([(vars.pta[g.id][k], 1),
(vars.a[r.stops.index(d)][k], -1)]), 'L', 0, 0))
for x, (g, r, _, d, k) in enumerate(groups.gp(0, True, False)):
cns.append('C10_4(%d' % x)
cs.append((linex([(vars.pta[g.id][k], 1),
(vars.a[r.stops.index(d)][k], -1),
(vars.pt[g.id][k], -M)]), 'G', -M, 0))
# for x, (g, r, o, _, k) in enumerate(groups.gp(0, True, False)):
# cns.append('C12_1(%d)' % x)
# cs.append((
# linex([(vars.ptd[g.id][k], 1),
# (vars.pt[g.id][k], -M)]),
# 'L', 0, 0))
# for x, (g, r, o, _, k) in enumerate(groups.gp(0, True, False)):
# cns.append('C12_2(%d)' % x)
# cs.append((
# linex([(vars.ptd[g.id][k], 1),
# (vars.d[r.stops.index(o)][k], -1)]),
# 'L', 0, 0))
# for x, (g, r, o, _, k) in enumerate(groups.gp(0, True, False)):
# cns.append('C12_3(%d)' % x)
# cs.append((
# linex([(vars.ptd[g.id][k], 1),
# (vars.d[r.stops.index(o)][k], -1),
# (vars.pt[g.id][k], -M)]),
# 'G', -M, 0))
# for x, (g, r) in enumerate(groups.gp(0, False)):
# cns.append('C12_4(%d' % x)
# cs.append((
# linex([(ptd, 1) for ptd in vars.ptd[g.id]] +
# [(vars.w[g.id], -1)]),
# 'E', g.alpha, 0))
for x, (g, r) in enumerate(groups.gp(0, False)):
cns.append('CLast(%d' % x)
cs.append(
(linex([(pta, 1)
for pta in vars.pta[g.id]] + [(vars.t[g.id], -1)]),
'E', g.alpha, 0))
for x, (g1, g2) in enumerate(combinations(groups, 2)):
cns.append('CComb(%d)' % x)
cs.append((linex([(vars.pt[g1.id][k], k)
for k in range(g1.routes[0].n_buses)] +
[(vars.pt[g2.id][k], -k)
for k in range(g2.routes[0].n_buses)]),
('L' if g1.alpha < g2.alpha else 'G'), 0, 0))
cs = zip(*cs)
self._cp.linear_constraints.add(lin_expr=cs[0],
senses=cs[1],
rhs=cs[2],
range_values=cs[3])
self._cp.linear_constraints.set_names(enumerate(cns))
def generate_constraints(self):
routes, groups = (self.sc.routes, self.sc.groups)
omega, theta = (self.sc.omega, self.sc.theta)
plan_horizon, M = (self.sc.plan_horizon, self.sc.M)
vars = self.vars
cs = []
cns = []
# C2: First bus
for r in routes:
if r.first_bus is not None:
cns.append('C2(%d)' % r.id)
cs.append(
(linex([(vars.d[r.id][0][0], 1)]), 'E', r.first_bus, 0))
# C3: Last bus
for r in routes:
if r.last_bus is not None:
cns.append('C3(%d)' % r.id)
cs.append((linex([(vars.d[r.id][0][r.n_buses - 1], 1)]), 'E',
r.last_bus, 0))
# C4: Headway range
for x, (r, s, k) in enumerate(routes.rsk(1)):
if r.hw_max != r.hw_min or (r.hw_max == r.hw_min and s == 0):
cns.append('C4(%d)' % x)
cs.append((linex([(vars.d[r.id][s][k + 1], 1),
(vars.d[r.id][s][k], -1)]), 'R', r.hw_max,
r.hw_min - r.hw_max))
# C5: Bus order
for x, (r, s, k) in enumerate(routes.rsk(1)):
cns.append('C5(%d)' % x)
cs.append((linex([(vars.d[r.id][s][k], 1),
(vars.d[r.id][s][k + 1], -1)]), 'L', 0, 0))
# C6: Departure-arrival relationship
for x, (r, s, k) in enumerate(routes.rsk()):
cns.append('C6(%d)' % x)
cs.append(
(linex([(vars.d[r.id][s][k], 1), (vars.a[r.id][s][k], -1),
(vars.h[r.id][s][k], -1)]), 'E', 0, 0))
# C7: Dwell time range
for x, (r, s, k) in enumerate(routes.rsk()):
cns.append('C7(%d)' % x)
cs.append((linex([(vars.h[r.id][s][k], 1)]), 'R', r.dw_max,
r.dw_min - r.dw_max))
# C8/11
for x, (r, s, k) in enumerate(routes.rsk(0, 1)):
cns.append('C8(%d)' % x)
cs.append((
linex([(vars.a[r.id][s + 1][k], 1), (vars.d[r.id][s][k], -1)] +
[(vars.db[r.id][s][k][n], -theta[r.id][s][n])
for n in range(plan_horizon)]), 'E', 0, 0))
# C9
for x, (r, s, k) in enumerate(routes.rsk()):
cns.append('C9(%d)' % x)
cs.append((linex([(vars.d[r.id][s][k], 1)] +
[(vars.db[r.id][s][k][n], -n)
for n in range(plan_horizon)]), 'E', 0, 0))
# C10
for x, (r, s, k) in enumerate(routes.rsk()):
cns.append('C10(%d)' % x)
cs.append((linex([(vars.db[r.id][s][k][n], 1)
for n in range(plan_horizon)]), 'E', 1, 0))
# C12
for x, (r, rp, s, k, kp) in enumerate(routes.tf()):
cns.append('C12(%d)' % x)
cs.append((linex([(vars.d[rp.id][rp.stops.index(s)][kp], 1),
(vars.a[r.id][r.stops.index(s)][k], -1),
(vars.c[r.id][rp.id][s][k][kp], -M)]), 'R',
omega[r.id][s][rp.id], -M))
# C13
for x, (r, rp, s, k, kp) in enumerate(routes.tf()):
cns.append('C13(%d)' % x)
cs.append((linex([(vars.d[rp.id][rp.stops.index(s)][kp], 1),
(vars.a[r.id][r.stops.index(s)][k], -1),
(vars.t[r.id][rp.id][s][k][kp], -1)]), 'L',
omega[r.id][s][rp.id], 0))
# C14
for x, (g, r, _, o, k) in enumerate(groups.gp(1, True, False)):
cns.append('C14(%d)' % x)
cs.append(
(linex([(vars.d[r.id][r.stops.index(o)][k], 1),
(vars.co[g.id][k], -M)]), 'R', g.alpha, -plan_horizon))
# C14b
for x, (g, r, _, o, k) in enumerate(groups.gp(1, True, False)):
cns.append('C14b(%d)' % x)
cs.append((linex([(vars.d[r.id][r.stops.index(o)][k],
1), (vars.to[g.id][k], -1)] +
([(vars.co[g.id][k - 1], -M)] if k != 0 else [])),
'L', g.alpha, 0))
# C15a
for x, (g, r, rp, s, k, kp) in enumerate(groups.gp(1, True, True)):
cns.append('C15(%d)' % x)
cs.append((linex(
[(vars.z[g.id][k][kp], 1), (vars.c[r.id][rp.id][s][k][kp],
-1), (vars.co[g.id][k], -1)] +
([(vars.c[r.id][rp.id][s][k][kp - 1], 1)] if kp != 0 else []) +
([(vars.co[g.id][k - 1], 1)] if k != 0 else [])), 'G', -1, 0))
# C15b
for x, (g, r, rp, s, k, kp) in enumerate(groups.gp(1, True, True)):
cns.append('C15b(%d)' % x)
cs.append((linex([(vars.z[g.id][k][kp],
1), (vars.c[r.id][rp.id][s][k][kp], -1)] +
([(vars.c[r.id][rp.id][s][k][kp - 1],
1)] if kp != 0 else [])), 'L', 0, 0))
# C15c
for x, (g, _, _, _, k, kp) in enumerate(groups.gp(1, True, True)):
cns.append('C15c(%d)' % x)
cs.append(
(linex([(vars.z[g.id][k][kp], 1), (vars.co[g.id][k], -1)] +
([(vars.co[g.id][k - 1], 1)] if k != 0 else [])), 'L',
0, 0))
# C15h
for x, (g, r, rp, s, k, kp) in enumerate(groups.gp(1, True, True)):
cns.append('C15h(%d)' % x)
cs.append((linex([(vars.p[r.id][rp.id][s][k][kp][g.id], 1),
(vars.z[g.id][k][kp], -M)]), 'L', 0, 0))
# C15i
for x, (g, r, rp, s, k, kp) in enumerate(groups.gp(1, True, True)):
cns.append('C15i(%d)' % x)
cs.append((linex([(vars.p[r.id][rp.id][s][k][kp][g.id], 1),
(vars.z[g.id][k][kp], -1)]), 'G', 0, 0))
# C3.5
print(groups.gp())
for x, (g, r, rp, s) in enumerate(groups.gp()):
cns.append('C3.5(%d)' % x)
cs.append((linex([(vars.p[r.id][rp.id][s][k][kp][g.id], 1)
for k in range(r.n_buses)
for kp in range(rp.n_buses)]), 'E', 1, 0))
# C3.6
for x, (g, r1, rp1, s1, r2, rp2, s2) in enumerate(groups.gp(2)):
cns.append('C3.6(%d)' % x)
cs.append(
(linex([(vars.p[r1.id][rp1.id][s1][k1][kp1][g.id], kp1 + 1)
for k1 in range(r1.n_buses)
for kp1 in range(rp1.n_buses)] +
[(vars.p[r2.id][rp2.id][s2][k2][kp2][g.id], -k2 - 1)
for k2 in range(r2.n_buses)
for kp2 in range(rp2.n_buses)]), 'E', 0, 0))
# C3.8
for x, (g, r, rp, s, k, kp) in enumerate(groups.gp(False, True, True)):
cns.append('C3.8(%d)' % x)
cs.append(
(linex([(vars.p[r.id][rp.id][s][k][kp][g.id], 1),
(vars.c[r.id][rp.id][s][k][kp], -1)]), 'L', 0, 0))
# C3.7a
for x, (g, r, rp, s, k, kp) in enumerate(groups.gp(False, True, True)):
cns.append('C3.7a(%d)' % x)
cs.append((linex([(vars.zt[r.id][rp.id][s][k][kp][g.id], 1),
(vars.p[r.id][rp.id][s][k][kp][g.id],
-plan_horizon)]), 'L', 0, 0))
# C3.7c
for x, (g, r, rp, s, k, kp) in enumerate(groups.gp(False, True, True)):
cns.append('C3.7c(%d)' % x)
cs.append((linex([(vars.zt[r.id][rp.id][s][k][kp][g.id], 1),
(vars.t[r.id][rp.id][s][k][kp], -1),
(vars.p[r.id][rp.id][s][k][kp][g.id],
-plan_horizon)]), 'G', -plan_horizon, 0))
# C3.7d
for x, (g, r, rp, s, k, kp) in enumerate(groups.gp(False, True, True)):
cns.append('C3.7d(%d)' % x)
cs.append(
(linex([(vars.zt[r.id][rp.id][s][k][kp][g.id], 1),
(vars.t[r.id][rp.id][s][k][kp], -1)]), 'L', 0, 0))
# C3.7e
for x, (g, trs) in enumerate(groups.gp(-1)):
cns.append('C3.7e(%d)' % x)
cs.append((linex([(vars.tg[g.id], 1)] +
[(vars.zt[r.id][rp.id][s][k][kp][g.id], -1)
for r, rp, s in trs for k in range(r.n_buses)
for kp in range(rp.n_buses)]), 'G', 0, 0))
cs = zip(*cs)
self._cp.linear_constraints.add(lin_expr=cs[0],
senses=cs[1],
rhs=cs[2],
range_values=cs[3])
self._cp.linear_constraints.set_names(enumerate(cns))
def generate_constraints(self, c=30, n=20):
routes = [1,2,3]
M = 60
route_length = [2,3,4]
max_capacity = c
n_buses = n
# demand = [[[9,9]], [[10,10]], [[24,25]]]
# demand = [[[138,141,155,157,146,162,180,174,165,182,194,167,100,116,144,182,202,207,205,209,221,218,211,221,224,208,203,201,181,174,170,156]], [[152,158,164,192,185,179,156,167,166,177,160,179,90,99,91,96,59,51,45,47,46,48,35,43,39,45,42,35,42,28,30,31]], [[366,386,395,436,443,432,436,430,426,442,480,516,366,323,392,496,516,485,491,491,499,464,479,424,418,417,388,386,360,357,350,355], [2,8,10,2,2,5,6,3,3,2,6,2,2,5,7,4,7,2,4,6,8,7,8,11,15,15,16,13,12,11,14,13]]]
demand = [[[591,662,708,542,823,871,836,681]], [[666,687,682,376,202,172,161,131]], [[1583,1741,1864,1577,1983,1866,1609,1422], [22,16,13,18,19,34,59,50]]]
vars = self.vars
cs = []
cns = []
cns.append('last')
cs.append((
linex([(vars.sw[0], -1)] +
[(vars.w[r][i], sum(map(lambda x: x[i], demand[r])))
for r in range(len(routes))
for i in range(len(demand[0][0]))]),
'E', 0, 0))
for r in range(len(routes)):
for i in range(len(demand[0][0])):
cns.append('C0(%d)' % (r*len(demand[0][0]) + i,))
cs.append((
linex([(vars.w[r][i], 1),
(vars.delta[r][i], -0.5)]),
'E', 0, 0))
for r in range(len(routes)):
for i in range(len(demand[0][0])):
cns.append('C1(%d)' % (r*len(demand[0][0]) + i,))
cs.append((linex([(vars.delta[r][i], 1)]), 'L', max_capacity / sum(map(lambda x: x[i]/60.0, demand[r])), 0))
for r in range(len(routes)):
for i in range(len(demand[0][0])):
cns.append('C22(%d)' % (r*len(demand[0][0]) + i,))
cs.append((
linex([(vars.ndelta[i][r][k], k+1)
for k in range(n_buses)]),
'G', route_length[r]*2, 0))
for r in range(len(routes)):
for i in range(len(demand[0][0])):
cns.append('C2(%d)' % (r*len(demand[0][0]) + i,))
cs.append((
linex([(vars.n[i][r][k], 1)
for k in range(n_buses)]),
'E', 1, 0))
if n_buses < len(vars.n[i][r]):
for r in range(len(routes)):
for i in range(len(demand[0][0])):
for k in range(len(vars.n[i][r]) - n_buses):
cns.append('C?(%d)' % k)
cs.append((linex([(vars.n[i][r][k+n_buses], 1)]), 'E', 0, 0))
for r in range(len(routes)):
for i in range(len(demand[0][0])):
for k in range(len(vars.n[i][r])):
# for k in range(n_buses):
cns.append('C3(%d)' % (r*len(demand[0][0]) + i,))
cs.append((
linex([(vars.ndelta[i][r][k], 1),
(vars.n[i][r][k], -M)]),
'L', 0, 0))
cns.append('C4(%d)' % (r*len(demand[0][0]) + i,))
cs.append((
linex([(vars.ndelta[i][r][k], 1),
(vars.delta[r][i], -1)]),
'L', 0, 0))
cns.append('C5(%d)' % (r*len(demand[0][0]) + i,))
cs.append((
linex([(vars.ndelta[i][r][k], 1),
(vars.n[i][r][k], -M),
(vars.delta[r][i], -1)]),
'G', -M, 0))
for i in range(len(demand[0][0])):
cns.append('C6(%d)' % i)
cs.append((
linex([(vars.n[i][r][k], k+1)
for k in range(n_buses)
for r in range(len(routes))]),
'L', n_buses, 0))
cs = zip(*cs)
self._cp.linear_constraints.add(
lin_expr=cs[0],
senses=cs[1],
rhs=cs[2],
range_values=cs[3])
self._cp.linear_constraints.set_names(enumerate(cns))
def generate_constraints(self):
r_max = 1
max_cap = 30
n_buses = 5
candidate_stops = [[55.785470, 12.523041], [55.786239, 12.522234],
[55.786541, 12.526961], [55.789277, 12.523946],
[55.786431, 12.521428], [55.785898, 12.520618],
[55.785138, 12.520728], [55.784637, 12.520459],
[55.782099, 12.514132], [55.782460, 12.519244],
[55.782226, 12.520035]]
demand_centroids = [[55.786344, 12.524316], [55.786344, 12.524316],
[55.786872, 12.5259], [55.78944, 12.525146],
[55.786745, 12.520469], [55.787486, 12.518472],
[55.787911, 12.518621], [55.785438, 12.519377],
[55.785154, 12.520434], [55.785241, 12.518699],
[55.782330, 12.513640], [55.781929, 12.521461]]
demands = [[
0.0, 59.04, 20.34, 17.0, 0.0, 15.87, 8.27, 2.94, 23.92, 34.73,
1.91, 15.26
],
[
47.32, 0.0, 7.31, 8.7, 0.0, 8.66, 5.07, 1.23, 11.37,
14.67, 0.97, 6.27
],
[
14.09, 7.53, 0.0, 4.51, 0.0, 2.02, 1.23, 0.39, 3.78,
3.75, 0.24, 1.67
],
[
14.64, 8.1, 4.75, 0.0, 0.0, 1.13, 0.81, 0.4, 2.07, 2.76,
0.27, 1.53
],
[
0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0,
0.0
],
[
14.1, 8.61, 2.06, 1.38, 0.0, 0.0, 20.35, 1.28, 6.48,
16.1, 0.25, 1.72
],
[
7.06, 4.96, 1.13, 1.09, 0.0, 20.74, 0.0, 0.39, 2.31,
6.38, 0.15, 0.86
],
[
2.3, 1.03, 0.34, 0.23, 0.0, 0.99, 0.22, 0.0, 5.55, 5.08,
0.07, 0.37
],
[
19.67, 11.18, 3.53, 1.91, 0.0, 7.05, 2.24, 6.45, 0.0,
28.64, 0.82, 4.56
],
[
31.91, 14.58, 3.91, 2.84, 0.0, 17.88, 6.71, 7.28,
28.09, 0.0, 1.15, 7.68
],
[
1.72, 0.94, 0.23, 0.3, 0.0, 0.22, 0.12, 0.1, 0.83, 1.13,
0.0, 1.2
],
[
13.15, 5.68, 1.69, 1.78, 0.0, 1.63, 0.82, 0.4, 4.24,
7.15, 1.11, 0.0
]]
# demands = [[0, 100, 3, 36, 0, 14, 13, 19, 42, 55, 5, 20], [190, 0, 2, 25, 0, 18, 6, 8, 40, 37, 5, 4], [13, 10, 0, 3, 0, 3, 0, 1, 5, 3, 0, 0], [137, 1, 0, 0, 0, 10, 4, 1, 17, 34, 1, 1], [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0], [47, 9, 2, 4, 0, 0, 63, 0, 0, 6, 2, 0], [46, 4, 0, 0, 0, 173, 0, 0, 0, 2, 2, 1], [61, 2, 0, 1, 0, 1, 0, 0, 78, 116, 0, 0], [116, 13, 1, 5, 0, 8, 0, 52, 0, 84, 16, 8], [167, 6, 0, 6, 0, 33, 13, 52, 230, 0, 23, 6], [1, 1, 0, 0, 0, 1, 0, 0, 7, 1, 0, 1], [97, 55, 0, 6, 0, 3, 2, 2, 47, 39, 9, 0]]
n_candidate_stops = len(candidate_stops)
n_demands = len(demands)
costs = []
for i in range(len(demands) + n_candidate_stops):
costs.append([])
for j in range(len(demands) + n_candidate_stops):
org = demand_centroids[i] if i < len(
demands) else candidate_stops[i - len(demands)]
des = demand_centroids[j] if j < len(
demands) else candidate_stops[j - len(demands)]
if i < len(demands) or j < len(demands):
costs[i].append(walktime(org, des))
else:
costs[i].append(
ridetime(i - len(demands), j - len(demands)))
dd = enumerate([(x, y) for x in range(n_demands)
for y in range(n_demands)])
dd_neq = enumerate([(x, y) for x in range(n_demands)
for y in range(n_demands) if x != y])
dd_eq = enumerate([(x, x) for x in range(n_demands)])
ddn = enumerate([(x, y, p) for x in range(n_demands)
for y in range(n_demands)
for p in range(n_demands + n_candidate_stops)])
vars = self.vars
cs = []
cns = []
for idx, (o, d) in dd:
if o != d:
cns.append('A1(%d)' % idx)
cs.append((linex([(vars.x[o][d][o][j], 1)
for j in range(n_demands + n_candidate_stops)
if j != o]), 'E', 1, 0))
else:
cns.append('A1(%d)' % idx)
cs.append((linex([(vars.x[o][d][o][j], 1)
for j in range(n_demands + n_candidate_stops)
]), 'E', 0, 0))
for idx, (o, d) in dd:
if o != d:
cns.append('A2(%d)' % idx)
cs.append((linex([(vars.x[o][d][i][d], 1)
for i in range(n_demands + n_candidate_stops)
if i != d]), 'E', 1, 0))
else:
cns.append('A2(%d)' % idx)
cs.append((linex([(vars.x[o][d][i][d], 1)
for i in range(n_demands + n_candidate_stops)
]), 'E', 0, 0))
# idx = 0
# for o in range(n_demands):
# for d in range(n_demands):
# if o != d:
# for i in range(n_demands + n_candidate_stops):
for idx, (o, d, i) in ddn:
if o != d:
rhs = 0
if i == o:
rhs = 1
elif i == d:
rhs = -1
cns.append('A3(%d)' % idx)
cs.append((linex(
[(vars.x[o][d][i][j], 1)
for j in range(n_demands + n_candidate_stops) if i != j] +
[(vars.x[o][d][j][i], -1)
for j in range(n_demands + n_candidate_stops) if i != j]),
'E', rhs, 0))
# idx += 1
idx = 0
for o in range(n_demands):
for d in range(n_demands):
for j in range(n_candidate_stops):
cns.append('A4(%d)' % idx)
cs.append((linex([(vars.x[o][d][i][j + n_demands], 1)
for i in range(n_demands)] +
[(vars.x[o][d][j + n_demands][i], 1)
for i in range(n_demands)]), 'L', 1, 0))
idx += 1
idx = 0
for o in range(n_demands):
for d in range(n_demands):
if o != d:
cns.append('A5(%d)' % idx)
cs.append(
(linex([(vars.w[o][d], 1)] +
[(vars.x[o][d][i][j], -costs[i][j])
for i in range(n_demands + n_candidate_stops)
for j in range(n_demands + n_candidate_stops)
if i == o or j == d]), 'E', 0, 0))
idx += 1
idx = 0
for o in range(n_demands):
for d in range(n_demands):
if o != d:
cns.append('A6(%d)' % idx)
cs.append(
(linex([(vars.r[o][d], 1)] +
[(vars.x[o][d][i][j], -costs[i][j])
for i in range(n_demands + n_candidate_stops)
for j in range(n_demands + n_candidate_stops)
if i != o and j != d]), 'E', 0, 0))
idx += 1
idx = 0
for o in range(n_demands):
for d in range(n_demands):
# cns.append('TIMEGUARANTEE(%d)' % idx)
# cs.append((
# linex([(vars.r[o][d], 1),
# (vars.w[o][d], 1)]),
# 'L', 20, 0))
# for j in range(n_demands):
# cns.append('MAXWALK(%d)' % idx)
# cs.append((linex([(vars.x[o][d][o][j], 1)]), 'E', 0, 0))
# idx += 1
# for i in range(n_demands):
# cns.append('MAXWALK(%d)' % idx)
# cs.append((linex([(vars.x[o][d][i][d], 1)]), 'E', 0, 0))
# idx += 1
for j in range(n_candidate_stops):
if costs[o][j + n_demands] > 5:
cns.append('A7(%d)' % idx)
cs.append((linex([(vars.x[o][d][o][j + n_demands], 1)
]), 'E', 0, 0))
idx += 1
for i in range(n_candidate_stops):
if costs[d][i + n_demands] > 5:
cns.append('A8(%d)' % idx)
cs.append((linex([(vars.x[o][d][i + n_demands][d], 1)
]), 'E', 0, 0))
idx += 1
for i in range(n_demands + n_candidate_stops):
for j in range(n_demands):
if j != d:
cns.append('A9(%d)' % idx)
cs.append(
(linex([(vars.x[o][d][i][j], 1)]), 'E', 0, 0))
idx += 1
for i in range(n_demands):
for j in range(n_demands + n_candidate_stops):
if i != o:
cns.append('A10(%d)' % idx)
cs.append(
(linex([(vars.x[o][d][i][j], 1)]), 'E', 0, 0))
idx += 1
idx = 0
for r in range(r_max):
# for i in range(n_candidate_stops):
# for j in range(n_candidate_stops):
# cns.append('B0(%d)' % idx)
# cs.append((
# linex([(vars.rx[r][i][j], 1)] +
# [(vars.x[o][d][i+n_demands][j+n_demands], -1)
# for o in range(len(demands))
# for d in range(len(demands))]),
# 'L', 0, 0))
# idx += 1
# idx = 0
# for i in range(n_candidate_stops):
# for j in range(n_candidate_stops):
# cns.append('B0-1(%d)' % idx)
# cs.append((
# linex([(vars.rx0[r][i][j], 1)] +
# [(vars.x[o][d][i+n_demands][j+n_demands], -1)
# for o in range(len(demands))
# for d in range(len(demands))]),
# 'L', 0, 0))
# idx += 1
idx = 0
for r in range(r_max):
for i in range(n_candidate_stops):
for j in range(n_candidate_stops):
cns.append('B0-2(%d)' % idx)
cs.append((linex([(vars.rx0[r][i][j], 1),
(vars.c[r], 1)]), 'L', 1, 0))
idx += 1
idx = 0
for i in range(n_candidate_stops):
for j in range(n_candidate_stops):
cns.append('B1(%d)' % idx)
cs.append(
(linex([(vars.rx[r][i][j], 1)
for r in range(r_max)] + [(vars.rx0[r][i][j], 1)
for r in range(r_max)] +
[(vars.x[o][d][i + n_demands][j + n_demands],
-1.0 / n_demands / n_demands)
for o in range(n_demands)
for d in range(n_demands)]), 'G', 0, 0))
idx += 1
idx = 0
for r in range(r_max):
cns.append('B2(%d)' % idx)
cs.append(
(linex([(vars.rx[r][n_candidate_stops][j], 1)
for j in range(n_candidate_stops + 1)]), 'E', 1, 0))
cns.append('B3(%d)' % idx)
cs.append((linex([(vars.rx[r][i][n_candidate_stops], 1)
for i in range(n_candidate_stops + 1)] +
[(vars.rx[r][i][n_candidate_stops + 1], 1)
for i in range(n_candidate_stops)]), 'E', 1, 0))
idx += 1
idx = 0
for r in range(r_max):
for i in range(n_candidate_stops):
cns.append('B4(%d)' % idx)
cs.append(
(linex([(vars.rx[r][i][j], 1)
for j in range(n_candidate_stops + 2) if i != j] +
[(vars.rx[r][j][i], -1)
for j in range(n_candidate_stops + 2) if i != j]),
'E', 0, 0))
idx += 1
idx = 0
for r in range(r_max):
for i in range(n_candidate_stops):
cns.append('B5(%d)' % idx)
cs.append((linex([
(vars.rx[r][j][i], 1) for j in range(n_candidate_stops + 2)
]), 'L', 1, 0))
cns.append('B6(%d)' % idx)
cs.append((linex([
(vars.rx[r][i][j], 1) for j in range(n_candidate_stops + 2)
]), 'L', 1, 0))
idx += 1
idx = 0
for r in range(r_max):
for i in range(n_candidate_stops):
cns.append('B7(%d)' % idx)
cs.append((linex([(vars.rx[r][i][i], 1)]), 'E', 0, 0))
idx += 1
for r in range(r_max):
cns.append('R9(%d)' % r)
cs.append((linex([
(vars.rx[r][n_candidate_stops][n_candidate_stops + 1], 1)
]), 'E', 0, 0))
idx = 0
for r in range(r_max):
for i in range(1, n_candidate_stops):
cns.append('B8-1(%d)' % idx)
cs.append(
(linex([(vars.rx[r][i][n_candidate_stops], 1)] +
[(vars.rx[r][k][n_candidate_stops], 1)
for k in range(n_candidate_stops + 1) if k < i]),
'L', 1, 0))
cns.append('B8-2(%d)' % idx)
cs.append((linex([(vars.rx[r][i][n_candidate_stops], 1)] +
[(vars.rx[r][j][i], -1)
for j in range(n_candidate_stops + 1)]), 'L',
0, 0))
cns.append('B8-3(%d)' % idx)
cs.append((linex([(vars.rx[r][i][n_candidate_stops], 1),
(vars.c[r], 1)]), 'L', 1, 0))
cns.append('B8-4(%d)' % idx)
cs.append(
(linex([(vars.rx[r][i][n_candidate_stops],
1), (vars.c[r], 1)] +
[(vars.rx[r][k][n_candidate_stops], 1)
for k in range(n_candidate_stops + 1) if k < i] +
[(vars.rx[r][j][i], -1)
for j in range(n_candidate_stops + 1)]), 'G', 0,
0))
idx += 1
idx = 0
for r in range(r_max):
for i in range(n_candidate_stops):
for j in range(n_candidate_stops):
if i != j:
cns.append('B9(%d)' % idx)
cs.append(
(linex([(vars.rq[r][i], 1), (vars.rq[r][j], -1),
(vars.rx[r][i][j], 1000)]), 'L', 999, 0))
idx += 1
idx = 0
for r in range(r_max):
for i in range(n_candidate_stops):
for j in range(n_candidate_stops):
cns.append('B10-1(%d)' % idx)
cs.append((linex([(vars.rx0[r][i][j], 1),
(vars.rx[r][i][n_candidate_stops], -1),
(vars.rx[r][n_candidate_stops][j], -1)]),
'G', -1, 0))
cns.append('B10-2(%d)' % idx)
cs.append((linex([(vars.rx0[r][i][j], 1),
(vars.rx[r][i][n_candidate_stops], -1)]),
'L', 0, 0))
cns.append('B10-3(%d)' % idx)
cs.append((linex([(vars.rx0[r][i][j], 1),
(vars.rx[r][n_candidate_stops][j], -1)]),
'L', 0, 0))
idx += 1
for r in range(r_max):
cns.append('C1(%d)' % r)
cs.append((linex([(vars.c[r], 1)] +
[(vars.rx[r][i][n_candidate_stops + 1], -1)
for i in range(n_candidate_stops)]), 'E', 0, 0))
idx = 0
for i in range(n_candidate_stops):
for j in range(n_candidate_stops):
if i != j:
cns.append('C3(%d)' % idx)
cs.append(
(linex([(vars.x[o][d][i + n_demands][j + n_demands],
demands[o][d]) for o in range(n_demands)
for d in range(n_demands)] +
[(vars.z1[r][i][j], -max_cap)
for r in range(r_max)]), 'L', 0, 0))
idx += 1
idx = 0
for r in range(r_max):
for i in range(n_candidate_stops):
for j in range(n_candidate_stops):
if i != j:
cns.append('C3-2(%d)' % idx)
cs.append(
(linex([(vars.z1[r][i][j], 1),
(vars.rx[r][i][j], -60)]), 'L', 0, 0))
cns.append('C3-3(%d)' % idx)
cs.append((linex([(vars.z1[r][i][j], 1),
(vars.f[r], -1)]), 'L', 0, 0))
cns.append('C3-3(%d)' % idx)
cs.append(
(linex([(vars.z1[r][i][j], 1), (vars.f[r], -1),
(vars.rx[r][i][j], -60)]), 'G', -60, 0))
idx += 1
idx = 0
for r in range(r_max):
for i in range(n_candidate_stops):
for j in range(n_candidate_stops):
if i != j:
cns.append('C5-2(%d)' % idx)
cs.append(
(linex([(vars.z2[r][i][j], 1),
(vars.rx[r][i][j], -60)]), 'L', 0, 0))
cns.append('C5-3(%d)' % idx)
cs.append((linex([(vars.z2[r][i][j], 1),
(vars.c[r], -60)]), 'L', 0, 0))
cns.append('C5-4(%d)' % idx)
cs.append((linex([(vars.z2[r][i][j], 1),
(vars.f[r], -1)]), 'L', 0, 0))
cns.append('C5-5(%d)' % idx)
cs.append((linex([(vars.z2[r][i][j], 1),
(vars.f[r], -1),
(vars.rx[r][i][j], -60),
(vars.c[r], -60)]), 'G', -120, 0))
idx += 1
cns.append('C5')
cs.append((linex([(vars.z1[r][i][j], 1) for r in range(r_max)
for i in range(n_candidate_stops)
for j in range(n_candidate_stops) if i != j] +
[(vars.z2[r][i][j], 1) for r in range(r_max)
for i in range(n_candidate_stops)
for j in range(n_candidate_stops) if i != j]), 'L',
n_buses, 0))
# idx = 0
# for o in range(n_demands):
# for d in range(n_demands):
# if o != d:
# for i in range(n_demands + n_candidate_stops):
# for j in range(n_demands + n_candidate_stops):
# if i != j:
# cns.append('SUB(%d)' % idx)
# cs.append((
# linex([(vars.q[o][d][i], 1),
# (vars.q[o][d][j], -1),
# (vars.x[o][d][i][j], 1000)]),
# 'L', 999, 0))
# idx += 1
# idx = 0
# for o in range(n_demands):
# for d in range(n_demands):
# if o != d:
# cns.append('C0(%d)' % idx)
# cs.append((
# linex([(vars.x[o][d][i][j], 1)
# for j in range(n_demands)
# for i in range(n_candidate_stops+n_demands)]),
# 'E', 1, 0))
# idx += 1
# idx = 0
# for o in range(n_demands):
# for d in range(n_demands):
# if o != d:
# cns.append('C4(%d)' % idx)
# cs.append((
# linex([(vars.x[o][d][i][j], 1)
# for i in range(n_demands)
# for j in range(n_demands)]),
# 'E', 0, 0))
# idx += 1
# idx = 0
# for o in range(n_demands):
# for d in range(n_demands):
# if o != d:
# cns.append('C5(%d)' % idx)
# cs.append((
# linex([(vars.t[o][d], 1)] +
# [(vars.x[o][d][i][j], -costs[i][j])
# for i in range(n_demands + n_candidate_stops)
# for j in range(n_demands + n_candidate_stops)]),
# 'E', 0, 0))
# idx += 1
# idx = 0
# for i in range(n_candidate_stops):
# for j in range(n_candidate_stops):
# if i != j:
# cns.append('C3(%d)' % idx)
# cs.append((
# linex([(vars.x[o][d][n_demands + i][n_demands + j], -demands[o][d])
# for o in range(n_demands)
# for d in range(n_demands)] +
# [(vars.D[i][j], 1)]),
# 'E', 0, 0))
# idx = 0
# for i in range(n_candidate_stops):
# for j in range(n_candidate_stops):
# if i != j:
# cns.append('C2(%d)' % idx)
# cs.append((
# linex([(vars.D[i][j], 1)] +
# [(vars.f[r], -max_cap)]),
# 'L', 0, 0))
# idx += 1
# idx = 0
# for r in range(r_max):
# cns.append('C4(%d)' % idx)
# cs.append((
# linex([(vars.T[r], 1)] +
# [(vars.rx[r][i][j], -ridetime(i, j))
# for i in range(n_candidate_stops)
# for j in range(n_candidate_stops)
# if i != j]),
# 'E', 0, 0))
# idx += 1
# idx = 0
# for r in range(r_max):
# for i in range(n_candidate_stops-1):
# cns.append('B8(%d)' % i)
# cs.append((
# linex([(vars.rx[r][i][n_candidate_stops], 1),
# (vars.rx[r][i+1][n_candidate_stops], -1)]),
# 'L', 0, 0))
# idx += 1
cs = zip(*cs)
self._cp.linear_constraints.add(lin_expr=cs[0],
senses=cs[1],
rhs=cs[2],
range_values=cs[3])
self._cp.linear_constraints.set_names(enumerate(cns))
def generate_constraints(self, **kwargs):
r_max = kwargs['r_max']
max_cap = kwargs['max_cap']
candidate_stops = kwargs['candidate_stops']
demand_centroids = kwargs['demand_centroids']
demands = kwargs['demands']
n_buses = kwargs['n_buses']
candidate_routes = kwargs['candidate_routes']
is_circular = kwargs['is_circular']
t = kwargs['route_time']
costs = kwargs['costs']
n_candidate_stops = len(candidate_stops)
n_demands = len(demands)
n_candidate_routes = len(candidate_routes)
def it(spec):
lengths = {
'd': n_demands,
's': n_candidate_stops,
'n': n_demands + n_candidate_stops,
'r': r_max,
'c': n_candidate_routes,
'b': n_buses + 1
}
if len(spec) == 1:
return enumerate(range(lengths[spec]))
return enumerate(product(*[range(lengths[x]) for x in spec]))
vars = self.vars
cs = []
cns = []
#///////////////////////////////////////////////////////////////////////
# Origin destination shortest path problem (ODSSP)
#///////////////////////////////////////////////////////////////////////
for idx, (o, d) in it('dd'):
if o != d:
cns.append('A1(%d)' % idx)
cs.append((linex([(vars.x[o][d][o][j], 1)
for j in range(n_demands + n_candidate_stops)
if j != o]), 'E', 1, 0))
else:
cns.append('A1(%d)' % idx)
cs.append((linex([(vars.x[o][d][o][j], 1)
for j in range(n_demands + n_candidate_stops)
]), 'E', 0, 0))
for idx, (o, d) in it('dd'):
if o != d:
cns.append('A2(%d)' % idx)
cs.append((linex([(vars.x[o][d][i][d], 1)
for i in range(n_demands + n_candidate_stops)
if i != d]), 'E', 1, 0))
else:
cns.append('A2(%d)' % idx)
cs.append((linex([(vars.x[o][d][i][d], 1)
for i in range(n_demands + n_candidate_stops)
]), 'E', 0, 0))
for idx, (o, d, i) in it('ddn'):
if o != d:
rhs = 0
if i == o:
rhs = 1
elif i == d:
rhs = -1
cns.append('A3(%d)' % idx)
cs.append((linex(
[(vars.x[o][d][i][j], 1)
for j in range(n_demands + n_candidate_stops) if j != i] +
[(vars.x[o][d][j][i], -1)
for j in range(n_demands + n_candidate_stops) if j != i]),
'E', rhs, 0))
for idx, (o, d, j) in it('dds'):
cns.append('A4(%d)' % idx)
cs.append((linex([(vars.x[o][d][i][j + n_demands], 1)
for i in range(n_demands)] +
[(vars.x[o][d][j + n_demands][i], 1)
for i in range(n_demands)]), 'L', 1, 0))
for idx, (o, d) in it('dd'):
if o != d:
cns.append('A5(%d)' % idx)
cs.append((linex([(vars.w[o][d], 1)] +
[(vars.x[o][d][i][j], -costs[i][j])
for i in range(n_demands + n_candidate_stops)
for j in range(n_demands + n_candidate_stops)
if i == o or j == d]), 'E', 0, 0))
for idx, (o, d) in it('dd'):
if o != d:
cns.append('A6(%d)' % idx)
cs.append(
(linex([(vars.r[o][d], 1)] +
[(vars.x[o][d][i + n_demands][j + n_demands],
-costs[i + n_demands][j + n_demands])
for i in range(n_candidate_stops)
for j in range(n_candidate_stops)]), 'E', 0, 0))
for idx, (o, d, j) in it('dds'):
if costs[o][j + n_demands] > 5:
cns.append('A7(%d)' % idx)
cs.append(
(linex([(vars.x[o][d][o][j + n_demands], 1)]), 'E', 0, 0))
for idx, (o, d, i) in it('dds'):
if costs[i + n_demands][d] > 5:
cns.append('A8(%d)' % idx)
cs.append(
(linex([(vars.x[o][d][i + n_demands][d], 1)]), 'E', 0, 0))
for idx, (o, d, i, j) in it('ddnd'):
if j != d:
cns.append('A9(%d)' % idx)
cs.append((linex([(vars.x[o][d][i][j], 1)]), 'E', 0, 0))
for idx, (o, d, i, j) in it('dddn'):
if i != o:
cns.append('A10(%d)' % idx)
cs.append((linex([(vars.x[o][d][i][j], 1)]), 'E', 0, 0))
for idx, (o, d) in it('dd'):
if o != d:
# if costs[o][d] == 0:
# f = 1
# else:
# f = 1.0/float(costs[o][d])/float(costs[o][d])
f = 0
cns.append(('A11-1(%d)' % idx))
cs.append((linex([(vars.s[o][d], 1), (vars.r[o][d], 1),
(vars.w[o][d], 1 - f), (vars.hod[o][d], 0.5)
]), 'E', costs[o][d] * (1 - f), 0))
cns.append(('A12(%d)' % idx))
cs.append(
(linex([(vars.r[o][d], 1), (vars.w[o][d], 1),
(vars.hod[o][d], 0.5)]), 'L', costs[o][d], 0))
else:
cns.append(('A11-2(%d)' % idx))
cs.append((linex([(vars.s[o][d], 1)]), 'E', 0, 0))
#///////////////////////////////////////////////////////////////////////
# Transit route network design problem (TRNDP)
#///////////////////////////////////////////////////////////////////////
# for idx, (r,i,j) in it('rss'):
# cns.append('B1(%d)' % idx)
# cs.append((
# linex([(vars.rx0[r][i][j], 1),
# (vars.c[r], 1)]),
# 'L', 1, 0))
# for idx, (r,i,j) in it('rss'):
# cns.append('B1(%d)' % idx)
# cs.append((
# linex([(vars.rx0[r][i][j], 1)] +
# [(vars.xc[r][c], -(candidate_routes[c][i][n_candidate_stops] + candidate_routes[c][n_candidate_stops][j] - 1))
# for c in range(n_candidate_routes)]),
# 'L', 0, 0))
for idx, (i, j) in it('ss'):
cns.append('B2(%d)' % idx)
cs.append(
(linex([(vars.rx[r][i][j], 1)
for r in range(r_max)] + [(vars.rx0[r][i][j], 1)
for r in range(r_max)] +
[(vars.x[o][d][i + n_demands][j + n_demands],
-1.0 / n_demands / n_demands)
for o in range(n_demands)
for d in range(n_demands)]), 'G', 0, 0))
for idx, (r, i, j) in it('rss'):
cns.append('B3(%d)' % idx)
cs.append((linex([(vars.rx[r][i][j], 1)] +
[(vars.xc[r][c], -candidate_routes[c][i][j])
for c in range(n_candidate_routes)]), 'E', 0, 0))
for idx, (r, i) in it('rs'):
cns.append('B4(%d)' % idx)
cs.append((linex([(vars.rx[r][i][n_candidate_stops], 1)] +
[(vars.xc[r][c],
-candidate_routes[c][i][n_candidate_stops])
for c in range(n_candidate_routes)]), 'E', 0, 0))
for idx, (r, i) in it('rs'):
cns.append('B5(%d)' % idx)
cs.append((linex([(vars.rx[r][i][n_candidate_stops + 1], 1)] +
[(vars.xc[r][c],
-candidate_routes[c][i][n_candidate_stops + 1])
for c in range(n_candidate_routes)]), 'E', 0, 0))
for idx, (r, j) in it('rs'):
cns.append('B8(%d)' % idx)
cs.append((linex([(vars.rx[r][n_candidate_stops][j], 1)] +
[(vars.xc[r][c],
-candidate_routes[c][n_candidate_stops][j])
for c in range(n_candidate_routes)]), 'E', 0, 0))
for idx, (r) in it('r'):
cns.append('B6(%d)' % idx)
cs.append((linex([(vars.xc[r][c], 1)
for c in range(n_candidate_routes)]), 'E', 1, 0))
for idx, (r) in it('r'):
cns.append('B7(%d)' % idx)
cs.append((linex([(vars.xn[r][n], 1)
for n in range(n_buses + 1)]), 'E', 1, 0))
# for idx, (r) in it('r'):
# cns.append('B3(%d)' % idx)
# cs.append((
# linex([(vars.rx[r][n_candidate_stops][j], 1)
# for j in range(n_candidate_stops+1)]),
# 'E', 1, 0))
# for idx, (r) in it('r'):
# cns.append('B4(%d)' % idx)
# cs.append((
# linex([(vars.rx[r][i][n_candidate_stops], 1)
# for i in range(n_candidate_stops+1)] +
# [(vars.rx[r][i][n_candidate_stops+1], 1)
# for i in range(n_candidate_stops)]),
# 'E', 1, 0))
# for idx, (r,i) in it('rs'):
# cns.append('B5(%d)' % idx)
# cs.append((
# linex([(vars.rx[r][i][j], 1)
# for j in range(n_candidate_stops+2)
# if i != j] +
# [(vars.rx[r][j][i], -1)
# for j in range(n_candidate_stops+2)
# if i != j]),
# 'E', 0, 0))
# for idx, (r,i) in it('rs'):
# cns.append('B6(%d)' % idx)
# cs.append((
# linex([(vars.rx[r][j][i], 1)
# for j in range(n_candidate_stops+2)]),
# 'L', 1, 0))
# for idx, (r,i) in it('rs'):
# cns.append('B7(%d)' % idx)
# cs.append((
# linex([(vars.rx[r][i][j], 1)
# for j in range(n_candidate_stops+2)]),
# 'L', 1, 0))
# for idx, (r,i) in it('rs'):
# cns.append('B8(%d)' % idx)
# cs.append((linex([(vars.rx[r][i][i], 1)]), 'E', 0, 0))
# for idx, (r) in it('r'):
# cns.append('B9(%d)' % r)
# cs.append((linex([(vars.rx[r][n_candidate_stops][n_candidate_stops+1], 1)]), 'E', 0, 0))
# for idx, (r,i,j) in it('rss'):
# if i != j:
# cns.append('B10(%d)' % idx)
# cs.append((
# linex([(vars.rq[r][i], 1),
# (vars.rq[r][j], -1),
# (vars.rx[r][i][j], 1000)]),
# 'L', 999, 0))
# for idx, (r,i) in it('rs'):
# if i != 0:
# cns.append('B11-1(%d)' % idx)
# cs.append((
# linex([(vars.rx[r][i][n_candidate_stops], 1)] +
# [(vars.rx[r][k][n_candidate_stops], 1)
# for k in range(n_candidate_stops+1)
# if k < i]),
# 'L', 1, 0))
# for idx, (r,i) in it('rs'):
# if i != 0:
# cns.append('B11-2(%d)' % idx)
# cs.append((
# linex([(vars.rx[r][i][n_candidate_stops], 1)] +
# [(vars.rx[r][j][i], -1)
# for j in range(n_candidate_stops+1)]),
# 'L', 0, 0))
# for idx, (r,i) in it('rs'):
# if i != 0:
# cns.append('B11-3(%d)' % idx)
# cs.append((
# linex([(vars.rx[r][i][n_candidate_stops], 1),
# (vars.c[r], 1)]),
# 'L', 1, 0))
# for idx, (r,i) in it('rs'):
# if i != 0:
# cns.append('B11-4(%d)' % idx)
# cs.append((
# linex([(vars.rx[r][i][n_candidate_stops], 1),
# (vars.c[r], 1)] +
# [(vars.rx[r][k][n_candidate_stops], 1)
# for k in range(n_candidate_stops+1)
# if k < i] +
# [(vars.rx[r][j][i], -1)
# for j in range(n_candidate_stops+1)]),
# 'G', 0, 0))
for idx, (r, i, j) in it('rss'):
cns.append('B12-1(%d)' % idx)
cs.append(
(linex([(vars.rx0[r][i][j], 1),
(vars.rx[r][i][n_candidate_stops], -1),
(vars.rx[r][n_candidate_stops][j], -1)]), 'G', -1, 0))
for idx, (r, i, j) in it('rss'):
cns.append('B12-2(%d)' % idx)
cs.append(
(linex([(vars.rx0[r][i][j], 1),
(vars.rx[r][i][n_candidate_stops], -1)]), 'L', 0, 0))
for idx, (r, i, j) in it('rss'):
cns.append('B12-3(%d)' % idx)
cs.append(
(linex([(vars.rx0[r][i][j], 1),
(vars.rx[r][n_candidate_stops][j], -1)]), 'L', 0, 0))
for idx, (r, i, j) in it('rss'):
cns.append('B13(%d)' % idx)
cs.append((linex([(vars.rx[r][i][j], 1),
(vars.rx0[r][i][j], 1)]), 'L', 1, 0))
#///////////////////////////////////////////////////////////////////////
# Frequency determination problem (FDP)
#///////////////////////////////////////////////////////////////////////
# for idx, (r) in it('r'):
# cns.append('C1(%d)' % r)
# cs.append((
# linex([(vars.c[r], 1)] +
# [(vars.rx[r][i][n_candidate_stops+1], -1)
# for i in range(n_candidate_stops)]),
# 'E', 0, 0))
# for idx, (i,j) in it('ss'):
# if i != j:
# cns.append('C3-1(%d)' % idx)
# cs.append((
# linex([(vars.x[o][d][i+n_demands][j+n_demands], demands[o][d])
# for o in range(n_demands)
# for d in range(n_demands)] +
# [(vars.zc[r][c], -max_cap*candidate_routes[c][i][j])
# for c in range(n_candidate_routes)
# for r in range(r_max)] +
# [(vars.z3[r][i][j], -max_cap*candidate_routes[c][i][j])
# for r in range(r_max)]),
# 'L', 0, 0))
for idx, (i, j) in it('ss'):
if i != j:
cns.append('C3-1(%d)' % idx)
cs.append((linex([(vars.x[o][d][i + n_demands][j + n_demands],
demands[o][d]) for o in range(n_demands)
for d in range(n_demands)] +
[(vars.z4[r][i][j], -max_cap)
for r in range(r_max)] +
[(vars.z5[r][i][j], -max_cap)
for r in range(r_max)]), 'L', 0, 0))
for idx, (r, c) in it('rc'):
cns.append('C3-2(%d)' % idx)
cs.append((linex([(vars.zc[r][c], 1),
(vars.f[r], -1)]), 'L', 0, 0))
cns.append('C3-3(%d)' % idx)
cs.append((linex([(vars.zc[r][c], 1),
(vars.xc[r][c], -60)]), 'L', 0, 0))
cns.append('C3-4(%d)' % idx)
cs.append((linex([(vars.zc[r][c], 1), (vars.f[r], -1),
(vars.xc[r][c], -60)]), 'G', -60, 0))
for idx, (r, i, j) in it('rss'):
if i != j:
cns.append('C3-2(%d)' % idx)
cs.append((linex([(vars.z4[r][i][j], 1),
(vars.rx[r][i][j], -60)]), 'L', 0, 0))
for idx, (r, i, j) in it('rss'):
if i != j:
cns.append('C3-3(%d)' % idx)
cs.append((linex([(vars.z4[r][i][j], 1),
(vars.f[r], -1)]), 'L', 0, 0))
for idx, (r, i, j) in it('rss'):
if i != j:
cns.append('C3-4(%d)' % idx)
cs.append((linex([(vars.z4[r][i][j], 1), (vars.f[r], -1),
(vars.rx[r][i][j], -60)]), 'G', -60, 0))
for idx, (r, i, j) in it('rss'):
if i != j:
cns.append('C3-5(%d)' % idx)
cs.append((linex([(vars.z5[r][i][j], 1),
(vars.rx0[r][i][j], -60)]), 'L', 0, 0))
for idx, (r, i, j) in it('rss'):
if i != j:
cns.append('C3-6(%d)' % idx)
cs.append((linex([(vars.z5[r][i][j], 1),
(vars.f[r], -1)]), 'L', 0, 0))
for idx, (r, i, j) in it('rss'):
if i != j:
cns.append('C3-7(%d)' % idx)
cs.append((linex([(vars.z5[r][i][j], 1), (vars.f[r], -1),
(vars.rx0[r][i][j], -60)]), 'G', -60, 0))
for idx, (r, i, j) in it('rss'):
if i != j:
cns.append('C3-5(%d)' % idx)
cs.append((linex([(vars.z3[r][i][j], 1),
(vars.rx0[r][i][j], -60)]), 'L', 0, 0))
for idx, (r, i, j) in it('rss'):
if i != j:
cns.append('C3-6(%d)' % idx)
cs.append((linex([(vars.z3[r][i][j], 1),
(vars.f[r], -1)]), 'L', 0, 0))
for idx, (r, i, j) in it('rss'):
if i != j:
cns.append('C3-7(%d)' % idx)
cs.append((linex([(vars.z3[r][i][j], 1), (vars.f[r], -1),
(vars.rx0[r][i][j], -60)]), 'G', -60, 0))
for idx, (r, i, j) in it('rss'):
if i != j:
cns.append('C3-8(%d)' % idx)
cs.append(
(linex([(vars.z3[r][i][j], 1)] +
[(vars.xc[r][c], -60)
for c in range(n_candidate_routes)]), 'L', 0, 0))
for idx, (r) in it('r'):
cns.append('C4(%d)' % idx)
cs.append((linex([(vars.xn[r][n], n) for n in range(n_buses + 1)] +
[(vars.zc[r][c], -t[c])
for c in range(n_candidate_routes)]), 'E', 0, 0))
cns.append('C5')
cs.append((linex([(vars.xn[r][n], n)
for n in range(n_buses + 1)]), 'L', n_buses, 0))
for idx, (r) in it('r'):
cns.append('C5(%d)' % idx)
cs.append((linex([(vars.f[r], 1)]), 'G', 0.2, 0))
for idx, (r) in it('r'):
cns.append('C6(%d)' % idx)
cs.append(
(linex([(vars.h[r], 1)] +
[(vars.z1[r][n][c], -t[c] / float(n) if n != 0 else 0)
for n in range(n_buses + 1)
for c in range(n_candidate_routes)]), 'E', 0, 0))
for idx, (r, n, c) in it('rbc'):
cns.append('C7-1(%d)' % idx)
cs.append((linex([(vars.z1[r][n][c], 1),
(vars.xn[r][n], -1)]), 'L', 0, 0))
cns.append('C7-2(%d)' % idx)
cs.append((linex([(vars.z1[r][n][c], 1),
(vars.xc[r][c], -1)]), 'L', 0, 0))
cns.append('C7-3(%d)' % idx)
cs.append((linex([(vars.z1[r][n][c], 1), (vars.xc[r][c], -1),
(vars.xn[r][n], -1)]), 'G', -1, 0))
for idx, (o, d, r, j, k) in it('ddrss'):
cns.append('C8-1(%d)' % idx)
cs.append(
(linex([(vars.z2[o][d][r][j][k], 1),
(vars.x[o][d][o][j + n_demands], -60)]), 'L', 0, 0))
cns.append('C8-2(%d)' % idx)
cs.append((linex([(vars.z2[o][d][r][j][k], 1),
(vars.x[o][d][j + n_demands][k + n_demands], -60)
]), 'L', 0, 0))
cns.append('C8-3(%d)' % idx)
cs.append((linex([(vars.z2[o][d][r][j][k], 1),
(vars.rx[r][j][k], -60),
(vars.rx0[r][j][k], -60)]), 'L', 0, 0))
# cns.append('C8-4(%d)' % idx)
# cs.append((
# linex([(vars.z2[o][d][r][j][k], 1),
# (vars.h[r], -1)]),
# 'L', 0, 0))
cns.append('C8-6(%d)' % idx)
cs.append((linex([
(vars.z2[o][d][r][j][k], 1), (vars.h[r], -1),
(vars.rx[r][j][k], -60), (vars.rx0[r][j][k], -60),
(vars.x[o][d][o][j + n_demands], -60),
(vars.x[o][d][j + n_demands][k + n_demands], -60)
]), 'G', -180, 0))
for idx, (o, d) in it('dd'):
cns.append('C9(%d)' % idx)
cs.append((linex([(vars.hod[o][d], 1)] +
[(vars.z2[o][d][r][j][k], -1)
for r in range(r_max)
for j in range(n_candidate_stops)
for k in range(n_candidate_stops) if j != k]),
'E', 0, 0))
cs = zip(*cs)
self._cp.linear_constraints.add(lin_expr=cs[0],
senses=cs[1],
rhs=cs[2],
range_values=cs[3])
self._cp.linear_constraints.set_names(enumerate(cns))
def generate_constraints(self):
r_max = 1
candidate_stops = [[55.785470, 12.523041],[55.786239, 12.522234],[55.786541, 12.526961],[55.789277, 12.523946],[55.786431, 12.521428],[55.785898, 12.520618],[55.785138, 12.520728],[55.784637, 12.520459],[55.782099, 12.514132],[55.782460, 12.519244],[55.782226, 12.520035]]
demand_centroids = [[55.786344,12.524316],[55.786344,12.524316],[55.786872,12.5259],[55.78944,12.525146],[55.786745,12.520469],[55.787486,12.518472],[55.787911,12.518621],[55.785438,12.519377],[55.785154,12.520434],[55.785241,12.518699],[55.783,12.512931],[55.781929,12.521461]]
demands = [[0, 100, 3, 36, 0, 14, 13, 19, 42, 55, 5, 20],
[190, 0, 2, 25, 0, 18, 6, 8, 40, 37, 5, 4],
[13, 10, 0, 3, 0, 3, 0, 1, 5, 3, 0, 0],
[137, 1, 0, 0, 0, 10, 4, 1, 17, 34, 1, 1],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
[47, 9, 2, 4, 0, 0, 63, 0, 0, 6, 2, 0],
[46, 4, 0, 0, 0, 173, 0, 0, 0, 2, 2, 1],
[61, 2, 0, 1, 0, 1, 0, 0, 78, 116, 0, 0],
[116, 13, 1, 5, 0, 8, 0, 52, 0, 84, 16, 8],
[167, 6, 0, 6, 0, 33, 13, 52, 230, 0, 23, 6],
[1, 1, 0, 0, 0, 1, 0, 0, 7, 1, 0, 1],
[97, 55, 0, 6, 0, 3, 2, 2, 47, 39, 9, 0]]
n_candidate_stops = len(candidate_stops)
n_buses = 11
costs = []
for i1, (stop_x1, stop_y1) in enumerate([[0,0]] + candidate_stops):
costs.append([])
for i2, (stop_x2, stop_y2) in enumerate([[0,0]] + candidate_stops):
costs[i1].append(275 * (abs(stop_x1 - stop_x2) + abs(stop_y1 - stop_y2)))
vars = self.vars
cs = []
cns = []
# idx = 0
# for r in range(r_max):
# for i in range(n_candidate_stops):
# for j in range(n_candidate_stops):
# cns.append('RTS(%d)' % idx)
# cs.append((
# linex([(vars.D[r][i][j], 1),
# (vars.x[r][i+1][j+1], -ridetime(candidate_stops[i], candidate_stops[j]))] +
# [(vars.x[r][i+1][x+1], -ridetime(candidate_stops[i], candidate_stops[x]))
# for x in range(n_candidate_stops)
# if x != i and x != j] +
# [(vars.xD[r][i][x][j], -1)
# for x in range(n_candidate_stops)
# if x != i and x != j]),
# 'E', 0, 0))
# idx += 1
# idx = 0
# for r in range(r_max):
# for i in range(n_candidate_stops):
# for x in range(n_candidate_stops):
# for j in range(n_candidate_stops):
# cns.append('XD1(%d)' % idx)
# cs.append((
# linex([(vars.xD[r][i][x][j], 1),
# (vars.D[r][x][j], -1)]),
# 'L', 0, 0))
# cns.append('XD2(%d)' % idx)
# cs.append((
# linex([(vars.xD[r][i][x][j], 1),
# (vars.x[r][i+1][x+1], -100)]),
# 'L', 0, 0))
# cns.append('XD3(%d)' % idx)
# cs.append((
# linex([(vars.xD[r][i][x][j], 1),
# (vars.D[r][x][j], -1),
# (vars.x[r][i+1][x+1], -100)]),
# 'G', -100, 0))
# idx += 1
# max_T = sum(list(map(lambda x: ridetime(x[0], x[1]), combinations(candidate_stops, 2))))
# idx = 0
# for r in range(r_max):
# for i in range(n_candidate_stops):
# for j in range(n_candidate_stops):
# cns.append('SD1(%d)' % idx)
# cs.append((
# linex([(vars.sD[r][i][j], 1),
# (vars.D[r][i][j], -max_T)]),
# 'L', 0, 0))
# cns.append('SXD2(%d)' % idx)
# cs.append((
# linex([(vars.sD[r][i][j], 1),
# (vars.D[r][i][j], -1/max_T)]),
# 'G', 0, 0))
# idx += 1
# cns.append('TS')
# cs.append((
# linex([(vars.D[r][i][j], -1)
# for r in range(r_max)
# for i in range(n_candidate_stops)
# for j in range(n_candidate_stops)] +
# [(vars.sD[r][i][j], walktime(candidate_stops[i], candidate_stops[j]))
# for r in range(r_max)
# for i in range(n_candidate_stops)
# for j in range(n_candidate_stops)] +
# [(vars.ts[0], -1)]),
# 'E', 0, 0))
# for i in range(n_candidate_stops):
# cns.append('STP(%d)' % i)
# cs.append((
# linex([(vars.st[i], 1)] +
# [(vars.x[r][i+1][j], -1)
# for r in range(r_max)
# for j in range(n_candidate_stops+1)]),
# 'L', 0, 0))
# idx = 0
# for r in range(r_max):
# cns.append('ITS(%d)' % idx)
# cs.append((
# linex([(vars.x[r][i+1][j+1], 1)
# for i in range(n_candidate_stops)
# for j in range(n_candidate_stops)
# if i != j]),
# 'L', 1, 0))
# idx += 1
# idx = 0
# for r in range(r_max):
# for i in range(n_candidate_stops + 1):
# for j in range(n_candidate_stops + 1):
# cns.append('XF1(%d)' % idx)
# cs.append((
# linex([(vars.xf[r][i][j], 1),
# (vars.f[r], -1)]),
# 'L', 0, 0))
# cns.append('XF2(%d)' % idx)
# cs.append((
# linex([(vars.xf[r][i][j], 1),
# (vars.x[r][i][j], -60)]),
# 'L', 0, 0))
# cns.append('XF3(%d)' % idx)
# cs.append((
# linex([(vars.xf[r][i][j], 1),
# (vars.f[r], -1),
# (vars.x[r][i][j], -60)]),
# 'G', -60, 0))
# idx += 1
for r in range(r_max):
cns.append('C1(%d)' % r)
cs.append((
linex([(vars.x[r][0][j], 1)
for j in range(n_candidate_stops+1)]),
'E', 1, 0))
# for r in range(r_max):
# cns.append('C2(%d)' % r)
# cs.append((
# linex([(vars.x[r][i][0], 1)
# for i in range(n_candidate_stops+1)]),
# 'E', 1, 0))
idx = 0
for r in range(r_max):
for j in range(n_candidate_stops):
cns.append('C3(%d)' % idx)
cs.append((
linex([(vars.x[r][i][j], 1)
for i in range(n_candidate_stops + 1)
if i != j] +
[(vars.x[r][j][i], -1)
for i in range(n_candidate_stops + 1)
if i != j]),
'E', 0, 0))
idx += 1
idx = 0
for r in range(r_max):
for j in range(n_candidate_stops):
cns.append('C4(%d)' % idx)
cs.append((
linex([(vars.x[r][i][j+1], 1)
for i in range(n_candidate_stops + 1)
if i != (j+1)]),
'L', 1, 0))
idx = 0
for r in range(r_max):
for i in range(n_candidate_stops):
cns.append('C5(%d)' % idx)
cs.append((
linex([(vars.x[r][i+1][j], 1)
for j in range(n_candidate_stops + 1)
if (i+1) != j]),
'L', 1, 0))
idx = 0
for r in range(r_max):
for i in range(n_candidate_stops):
cns.append('C6(%d)' % idx)
cs.append((linex([(vars.x[r][i+1][i+1], 1)]), 'E', 0, 0))
# idx = 0
# for r in range(r_max):
# for i in range(n_candidate_stops + 1):
# for j in range(n_candidate_stops + 1):
# cns.append('C7-1(%d)' % idx)
# cs.append((
# linex([(vars.zxf[r][i][j], 1),
# (vars.xf[r][i][j], -1)]),
# 'L', 0, 0))
# cns.append('C7-2(%d)' % idx)
# cs.append((
# linex([(vars.zxf[r][i][j], 1),
# (vars.x[r][0][0], 60)]),
# 'L', 60, 0))
# cns.append('C7-3(%d)' % idx)
# cs.append((
# linex([(vars.zxf[r][i][j], 1),
# (vars.xf[r][i][j], -1),
# (vars.x[r][0][0], 60)]),
# 'G', 0, 0))
# idx += 1
# cns.append('C8')
# cs.append((
# linex([(vars.zxf[r][i+1][j+1], 2*costs[i+1][j+1])
# for r in range(r_max)
# for i in range(n_candidate_stops)
# for j in range(n_candidate_stops)]),
# 'L', n_buses, 0, 0))
# for r in range(r_max):
# cns.append('C9(%d)' % r)
# cs.append((
# linex([(vars.x[r][i+1][j+1], costs[i+1][j+1])
# for i in range(n_candidate_stops)
# for j in range(n_candidate_stops)
# if i != j] +
# [(vars.T[r], -1)]),
# 'E', 0, 0))
# Stops for demand origin/destination
idx = 0
for p in range(len(demands)):
for q in range(len(demands)):
for c1 in range(n_candidate_stops):
for c2 in range(n_candidate_stops):
cns.append('S1(%d)' % idx)
cs.append((
linex([(vars.s[p][q][c1][c2], 1),
(vars.adm[c1+1][c2+1], -1)]),
'L', 0, 0))
idx += 1
# Stop for demand origin
idx = 0
for p in range(len(demands)):
for q in range(len(demands)):
cns.append('S2(%d)' % idx)
cs.append((
linex([(vars.s[p][q][c1][c2], 1)
for c1 in range(n_candidate_stops)
for c2 in range(n_candidate_stops)]),
'E', 1, 0))
idx += 1
# # Sub tour elimination
# idx = 0
# for r in range(r_max):
# for i in range(n_candidate_stops):
# for j in range(n_candidate_stops):
# if i != j:
# cns.append('SUB(%d)' % idx)
# cs.append((
# linex([(vars.q[r][i], 1),
# (vars.q[r][j], -1),
# (vars.x[r][i+1][j+1], 1000)]),
# 'L', 999, 0))
# idx += 1
# Admissibility
idx = 0
for i in range(n_candidate_stops+1):
for j in range(n_candidate_stops+1):
cns.append('ADM(%d)' % idx)
cs.append((
linex([(vars.adm[i][j], 1)] +
[(vars.x[r][i][j], -1)
for r in range(r_max)] +
[(vars.xadm[r][i][x][j], -1)
for r in range(r_max)
for x in range(n_candidate_stops + 1)]),
'L', 0, 0))
# XADM
idx = 0
for r in range(r_max):
for i in range(n_candidate_stops+1):
for j in range(n_candidate_stops+1):
for x in range(n_candidate_stops+1):
cns.append('XADM1(%d)' % idx)
cs.append((
linex([(vars.xadm[r][i][x][j], 1),
(vars.x[r][i][x], -1)]),
'L', 0, 0))
cns.append('XADM2(%d)' % idx)
cs.append((
linex([(vars.xadm[r][i][x][j], 1),
(vars.adm[x][j], -1)]),
'L', 0, 0))
cns.append('XADM3(%d)' % idx)
cs.append((
linex([(vars.xadm[r][i][x][j], 1),
(vars.adm[x][j], -1),
(vars.x[r][i][x], -1)]),
'G', -1, 0))
idx += 1
idx = 0
for p in range(len(demands)):
for q in range(len(demands)):
cns.append("WAL(%d)" % idx)
cs.append((
linex([(vars.w[p][q], 1)] +
[(vars.s[p][q][c1][c2], -walktime(demand_centroids[p], candidate_stops[c1])-walktime(demand_centroids[p], candidate_stops[c2]))
for c1 in range(n_candidate_stops)
for c2 in range(n_candidate_stops)]),
'E', 0, 0))
idx = 0
for r in range(r_max):
for c2 in range(n_candidate_stops):
cns.append("DEL(%d)" % idx)
cs.append((
linex([(vars.delta[r][c2], 1)] +
[(vars.x[r][i+1][c2+1], -1)
for i in range(n_candidate_stops)]),
'E', 0, 0))
idx += 1
idx = 0
for r in range(r_max):
for c2 in range(n_candidate_stops):
cns.append("DELF1(%d)" % idx)
cs.append((
linex([(vars.deltaf[r][c2], 1),
(vars.f[r], -1)]),
'L', 0, 0))
cns.append("DELF2(%d)" % idx)
cs.append((
linex([(vars.deltaf[r][c2], 1),
(vars.delta[r][c2], -60)]),
'L', 0, 0))
cns.append("DELF3(%d)" % idx)
cs.append((
linex([(vars.deltaf[r][c2], 1),
(vars.f[r], -1),
(vars.delta[r][c2], -60)]),
'G', -60, 0))
idx += 1
idx = 0
for c in range(n_candidate_stops):
cns.append("CAP(%d)" % idx)
cs.append((
linex([(vars.deltaf[r][c], 40)
for r in range(r_max)] +
[(vars.s[p][q][c1][c], -demands[p][q]/15.0)
for c1 in range(n_candidate_stops)
for p in range(len(demands))
for q in range(len(demands[0]))]),
'G', 0, 0))
idx += 1
# idx = 0
# for r in range(r_max):
# cns.append("FRQ(%d)" % idx)
# cs.append((
# linex([(vars.f[r], 1),
# (vars.x[r][0][0], 0.033)]),
# 'G', 0.033, 0))
# idx += 1
cs = zip(*cs)
self._cp.linear_constraints.add(
lin_expr=cs[0],
senses=cs[1],
rhs=cs[2],
range_values=cs[3])
self._cp.linear_constraints.set_names(enumerate(cns))