CN105956681A - Drop-and-pull transport dynamic path planning method based on receding horizon optimization - Google Patents

Abstract

The invention discloses a drop-and-pull transport dynamic path planning method based on receding horizon optimization. The drop-and-pull transport dynamic path planning method is characterized by comprising the following steps of 1, dividing vehicle queue operation time into a plurality of time fragments by means of a rolling optimization method; 2, acquiring an inner dispensing cost matrix of a plurality of future time segments at each time segment; 3, establishing a drop-and-pull transport dynamic path planning model based on receding horizon optimization by means of the dispensing cost matrix; 4, solving an optimal plan in the plurality of future time segments by means of the drop-and-pull transport dynamic path planning model; and 5, randomly generating an initial solution, performing iterative solving on the initial solution by means of a simulated annealing algorithm, thereby obtaining an optimal solution, and executing the optimal solution. The drop-and-pull transport dynamic path planning method can perform structured decision on a drop-and-pull transport dynamic path planning problem and quickly generates a dynamic path planning scheme, thereby improving real-time performance, rationality and accuracy of the dynamic path planning scheme and reducing transport cost.

Description

A kind of Drop and pull transport dynamic path planning method optimized based on rolling time horizon

Technical field

The present invention relates to roller and optimization method, a kind of Drop and pull transport dynamic route optimized based on rolling time horizon Planing method, belongs to Combinatorial Optimization, dynamic decision or dynamic programming field.

Background technology

Drop and pull transport refers to that tractor unloads self-contained trailer at cargo handling operation point, and changes other trailers and continue to run with Total logistics cost mode.Drop and pull transport has obtained popularization and application widely the most in the world, in practice it has proved that be a kind of efficient, green Color, advanced freight transportation organizing mode, to reducing logistics cost, promote modern logistics development, improves country's macroeconomic fortune Row quality, the most significant.

Static Drop and pull transport path planning problem is unobstructed can be described as follows: utilize tractor trailer converter dolly, from parking lot by based on Draw and access Drop and pull transport user, after completing whole visiting demand, return to parking lot.Wherein the demand of Drop and pull transport user can be divided into two Class: individually need tractor access and need tractor trailer converter dolly jointly to access.Meanwhile, the operation of pulling of tractor is only capable of Particular station completes.Problem is intended to, by constructing suitable route or travel by vehicle and getting rid of hanging method, meeting relevant constraint Meanwhile, make total distribution cost minimum.

At present, the correlational study both at home and abroad container being hung surely to transportation problem is more, but specializes in Drop and pull transport scheduling problem Document relatively fewer；Research simultaneously for Drop and pull transport scheduling problem is appointed main with the path planning in the environment of static state and fortune Defeated Model Design is main, less, to delivery system each several part in scheduling process for the correlational study dynamically and under uncertain environment Between concertedness also take into full account, thus achievement in research is difficult to successfully manage the impact that actual dispensing environmental change brings.And Core people's task of dynamic dispatching makes scheduling scheme when in the face of time-varying, complicated actual implementation environment, it is possible to preferably Implementation is collaborative with enforcement, obtains optimal actual implementation effect.

Summary of the invention

The present invention, in order to reduce the deviation produced in fleet's actual moving process with path planning scheme, overcomes existing route planning side The weak point that method exists, proposes a kind of Drop and pull transport dynamic path planning method optimized based on rolling time horizon, to can be to base Drop and pull transport active path planning problem in rolling time horizon optimization carries out structured decision, and quickly provides active path planning Scheme, thus improve the real-time of path planning scheme, reasonability and accuracy and reduce cost of transportation.

The present invention solves that technical problem adopts the following technical scheme that

The feature of a kind of Drop and pull transport dynamic path planning method based on rolling time horizon optimization of the present invention is: be applied to by 1 car A, R the destination node U in field₁, V road-net node U₂And the delivery service that P tractor position node S is formed In region, an external foursquare summit in described delivery service region is set to zero o, will be with described zero o Both horizontally and vertically go up two adjacent sides being connected are respectively set to x-axis and y-axis, thus form right angle coordinate system xoy；Institute State in coordinate system xoy:

The position of definition parking lot A is (x₀,y₀), define described P tractor all from the position (x of described parking lot A₀,y₀Go out in) Send out；Described P tractor is designated as K={K⁽¹⁾,K⁽²⁾,…,K^(p),…,K^(P), 1≤p≤P, K^(p)Represent pth traction Car；

R destination node is designated as Represent the r target joint Point, the r destination nodePosition be designated as (x_r,y_r)；By the r destination nodeThe time of access the earliest be designated as e_r； By the r destination nodeThe time of access at the latest be designated as l_r；e_rAnd l_rConstitute the time window [e of destination node_r,l_r]；

V road-net node is designated as Represent v Road-net node；The v road-net nodePosition be designated as (x_R+v,y_R+v)；

P tractor K position node is designated as S={S^(R+V+1),S^(R+V+2),…,S^(R+V+p),…,S^(R+V+P), S^(R+V+p) Represent the R+V+p node, represent pth tractor K^(p)Position；Pth tractor K^(p)Position node S^(R+S+p)Position be designated as (x_R+V+p,y_R+V+p)；

By described 1 parking lot A, R destination node U₁, V road-net node U₂, P tractor position node S, It is designated as point set U={A, U₁,U₂, S}, the most described parking lot A represent the 0th node；R destination node U₁Represent the 1st node To the R node；V road-net node U₂Represent that the R+1 node is to the R+S node；P tractor position Node S represents that R+V+1 node is to the R+V+P node；

Definition limit collection E={<i, j>| i, j ∈ U, i ≠ j} represent the straight line path set of any two node,<i, j>in described point set U Represent the straight line path between any i-th node and any jth node；Remember in described point set U that any two is internodal to join Sending Cost matrix is C, and C={c_ij|i,j∈U,i≠j}；c_ijRepresent the distribution cost between i-th node and jth node；

Described Drop and pull transport dynamic path planning method based on rolling time horizon optimization is to carry out as follows:

Step 1, by activity duration [0, the max l of all tractor K_r] it is divided into N number of continuous print time slice at equal intervals, it is designated as {[0,Δt],[Δt,2Δt],…,[(n-1)Δt,nΔt],…,[(N-1)Δt,maxl_r]}；When wherein [(n-1) Δ t, n Δ t] represents n-th Between fragment；Δ t represents divided interval；

Step 2, make n=1；Actual travel route to the n-th to the n-th+m time period [(n-1) Δ t, (n+m) Δ t] interior vehicle Plan；1≤m≤N；N is time slice sum；

Step 3, make the n-th to the n-th+m time period [(n-1) Δ t, (n+m) Δ t] interior pth tractor K^(p)Place initial bit Put nodeThus initialize all tractor places initial position node；

Step 4, obtain described i-th node and jth node in the n-th to the n-th+m time period [(n-1) Δ t, (n+m) Δ t] Between distribution costThus obtain distribution cost Matrix C_n；

Step 5, according to pth tractor K^(p)Destination node to be accessed, according to destination node to be accessed the earliest The sequencing of access time, will fall into all destination nodes in the n-th to the n-th+m time period [(n-1) Δ t, (n+m) Δ t] Composition pth tractor K successively^(p)Initial access sequence

Step 6, taking-up pth tractor K^(p)Initial access sequenceIn last destination node as purpose save Point, is designated asThus form the access sequence of renewal

Step 7, according to pth tractor K^(p)Place initial position nodeDistribution cost Matrix C_n, access sequenceDestination nodeSet up the Drop and pull transport active path planning model optimized based on rolling time horizon；

Step 8, stochastic generation n-th initial solution X_n；

Step 9, utilize simulated annealing to described n-th initial solution X_nIt is iterated solving, it is thus achieved that n-th optimal solution；

Step 10, n+1 is assigned to n；

Step 11, the n-th to the n-th+m time period [(n-1) Δ t, (n+m) Δ t] interior pth tractor K of renewal^(p)Position NodeDistribution cost Matrix C_n；

Step 12, before [(n-2) Δ t, (n-1) Δ t] terminates (n-1)th time period, obtain pth tractor K^(p)Arrive n-th Access sequence in n-th+m time period [(n-1) Δ t, (n+m) Δ t]Described access sequenceIncluding: (n-1)th Destination node and pth the tractor K of access it is not fully complete in time period [(n-2) Δ t, (n-1) Δ t]^(p)Target to be accessed The time of access the earliest of node falls into the destination node of the n-th to the n-th+m time period [(n-1) Δ t, (n+m) Δ t]；

Step 13, the n-th to the n-th+m time period [(n-1) Δ t, (n+m) Δ t] interior pth tractor K of taking-up^(p)Access sequence RowLast destination node as destination nodeThus form the access sequence of renewal

Step 14, according to pth tractor K^(p)Place initial position nodeDistribution cost Matrix C_n, access sequence RowDestination nodeInput in described Drop and pull transport active path planning model；

Step 15, initial solution X of stochastic generation n-th_n,

Step 16, utilize simulated annealing to described n-th initial solution X_nIt is iterated solving, it is thus achieved that n-th optimal solution；

Step 17, judge whether n=N sets up, if setting up, then it represents that all tractor K fulfil assignment；Otherwise, step is returned Rapid 10.

The feature of Drop and pull transport dynamic path planning method based on rolling time horizon optimization of the present invention is, in described step 7 The model of Drop and pull transport active path planning based on rolling time horizon optimization be:

(1) object function:

$\min Z=\underset{i\&Element;U}{\mathrm{\&Sigma;}}\underset{j\&Element;U}{\mathrm{\&Sigma;}}\underset{p=1}{\overset{P}{\mathrm{\&Sigma;}}}{c}_{ij}\&CenterDot;x_{ij}^{\left(p\right)}---\left(1\right)$

(2) constraints

$\underset{j}{\mathrm{\&Sigma;}}\underset{p=1}{\overset{P}{\mathrm{\&Sigma;}}}{x}_{S_n^{(R+S+p)}j}^{\left(p\right)}=P---\left(2\right)$

$\underset{i}{\mathrm{\&Sigma;}}\underset{p=1}{\overset{P}{\mathrm{\&Sigma;}}}{x}_{{\mathrm{iE}}_p^{\left(n\right)}}^{\left(p\right)}=P---\left(3\right)$

${\mathrm{AT}}_{S_n^{(R+S+p)}}^{\left(p\right)}=Time---\left(4\right)$

$x_{ij}^{\left(p\right)}=1\&DoubleRightArrow;{\mathrm{AT}}_i^{\left(p\right)}+w_i+t_{ij}\&le;{\mathrm{AT}}_j^{\left(p\right)}---\left(5\right)$

$e_r\&le;{\mathrm{AT}}_r^{\left(p\right)}\&le;l_r---\left(6\right)$

$\underset{i\&Element;U}{\mathrm{\&Sigma;}}\underset{p=1}{\overset{P}{\mathrm{\&Sigma;}}}{x}_{iw}^{\left(p\right)}=\underset{i\&Element;U}{\mathrm{\&Sigma;}}\underset{p=1}{\overset{P}{\mathrm{\&Sigma;}}}{x}_{wj}^{\left(p\right)},w\&Element;(U_2\&cup;H_p^{\left(n\right)})---\left(7\right)$

${\mathrm{AT}}_{H_p^{\left(n\right)\left(z\right)}}^{\left(p\right)}\&le;{\mathrm{AT}}_{H_p^{\left(n\right)(z+1)}}^{\left(p\right)}---\left(8\right)$

$\underset{i\&Element;U}{\mathrm{\&Sigma;}}\underset{p=1}{\overset{P}{\mathrm{\&Sigma;}}}{x}_{iu}^{\left(p\right)}=\underset{j\&Element;U}{\mathrm{\&Sigma;}}\underset{p=1}{\overset{P}{\mathrm{\&Sigma;}}}{x}_{uj}^{\left(p\right)}=1,u\&Element;H_p^{\left(n\right)}---\left(9\right)$

$x_{ij}^{\left(p\right)}=0or1---\left(10\right)$

Formula (1) is object function, represents that total distribution cost of described tractor K minimizes；WhenTime, represent that pth is led Draw car K^(p)Straight line path ＜ i, j ＞ between described i-th node and jth node, whenTime, represent that pth is led Draw car K^(p)Without straight line path ＜ i, j ＞ between described i-th node and jth node；

Formula (2) represents pth tractor K^(p)Must be from nodes of locationsSet out；WhenTime, represent pth Tractor K^(p)Through nodes of locationsAnd the path between jth nodeWhenTime, table Show pth tractor K^(p)Without nodes of locationsAnd the path between jth node

Formula (3) represents that pth tractor must arrive destination nodeWhenTime, represent pth tractor K^(p)Warp Cross i-th node and destination nodeBetween pathWhenTime, represent pth tractor K^(p)Without Cross i-th node and destination nodeBetween path

Formula (4) is time windows constraints, represents and makes pth tractor K^(p)In nodes of locationsTime be current time Time；Represent pth tractor K^(p)Arriving at location nodeTime；

Formula (5) is worked as representingDuring establishment, it is thus achieved that pth tractor K^(p)The access time on jth node, AT_i ^(p) Represent pth tractor K^(p)Arrive at the time of i-th node；w_iRepresent pth tractor K^(p)At i-th node Waiting time；t_ijRepresent pth tractor K^(p)From the internodal running time of i-th node to jth；Table Show expression pth tractor K^(p)Arrive at the time of jth node；

Formula (6) is time windows constraints, represents pth tractor K^(p)Arrive the r destination nodeTime access the earliest e_rAccess time l at the latest_rBetween；

Formula (7) represents all road-net node U₂And access sequenceIn destination node meet come in and go out balance；Represent pth Tractor K^(p)Straight line path ＜ i, w between described i-th node and the w node >, whenTime, represent pth Tractor K^(p)Without straight line path<i, w>between described i-th node and the w node；WhenTime, represent pth Tractor K^(p)Through described straight line path<w, j>between w node and jth node, whenTime, represent pth Tractor K^(p)Without described straight line path between w node and jth node；

Formula (8) represents access sequenceThe visit of interior joint time order constrained；DescribedRepresent pth tractor K^(p)Arrive Access sequenceIn time of z destination node；DescribedRepresent pth tractor K^(p)Arrive access sequenceIn time of z+1 destination node；

Formula (9) represents access sequenceIn node, only by one traction train number access once；WhenTime, represent pth Tractor K^(p)Straight line path<i, u>between described i-th node and the u node, whenTime, represent pth Tractor K^(p)Without straight line path<i, u>between described i-th node and the u node；WhenTime, represent pth Tractor K^(p)Through described straight line path between u node and jth node<u,j>, whenTime, represent pth Tractor K^(p)Without described straight line path between u node and jth node<u,j>；

Formula (10) represents that the value of decision variable is " 0 " or " 1 ".

Compared with prior art, beneficial effects of the present invention is embodied in:

1, the present invention is compared to static Drop and pull transport paths planning method, it is proposed that apply to rolling time horizon optimization method solve to move The Drop and pull transport vehicle dispatching problem of state, breaches the limitation of original static path planning method.Based on rolling time horizon optimization Drop and pull transport dynamic path planning method overcome its path planning scheme perform during control in terms of deficiency, thus Can in real time, be dynamically generated path planning scheme, enable path planning scheme preferably to meet actual execution environment.

2, the present invention proposes rolling time horizon optimization method framework.It is excellent that rolling time horizon optimization method uses roller finite time-domain Changing strategy, the most disposable off-line of optimisation strategy of i.e. full task process completes, but enters the most online OK.Concrete path optimization's strategy is to enter in following finite time fragment (optimization time window) based on current sample time OK, execution route optimum results and in optimization time window based on the sampling time.When next sampling instant arrives, this Optimize time window together with time move, therefore the optimization of this rolling time horizon introduce dynamic distribution cost, its dynamic is embodied in difference Sampling instant dynamic cost matrix also can change constantly.For simplified model, present invention provide that from the beginning of fleet performs task, Fixing is a sampling instant at interval of Δ t.

3, the present invention designs and establishes Drop and pull transport dynamic vehicle paths planning method model so that when each objective optimization Between in window, this model can obtain standardization and solve.What the derivation algorithm of model was chosen is simulated annealing, simulated annealing Algorithm has the feature of strong robustness, versatility, such that it is able to quickly provide the path planning scheme of optimization.

4, the present invention is by being dynamically generated distribution cost matrix, it is achieved that the Drop and pull transport dynamic route optimized based on rolling time horizon The dynamic of planing method.Appoint in advance road surface unobstructed in the case of road surface cost on the basis of, in conjunction with during to following one section The traffic capacity in interior section and the predictive coefficient of situation, just can be met the dynamic route Cost matrix of present case.Fortune Can the dynamic of reaction model well with dynamic route Cost matrix.

Accompanying drawing explanation

Fig. 1 is implementation flow chart of the present invention；

Fig. 2 is present invention Drop and pull transport active path planning simplified schematic diagram based on billowing dynamic Optimization of Time Domain.

Detailed description of the invention

In the present embodiment, a kind of Drop and pull transport dynamic path planning method based on rolling time horizon optimization is to be applied to by 1 car A, R the destination node U in field₁, V road-net node U₂And the delivery service that P tractor position node S is formed In region, an external foursquare summit in delivery service region is set to zero o, by be connected with zero o Two adjacent sides both horizontally and vertically gone up are respectively set to x-axis and y-axis, thus form right angle coordinate system xoy；In coordinate system In xoy:

The position of definition parking lot A is (x₀,y₀), define P tractor all from the position (x of parking lot A₀,y₀Set out in)；By P Tractor is designated as K={K⁽¹⁾,K⁽²⁾,…,K^(p),…,K^(P), 1≤p≤P, K^(p)Represent pth tractor；P tractor With P tractor position node one_to_one corresponding；

R destination node is designated as Represent the r target joint Point, the r destination nodePosition be designated as (x_r,y_r)；By the r destination nodeThe time of access the earliest be designated as e_r； By the r destination nodeThe time of access at the latest be designated as l_r；e_rAnd l_rConstitute the time window [e of destination node_r,l_r]；

V road-net node is designated as Represent v Road-net node；The v road-net nodePosition be designated as (x_R+v,y_R+v)；

P tractor K position node is designated as S={S^(R+V+1),S^(R+V+2),…,S^(R+V+p),…,S^(R+V+P), S^(R+V+p) Represent the R+V+p node, represent pth tractor K^(p)Position；Pth tractor K^(p)Position node S^(R+S+p)Position be designated as (x_R+V+p,y_R+V+p)；P represents the quantity of tractor, also illustrates that the number of tractor position node Amount；

By 1 parking lot A, R destination node U₁, V road-net node U₂, P tractor position node S, it is designated as Point set U={A, U₁,U₂, S}, then parking lot A represents the 0th node；R destination node U₁Represent that the 1st node is to R Individual node；V road-net node U₂Represent that the R+1 node is to the R+S node；P tractor position node S table Show that R+V+1 node is to the R+V+P node；Network has altogether R+V+P node；

Definition limit collection E={<i, j>| i, j ∈ U, i ≠ j} represent the straight line path set of any two node in point set U, ＜ i, j ＞ table Show the straight line path between any i-th node and any jth node；The internodal distribution cost of any two in note point set U Matrix is C, and C={ci_j|i,j∈U,i≠j}；ci_jRepresent the distribution cost between i-th node and jth node；

In the present embodiment, as it is shown in figure 1, Drop and pull transport dynamic path planning method based on rolling time horizon optimization is by following step Suddenly carry out:

Step 1, by activity duration [0, the max l of all tractor K_r] it is divided into N number of continuous print time slice at equal intervals, to lead Draw car and access the end as whole tasks of last destination node, max l here_rTo illustrate whole task possible the latest End time, be designated as [0, Δ t], [Δ t, 2 Δ t] ..., [(n-1) Δ t, n Δ t] ..., [(N-1) Δ t, max l_r]}；Wherein [(n-1) Δ t, n Δ t] represents the n-th time slice；Δ t represents divided interval；

Step 2, make n=1；Actual travel route to the n-th to the n-th+m time period [(n-1) Δ t, (n+m) Δ t] interior vehicle Plan；1≤m≤N；N is time slice sum；In the value of m represents each Rolling Planning m+1 time period Destination node；

Step 3, make the n-th to the n-th+m time period [(n-1) Δ t, (n+m) Δ t] interior pth tractor K^(p)Place initial bit Put nodeThus initialize all tractor places initial position node；Tractor most starts to rest in Parking lot, this step is by all for acquisition tractor places initial position node；

Step 4, the n-th to the n-th+m time period [(n-1) Δ t, (n+m) Δ t] of acquisition are between interior i-th node and jth node Distribution costThus obtain distribution cost Matrix C_n；Obtain distribution cost Matrix C_nMethod: appoint road in advance Road surface cost C in the case of face is unobstructed⁰On the basis of, in conjunction with to the traffic capacity in section in following a period of time and the prediction of situation Coefficient, just can obtain reacting the traffic capacity in section and situation dynamic route Cost matrix；

Step 5, according to pth tractor K^(p)Destination node to be accessed, according to destination node to be accessed the earliest The sequencing of access time, will fall into all destination nodes in the n-th to the n-th+m time period [(n-1) Δ t, (n+m) Δ t] Composition pth tractor K successively^(p)Initial access sequence

Step 6, taking-up pth tractor K^(p)Initial access sequenceIn last destination node as purpose save Point, is designated asThus form the access sequence of renewal BeOn the basis of, eliminateIn Obtained by later destination node；

Step 7, according to pth tractor K^(p)Place initial position nodeDistribution cost Matrix C_n, access sequenceDestination nodeSet up the Drop and pull transport active path planning model optimized based on rolling time horizon；Dynamic route is utilized to advise Draw model to carry out path planning and need to obtain 4 parameter informations in advance；

The model of the Drop and pull transport active path planning based on rolling time horizon optimization in step 7 is:

(1) object function:

$\min Z=\underset{i\&Element;U}{\mathrm{\&Sigma;}}\underset{j\&Element;U}{\mathrm{\&Sigma;}}\underset{p=1}{\overset{P}{\mathrm{\&Sigma;}}}{c}_{ij}\&CenterDot;x_{ij}^{\left(p\right)}---\left(1\right)$

(2) constraints

$\underset{j}{\mathrm{\&Sigma;}}\underset{p=1}{\overset{P}{\mathrm{\&Sigma;}}}{x}_{S_n^{(R+S+p)}j}^{\left(p\right)}=P---\left(2\right)$

$\underset{i}{\mathrm{\&Sigma;}}\underset{p=1}{\overset{P}{\mathrm{\&Sigma;}}}{x}_{{\mathrm{iE}}_p^{\left(n\right)}}^{\left(p\right)}=P---\left(3\right)$

${\mathrm{AT}}_{S_n^{(R+S+p)}}^{\left(p\right)}=Time---\left(4\right)$

$x_{ij}^{\left(p\right)}=1\&DoubleRightArrow;{\mathrm{AT}}_i^{\left(p\right)}+w_i+t_{ij}\&le;{\mathrm{AT}}_j^{\left(p\right)}---\left(5\right)$

$e_r\&le;{\mathrm{AT}}_r^{\left(p\right)}\&le;l_r---\left(6\right)$

$\underset{i\&Element;U}{\mathrm{\&Sigma;}}\underset{p=1}{\overset{P}{\mathrm{\&Sigma;}}}{x}_{iw}^{\left(p\right)}=\underset{i\&Element;U}{\mathrm{\&Sigma;}}\underset{p=1}{\overset{P}{\mathrm{\&Sigma;}}}{x}_{wj}^{\left(p\right)},w\&Element;(U_2\&cup;H_p^{\left(n\right)})---\left(7\right)$

${\mathrm{AT}}_{H_p^{\left(n\right)\left(z\right)}}^{\left(p\right)}\&le;{\mathrm{AT}}_{H_p^{\left(n\right)(z+1)}}^{\left(p\right)}---\left(8\right)$

$\underset{i\&Element;U}{\mathrm{\&Sigma;}}\underset{p=1}{\overset{P}{\mathrm{\&Sigma;}}}{x}_{iu}^{\left(p\right)}=\underset{j\&Element;U}{\mathrm{\&Sigma;}}\underset{p=1}{\overset{P}{\mathrm{\&Sigma;}}}{x}_{uj}^{\left(p\right)}=1,u\&Element;H_p^{\left(n\right)}---\left(9\right)$

$x_{ij}^{\left(p\right)}=0or1---\left(10\right)$

Formula (1) is object function, represents that total distribution cost of tractor K minimizes；WhenTime, represent pth tractor K^(p)Straight line path ＜ i, j ＞ between i-th node and jth node, whenTime, represent pth tractor K^(p)No Straight line path ＜ i, j ＞ between i-th node and jth node；

Formula (2) represents pth tractor K^(p)Must be from nodes of locationsSet out；WhenTime, represent pth Tractor K^(p)Through nodes of locationsAnd the path between jth nodeWhenTime, table Show pth tractor K^(p)Without nodes of locationsAnd the path between jth node

Formula (3) represents that pth tractor must arrive destination nodeWhenTime, represent pth tractor K^(p)Warp Cross i-th node and destination nodeBetween pathWhenTime, represent pth tractor K^(p)Without Cross i-th node and destination nodeBetween path

Formula (4) is time windows constraints, represents and makes pth tractor K^(p)In nodes of locationsTime be current time Time；Represent pth tractor K^(p)Arriving at location nodeTime；

Formula (5) is worked as representingDuring establishment, it is thus achieved that pth tractor K^(p)The access time on jth node, AT_i ^(p) Represent pth tractor K^(p)Arrive at the time of i-th node；w_iRepresent pth tractor K^(p)At i-th node Waiting time；t_ijRepresent pth tractor K^(p)From the internodal running time of i-th node to jth；Table Show expression pth tractor K^(p)Arrive at the time of jth node；

Formula (6) is time windows constraints, represents pth tractor K^(p)Arrive the r destination nodeTime access the earliest e_rAccess time l at the latest_rBetween；

Formula (7) represents all road-net node U₂And access sequenceIn destination node meet come in and go out balance；Represent pth Tractor K^(p)Straight line path ＜ i between i-th node and the w node, w ＞, whenTime, represent that pth is led Draw car K^(p)Without straight line path ＜ i between i-th node and the w node, w ＞；WhenTime, represent pth traction Car K^(p)Through the straight line path between w node and jth node < w, j ＞, whenTime, represent pth tractor K^(p)Without the straight line path between w node and jth node；

Formula (8) represents access sequenceThe visit of interior joint time order constrained；Represent pth tractor K^(p)Arrive and visit Ask sequenceIn time of z destination node；Represent pth tractor K^(p)Arrive access sequenceIn The time of z+1 destination node；Represent access sequenceIn the z destination node；

Formula (9) represents access sequenceIn node, only by one traction train number access once；WhenTime, represent pth Tractor K^(p)Straight line path ＜ i between i-th node and the u node, u ＞, whenTime, represent that pth is led Draw car K^(p)Without straight line path ＜ i between i-th node and the u node, u ＞；WhenTime, represent pth traction Car K^(p)Through the straight line path ＜ u, j ＞ between u node and jth node, whenTime, represent pth tractor K^(p)Without the straight line path ＜ u, j ＞ between u node and jth node；

Formula (10) represents that the value of decision variable is " 0 " or " 1 ".

The model of the Drop and pull transport active path planning optimized based on rolling time horizon according to Fig. 2, at certain timing node, to not Carry out the destination node in several time slices and carry out dynamic programming.It is illustrated as 2 tractors, corresponding 2 tractors The path planning scheme schematic diagram that position node, 7 destination nodes, and road-net node produce.

Step 8, stochastic generation n-th initial solution X_n；

Step 9, utilize simulated annealing to n-th initial solution X_nIt is iterated solving, it is thus achieved that n-th optimal solution；

Step 10, n+1 is assigned to n；

Step 11, the n-th to the n-th+m time period [(n-1) Δ t, (n+m) Δ t] interior pth tractor K of renewal^(p)Position NodeDistribution cost Matrix C_n；And C_nPreparation method with step 3, step 4；

Step 12, in order to preferably subsequent time slice be planned, Δ T time in advance of can arranging here, time the most each Between before sheet terminates, carry previous Δ T duration and subsequent time slice planned, herein for simplified model, make Δ T=0；? Before (n-1)th time period [(n-2) Δ t, (n-1) Δ t] terminates, obtain pth tractor K^(p)The the n-th to the n-th+m time Access sequence in section [(n-1) Δ t, (n+m) Δ t]Access sequenceIncluding: (n-1)th time period Destination node and pth the tractor K of access it is not fully complete in [(n-2) Δ t, (n-1) Δ t]^(p)Destination node to be accessed The time that accesses the earliest falls into the destination node of the n-th to the n-th+m time period [(n-1) Δ t, (n+m) Δ t]；HereinAcquisition Mode is different with step 5；

Step 13, the n-th to the n-th+m time period [(n-1) Δ t, (n+m) Δ t] interior pth tractor K of taking-up^(p)Access sequence RowLast destination node as destination nodeThus form the access sequence of renewal

Step 14, according to pth tractor K^(p)Place initial position nodeDistribution cost Matrix C_n, access sequence RowDestination nodeIn input Drop and pull transport active path planning model；Active path planning model is utilized to carry out path Planning needs to obtain 4 parameter informations in advance；

Step 15, initial solution X of stochastic generation n-th_n,

Step 16, utilize simulated annealing to n-th initial solution X_nIt is iterated solving, it is thus achieved that n-th optimal solution；

Step 17, judge whether n=N sets up, if setting up, then it represents that all tractor K fulfil assignment；Otherwise, step is returned Rapid 10.

Claims (2)

1. the Drop and pull transport dynamic path planning method optimized based on rolling time horizon, is characterized in that: be applied to by 1 parking lot A, R destination node U₁, V road-net node U₂And the delivery service district that P tractor position node S is formed In territory, an external foursquare summit in described delivery service region is set to zero o, will be with described zero o phase Both horizontally and vertically go up two adjacent sides even are respectively set to x-axis and y-axis, thus form right angle coordinate system xoy；Described In coordinate system xoy:

The position of definition parking lot A is (x₀,y₀), define described P tractor all from the position (x of described parking lot A₀,y₀Go out in) Send out；Described P tractor is designated as K={K⁽¹⁾,K⁽²⁾,…,K^(p),…,K^(P), 1≤p≤P, K^(p)Represent pth traction Car；

R destination node is designated as Represent the r target joint Point, the r destination nodePosition be designated as (x_r,y_r)；By the r destination nodeThe time of access the earliest be designated as e_r； By the r destination nodeThe time of access at the latest be designated as l_r；e_rAnd l_rConstitute the time window [e of destination node_r,l_r]；

V road-net node is designated as Represent v Road-net node；The v road-net nodePosition be designated as (x_R+v,y_R+v)；

P tractor K position node is designated as S={S^(R+V+1),S^(R+V+2),…,S^(R+V+p),…,S^(R+V+P), S^(R+V+p) Represent the R+V+p node, represent pth tractor K^(p)Position；Pth tractor K^(p)Position node S^(R+S+p)Position be designated as (x_R+V+p,y_R+V+p)；

By described 1 parking lot A, R destination node U₁, V road-net node U₂, P tractor position node S, It is designated as point set U={A, U₁,U₂, S}, the most described parking lot A represent the 0th node；R destination node U₁Represent the 1st node To the R node；V road-net node U₂Represent that the R+1 node is to the R+S node；P tractor position Node S represents that R+V+1 node is to the R+V+P node；

Definition limit collection E={<i, j>| i, j ∈ U, i ≠ j} represent the straight line path set of any two node,<i, j>in described point set U Represent the straight line path between any i-th node and any jth node；Remember in described point set U that any two is internodal to join Sending Cost matrix is C, and C={c_ij|i,j∈U,i≠j}；c_ijRepresent the distribution cost between i-th node and jth node；

Described Drop and pull transport dynamic path planning method based on rolling time horizon optimization is to carry out as follows:

Step 1, by the activity duration [0, maxl of all tractor K_r] it is divided into N number of continuous print time slice at equal intervals, it is designated as {[0,Δt],[Δt,2Δt],…,[(n-1)Δt,nΔt],…,[(N-1)Δt,maxl_r]}；When wherein [(n-1) Δ t, n Δ t] represents n-th Between fragment；Δ t represents divided interval；

Step 2, make n=1；Actual travel route to the n-th to the n-th+m time period [(n-1) Δ t, (n+m) Δ t] interior vehicle Plan；1≤m≤N；N is time slice sum；

Step 3, make the n-th to the n-th+m time period [(n-1) Δ t, (n+m) Δ t] interior pth tractor K^(p)Place initial bit Put nodeThus initialize all tractor places initial position node；

Step 4, obtain described i-th node and jth node in the n-th to the n-th+m time period [(n-1) Δ t, (n+m) Δ t] Between distribution costThus obtain distribution cost Matrix C_n；

Step 5, according to pth tractor K^(p)Destination node to be accessed, according to destination node to be accessed the earliest The sequencing of access time, will fall into all destination nodes in the n-th to the n-th+m time period [(n-1) Δ t, (n+m) Δ t] Composition pth tractor K successively^(p)Initial access sequence

Step 6, taking-up pth tractor K^(p)Initial access sequenceIn last destination node as purpose save Point, is designated asThus form the access sequence of renewal

Step 7, according to pth tractor K^(p)Place initial position nodeDistribution cost Matrix C_n, access sequenceDestination nodeSet up the Drop and pull transport active path planning model optimized based on rolling time horizon；

Step 8, stochastic generation n-th initial solution X_n；

Step 9, utilize simulated annealing to described n-th initial solution X_nIt is iterated solving, it is thus achieved that n-th optimal solution；

Step 10, n+1 is assigned to n；

Step 11, the n-th to the n-th+m time period [(n-1) Δ t, (n+m) Δ t] interior pth tractor K of renewal^(p)Position NodeDistribution cost Matrix C_n；

Step 12, before [(n-2) Δ t, (n-1) Δ t] terminates (n-1)th time period, obtain pth tractor K^(p)Arrive n-th Access sequence in n-th+m time period [(n-1) Δ t, (n+m) Δ t]Described access sequenceIncluding: (n-1)th Destination node and pth the tractor K of access it is not fully complete in time period [(n-2) Δ t, (n-1) Δ t]^(p)Target to be accessed The time of access the earliest of node falls into the destination node of the n-th to the n-th+m time period [(n-1) Δ t, (n+m) Δ t]；

Step 13, the n-th to the n-th+m time period [(n-1) Δ t, (n+m) Δ t] interior pth tractor K of taking-up^(p)Access sequence RowLast destination node as destination nodeThus form the access sequence of renewal

Step 14, according to pth tractor K^(p)Place initial position nodeDistribution cost Matrix C_n, access sequence RowDestination nodeInput in described Drop and pull transport active path planning model；

Step 15, initial solution X of stochastic generation n-th_n,

Step 16, utilize simulated annealing to described n-th initial solution X_nIt is iterated solving, it is thus achieved that n-th optimal solution；

Step 17, judge whether n=N sets up, if setting up, then it represents that all tractor K fulfil assignment；Otherwise, step is returned Rapid 10.

The Drop and pull transport dynamic path planning method optimized based on rolling time horizon the most according to claim 1, is characterized in that, The model of the Drop and pull transport active path planning based on rolling time horizon optimization in described step 7 is:

(1) object function:

$\min Z=\underset{i\&Element;U}{\&Sigma;}\underset{j\&Element;U}{\&Sigma;}\underset{p=1}{\overset{P}{\&Sigma;}}{c}_{ij}\&CenterDot;x_{ij}^{\left(p\right)}---\left(1\right)$

(2) constraints

$\underset{j}{\&Sigma;}\underset{p=1}{\overset{P}{\&Sigma;}}{x}_{S_n^{(R+S+p)}j}^{\left(p\right)}=P---\left(2\right)$

$\underset{i}{\&Sigma;}\underset{p=1}{\overset{P}{\&Sigma;}}{x}_{{\mathrm{iE}}_p^{\left(n\right)}}^{\left(p\right)}=P---\left(3\right)$

${\mathrm{AT}}_{S_n^{(R+S+p)}}^{\left(p\right)}=Time---\left(4\right)$

$x_{ij}^{\left(p\right)}=1\&DoubleRightArrow;{\mathrm{AT}}_i^{\left(p\right)}+w_i+t_{ij}\&le;{\mathrm{AT}}_j^{\left(p\right)}---\left(5\right)$

$e_r\&le;{\mathrm{AT}}_r^{\left(p\right)}\&le;l_r---\left(6\right)$

$\underset{i\&Element;U}{\&Sigma;}\underset{p=1}{\overset{P}{\&Sigma;}}{x}_{iw}^{\left(p\right)}=\underset{j\&Element;U}{\&Sigma;}\underset{p=1}{\overset{P}{\&Sigma;}}{x}_{wj}^{\left(p\right)},w\&Element;(U_2\&cup;H_p^{\left(n\right)})---\left(7\right)$

${\mathrm{AT}}_{H_p^{\left(n\right)\left(z\right)}}^{\left(p\right)}\&le;{\mathrm{AT}}_{H_p^{\left(n\right)(z+1)}}^{\left(p\right)}---\left(8\right)$

$\underset{i\&Element;U}{\&Sigma;}\underset{p=1}{\overset{P}{\&Sigma;}}{x}_{iu}^{\left(p\right)}=\underset{j\&Element;U}{\&Sigma;}\underset{p=1}{\overset{P}{\&Sigma;}}{x}_{uj}^{\left(p\right)}=1,u\&Element;H_p^{\left(n\right)}---\left(9\right)$

$x_{ij}^{\left(p\right)}=0or1---\left(10\right)$

Formula (1) is object function, represents that total distribution cost of described tractor K minimizes；WhenTime, represent that pth is led Draw car K^(p)Straight line path<i, j>between described i-th node and jth node, whenTime, represent that pth is led Draw car K^(p)Without straight line path<i, j>between described i-th node and jth node；

Formula (2) represents pth tractor K^(p)Must be from nodes of locationsSet out；WhenTime, represent pth Tractor K^(p)Through nodes of locationsAnd the path between jth nodeWhenTime, table Show pth tractor K^(p)Without nodes of locationsAnd the path between jth node

Formula (3) represents that pth tractor must arrive destination nodeWhenTime, represent pth tractor K^(p)Warp Cross i-th node and destination nodeBetween pathWhenTime, represent pth tractor K^(p)Without Cross i-th node and destination nodeBetween path

Formula (4) is time windows constraints, represents and makes pth tractor K^(p)In nodes of locationsTime be current time Time；Represent pth tractor K^(p)Arriving at location nodeTime；

Formula (5) is worked as representingDuring establishment, it is thus achieved that pth tractor K^(p)The access time on jth node, Represent pth tractor K^(p)Arrive at the time of i-th node；w_iRepresent pth tractor K^(p)At i-th node Waiting time；t_ijRepresent pth tractor K^(p)From the internodal running time of i-th node to jth；Table Show expression pth tractor K^(p)Arrive at the time of jth node；

Formula (6) is time windows constraints, represents pth tractor K^(p)Arrive the r destination nodeTime access the earliest e_rAccess time l at the latest_rBetween；

Formula (7) represents all road-net node U₂And access sequenceIn destination node meet come in and go out balance；Represent pth Tractor K^(p)Straight line path<i, w>between described i-th node and the w node, whenTime, represent pth Tractor K^(p)Without straight line path<i, w>between described i-th node and the w node；WhenTime, represent pth Tractor K^(p)Through described straight line path<w, j>between w node and jth node, whenTime, represent pth Tractor K^(p)Without described straight line path between w node and jth node；

Formula (8) represents access sequenceThe visit of interior joint time order constrained；DescribedRepresent pth tractor K^(p)Arrive Access sequenceIn time of z destination node；DescribedRepresent pth tractor K^(p)Arrive access sequenceIn time of z+1 destination node；

Formula (9) represents access sequenceIn node, only by one traction train number access once；WhenTime, represent pth Tractor K^(p)Straight line path<i, u>between described i-th node and the u node, whenTime, represent pth Tractor K^(p)Without straight line path<i, u>between described i-th node and the u node；WhenTime, represent pth Tractor K^(p)Through described straight line path between u node and jth node<u,j>, whenTime, represent pth Tractor K^(p)Without described straight line path between u node and jth node<u,j>；

Formula (10) represents that the value of decision variable is " 0 " or " 1 ".