CN107239883A

CN107239883A - A kind of dispatching method of Car sharing vehicle

Info

Publication number: CN107239883A
Application number: CN201710333494.9A
Authority: CN
Inventors: 马万经; 刘奇
Original assignee: Tongji University
Current assignee: Tongji University
Priority date: 2017-05-12
Filing date: 2017-05-12
Publication date: 2017-10-10
Anticipated expiration: 2037-05-12
Also published as: CN107239883B

Abstract

The present invention relates to a kind of dispatching method of Car sharing vehicle, comprise the following steps：1) data setting website failure probability is dispatched according to the history of each website and obtains corresponding threshold value；2) the minimum object function as scheduling model of the difference of the vehicle number of each website and threshold value using after the completion of scheduling, and constraints and preferential setting are set up, scheduling model is set up, and for needing the event of re-optimization to carry out the reset of threshold value；3) scheduling model solve and obtain corresponding scheduling strategy.Compared with prior art, the present invention has the advantages that model hypothesis, parameter are less, demarcation is simple, scientific and effective, Optimized model target is reasonable, realizing convenient, solution, rapid, model constraints meets actual requirement.

Description

Vehicle scheduling method of automobile sharing system

Technical Field

The invention relates to the field of vehicle allocation of an automobile sharing system, in particular to a scheduling method of vehicles of the automobile sharing system.

Background

The existing vehicle sharing system vehicle dispatching method mainly comprises an artificial experience judgment method, a static linear programming method and a dynamic random programming method:

1. (threshold-based) artificial experience judgment method

Setting upper and lower thresholds for the number of vehicles at the station according to experience; comparing the current vehicle number with a threshold value to obtain a dispatching requirement; and then the staff judges by themselves to obtain a scheduling scheme.

The disadvantages are as follows: the threshold value setting has no scientific basis; scheduling scheme generation is not optimized.

2. Static linear programming method (based on threshold value and taking minimum cost as target)

Setting upper and lower thresholds for the number of vehicles at the station; scheduling cost between additional sites; and solving the vehicle allocation scheme by utilizing linear programming with the aim of minimum cost.

The disadvantages are as follows: the threshold value setting has no scientific basis; vehicle scheduling is a daily short-term decision; factors related to the scheduling cost, such as the number of employees, are fixed and invariable factors in a short period; so costs are not justified as a target; static models cannot cope with dynamic changes of the system.

3. (reliability-based) dynamic stochastic programming method

Factors such as user requirements and service time are used as random variables, reliability indexes are introduced, and a vehicle allocation scheme is dynamically optimized and solved.

The disadvantages are as follows: in the use of the model, parameter calibration is complex, the requirement on data volume is high, and the problem of data sparsity is easy to exist; stochastic programming solves the problem.

Disclosure of Invention

The invention aims to overcome the defects of the prior art and provides the dispatching method of the automobile sharing system vehicle, which has the advantages of less model hypothesis, fewer parameters, simple calibration, scientificity and effectiveness, reasonable model optimization target, convenient realization, quick solution and model constraint condition according with the practical requirement.

A scheduling method of vehicles of an automobile sharing system comprises the following steps:

1) setting station failure probability according to historical scheduling data of each station and acquiring a corresponding threshold;

2) taking the minimum difference between the number of vehicles at each station after scheduling and the threshold as an objective function of a scheduling model, establishing constraint conditions and priority setting, establishing the scheduling model, and resetting the threshold for the event needing to be re-optimized;

3) and solving the scheduling model to obtain a corresponding scheduling strategy.

In step 2), the objective function of the scheduling model is as follows:

wherein x is_ijα for the number of vehicles to be dispatched between the station requiring to dispatch vehicles to the station j requiring to dispatch vehicles within the optimization period T_i、α_jTo add coefficients to the priority, I_excessTo call out a set of stations for vehicles, I_shortageTo call in a collection of stations for a vehicle.

In the step 2), the priority setting includes site priority and state priority, the site priority is represented by setting a priority additional coefficient, and the state priority is set such that the priority of a station with a full station is higher than that of a station with an empty station.

In the step 2), the constraint conditions include a scheduling amount upper limit constraint, an employee trip chain length constraint, a cruising constraint, a node conservation constraint, a single scheduling distance limit constraint and a feasibility constraint.

The upper limit of the scheduling amount is restricted to the scheduling task number, the number of the scheduled vehicles of each station is ensured to be not more than the vehicles actually required to be scheduled, namely not more than the scheduling requirement, and the expression is as follows:

wherein Veh_iThe current number of vehicles for station i,for the upper threshold of station i,is the lower threshold for station i.

The length constraint of the employee trip chain is as follows:

wherein,for the scheduling task vector of stations i to j,is Euclidean space RⁿBase vector of, Dist_ijDistance between station i and station j, V is the speed of travel, TL_kIs the practical travel length limit value of the employee K, K is the employee number set,the kth dimension of the scheduling task vector for sites i through j, i.e., the number of times the kth worker goes from site i to site j.

The endurance constraint is as follows:

wherein,andrespectively the 1 st maximum driving range and the 2 nd maximum driving range in the vehicle at the station i,is composed ofKth of (1)₁The number of the dimensions is one,is composed ofKth of (1)₂The number of the dimensions is one,for site i to site j₁The distance between the two or more of the two or more,respectively site i to site j₂The first formula shows that the distance of any employee from the station i to the scheduled task at any station is less than the maximum value of the range of the vehicle at the station i. The second formula shows that any two employees from station i, going to any station (i.e. to any j)₁,j₂∈I,k₁,k₂∈ K), less than the sum of the two largest values of range of the vehicle at station i

The node conservation constraint is as follows:

wherein,and (4) status identification for judging whether the employee k is at the station i.

The single scheduling distance limit constraint is as follows:

wherein Dist_maxThe maximum distance limit for a single scheduling task.

The feasibility constraints are:

wherein Fasb_ij(k) Adding feasibility constraints to site i through j employee k.

The purpose of the invention can be realized by the following technical scheme:

compared with the prior art, the invention has the following advantages:

firstly, model hypothesis, less parameters and simple calibration

Only two assumptions (X) are involved in this wear protection_nRandomness of (c) and stability of parameter p), three parameters (optimization period T, full load control probability Prob)_uAnd the idle control probability Prob_l) (ii) a The verification of hypothesis and the calibration of parameters are convenient for the actual system.

Secondly, the science and the effectiveness

In the invention, the concept of 'threshold value' is strictly defined, and a scientific calculation method is provided, so that the non-tightness in the past experience method is eliminated.

Thirdly, optimizing the model target to be reasonable

The optimization in the aspect aims to recover the stations in the network as normal as possible under the condition that long-term factors such as the number of workers are fixed; compared with other methods, the method has the advantage that the system cost is optimized, and the method is more reasonable.

Fourthly, the realization is convenient, the solution is rapid

The target and the constraint of the optimization model are in a linear form, so that the optimization model can be conveniently realized by software, and the linear form model is quickly solved to meet the actual requirement.

Fifthly, the model constraint conditions meet the practical requirements

In the optimization model, the influence of the characteristics of the electric automobile on the generation of the scheduling scheme is considered; the length constraint of the travel chain and the constraint of the battery endurance mileage are particularly considered; for this purpose, special linear methods, namely a vector optimization method and a linear decomposition method, are proposed.

Drawings

FIG. 1 is a flow chart of the method of the present invention.

FIG. 2 is a data flow diagram of the method of the present invention.

Fig. 3 is a schematic diagram of a scheduling scheme in an embodiment.

Detailed Description

The invention is described in detail below with reference to the figures and specific embodiments.

Examples

As shown in FIG. 1, the process of the present invention is as follows:

1) simulating the change of the station state by using a Random Walk Model with Barriers, and calibrating parameters by using historical order data to further calculate the station failure probability;

2) defining a threshold value by the failure probability, and obtaining the scheduling requirement of the station vehicle (the method is called a failure probability control method) in the process;

3) comparing the current state of the station with a threshold value to obtain the scheduling requirement of each station;

4) optimizing and solving a vehicle scheduling scheme and a personnel task allocation scheme by utilizing a network flow model; the aim is to ensure that as many stations in the network as possible work normally under the condition of limited resources;

5) and (4) dynamically operating the scheduling model, terminating the current scheduling task when the system state is changed, and optimizing according to the steps.

As shown in FIG. 2, the data flow of the method of the present invention includes the calculation of scheduling requirements, the generation of scheduling schemes and the parameter calibration.

1) Scheduling demand calculation

Upper (lower) threshold Thrd^U(Thrd^L): defined as the upper (lower) limit of the number of vehicles that the station should hold, so that the probability of the station being full (empty) is less than a given value Prob_u(Prob_l) (ii) a And the maximum (small) value satisfying the condition is shown in the formulas (1) and (2). p is a radical of_zAnd q is_zRespectively representing the probability of full and empty failure of a station within the optimization period T. Prob_uThe station full load control probability is obtained; prob_lThe probability of controlling the station vacancy is a set parameter.

Scheduling requirement Need: the difference between the current vehicle number Veh and the threshold Thrd at the station is shown in formula (3). When the number of vehicles at the station is higher than the upper threshold, Need is the number of vehicles at the station minus the upper threshold, and then Need is greater than zero, which indicates that the vehicles Need to be called out; when the number of vehicles at the station is lower than the lower threshold, Need is the number of vehicles at the station minus the lower threshold, and Need is less than zero at this time, indicating that the vehicles Need to be called.

And taking or returning the station as an event. By X_nIndicating the nth event of the site. When the nth event is returning, X_nTaking the value + 1; when the nth event is a car borrowing, X_nThe value is-1, see formula (5). Let X_nThe probability of taking the value +1 is p, and the probability of taking the value-1 is q ═ 1-p. p is the model parameter that needs to be calibrated with the historical order.

p{X_n＝+1}＝p,p{X_n＝-1}＝1-p (5)

A random loitering model is used to calculate the probability. Let N be the number of expected events within the optimization period T, i.e., the number of steps (epochs) in the random loitering model. Where v is used_z,nAnd u_z,nRespectively representing the probability of full and empty failures occurring at the station just at step n. p is a radical of_z、q_zAnd u_z,nSee equations (6) to (7); by replacing the positions of p and q in the formula (7) with z by (a-z), the formula (8) is obtained, and v is obtained_z,nThe value of (c).

2) Scheduling scheme generation

Optimization variable, x, of the present algorithm_ijAll dispatchers between sites i to j are representedThe total driving times in the optimized period T, when I belongs to the station needing to call out the vehicle (I ∈ I)_excess) And j needs to call the station of the vehicle (j ∈ I)_shortage) Then x_ijRepresenting the total number of scheduled actions between stations i to j, i.e. the total number of vehicles scheduled. X in addition to this_ijThe travel path for the dispatcher to complete the dispatch is represented. By Veh_iRepresenting the number of vehicles at station i at the start of the optimization cycle.

The objective function is to meet the scheduling requirements of all stations as much as possible, that is, to minimize the difference between the number of vehicles and the threshold value of each station after scheduling is completed. When station i is a station requiring a vehicle to be called out,representing the total number of vehicles called out from i;representing the difference between the number of vehicles at the station and the upper threshold value after the task is finished. The value of index j is taken together with I in summation_shortage(number set of stations to which vehicles are to be tuned) because the scheduling model in this section uses Inventory-balancing (Inventory-balancing) scheduling policies. The calculation is similarly performed for the station that needs to call in the vehicle. Summing all stations yields the total difference between the number of vehicles at all stations and the threshold, see equation (9).

The constant term in (9) is removed, the optimization result is not influenced, and a more concise form is obtained as shown in formula (10). The influence of the threshold is reflected in the judgment of the calling-out required ("outside") and calling-in required ("short") sites, and the "upper limit constraint of the scheduling amount" mentioned below.

For an actual system, generally, the sum of scheduling tasks of all sites is too large to be completed; it needs to be determined which scheduling tasks have priority; two priority settings are given here:

i) site prioritization

If some stations are deemed to be more important and need priority to ensure normal status, priority may be given, for example, to some important transportation hub stations or stations with significant marketing significance_iThe method is realized by the formulas (9) and (10).

ii) state priority

The user can not return the vehicle after the station is fully loaded, or the vehicle can not be charged after the station is returned, so that the station is generally considered to be more serious than the station is empty. Attempts to give different priority to full and empty fail control have been made where two more approaches are possible. One is to control the probability Prob of full load of the station_uSetting ground ratio station vacancy control probability Prob_lLower, i.e., less probability of allowing a site to be fully loaded. Alternatively, a person may be required to have to be sent to address the scheduling needs of a fully loaded site, see equation (11).

Six types of constraints are listed in the scheme generation model, see i) -iv).

i) Upper bound of scheduling amount

The number of scheduling tasks should ensure that no more vehicles per station are scheduled than actually needed, i.e., no more scheduling needs. As shown in equations (12) and (13), the left side of the No. greater than or equal to is the number of vehicles at the scheduled station, and the right side is the threshold value of the station.

ii) employee travel chain length constraints

Typically, a dispatcher will start from a dispatch center, work a shift (e.g., 4 hours), and then return to the dispatch center to rest or change shifts. The time from the current moment to the end of the shift, that is, the available time of the dispatcher k, is recorded as TL_k. The algorithm provides a method for constraining the length of the scheduling staff travel chain, and ensures that the constraint is in a linear form (the method is called as a vector optimization method).

The set-up is started in the optimization cycle and has K employees. Each x is_ijThe variable is 'split' into a sum of K variable values, using a K-dimensional vectorAnd (4) showing.The k dimension of (i.e. theThe number of vehicles dispatched by the kth employee from station i to station j is indicated. Under this representation approach, the "tracking" of the paths that implement the individual dispatch personnel can be achieved. Are clearly defined by the definition.Is equal toNamely, formula (14).Representing the station i to station j travel time.The time length of the k-th worker for executing the task from the station i to the station j is represented; if employee k does not pass through the ij road segment, i.e.The corresponding k-th dimension component is zero. Summing i and j then we get the total path length the kth employee has taken in the network. The length needs to be less than or equal to the available time TL of the employee_kSee equation (15). Considering that the number of employees in a real-world system is generally small, a constraint (16) can be added, so that the same employee cannot execute more than one task at a pair of sites in an optimization period, and the model becomes a 0-1 planning model; this allows the model to be solved more quickly. Of course, this constraint may not be used, but only requiredTaking integers, the general integer programming model is obtained.

iii) endurance constraints

When the electric automobile system generates a scheduling scheme, endurance constraint is required to be added to ensure that the generated scheduling task is practical and feasible; ensuring that a situation in which the vehicle is dispatched to half without power does not occur. This requires that the distance of the nth farthest dispatched task at station i is less than the nth largest cruising power of the vehicles at that stationThese conditions are non-linear. The algorithm adopts a group of linear substitutes, and is called as a linear decomposition method. The nth group of constraints requires that the sum of n tasks with the largest station distance is smaller than the sum of the largest n endurance electric quantities. It is to be noted that these conditions are necessarily insufficient conditions; the condition is mainly adopted to solve more rapidly and basically meet the actual requirement. The nth group of constraints actually comprisesAnd (4) constraint. If all the cases are to be considered, then there will be 2ⁿAnd (4) a condition. To avoid an exponential increase in the number of conditions in this set of conditions, it may be simplified to consider only the first few sets of conditions. Generally, scheduling tasks between the same pair of stations are not too many, and only the first two groups or the first three groups of conditions can be considered; this simplification is practical. The case where only the first two sets of conditions are considered includes equations (17) and (18). If a constant is added in the condition, a certain amount of electric quantity remaining after the vehicle is adjusted to a target station can be additionally required; to ensure that the vehicle can be used immediately. It should be noted here that, for simplicity, the charging process of the vehicle in the optimization cycle is omitted in the present model. The vehicle range takes the value at the beginning of the optimization cycle and is constant throughout the optimization cycle.

iv) node conservation constraints

This set of constraints is used to ensure that employees do not "die" from the network. Under the new vector representation method, workers are required to keep conservation at each site and each dimension", see formula (19); i.e., i and k need to be taken through the number set.Indicating the number of entries to the site from other sites.Indicating the number of departures from that site to other sites.

v) single scheduling distance limits

This group constraint is used to avoid some too far distance scheduling. Scheduling over long distances consumes a lot of time and manpower, and in practice, such scheduling is often avoided, which is a limit often added by system operators, see equation (20). Dist_maxAnd (4) a parameter set by the operator, namely an upper limit.

vi) feasibility constraints

The set of constraints is expressed inAdditional constraints. To implement a "vehicle number balance" scheduling strategy, such constraints need to be added, requiring that, in addition to the following station OD pairs, the fasts between other station OD pairs_ij(k) Zero, i.e., no dispatcher is allowed to pass.

Departure Link (Staff department link): connecting an OD pair between a dispatching center from a dispatcher and a site needing to be dispatched (the dispatcher performs a first task before);

scheduling links (I ∈ I) connecting sites needing to be called_excess) And the site to be called (j ∈ I)_shortage) OD pair in between (dispatcher performs one task);

non-scheduling links (Staff rebalaning links) connection needs to be called in (I ∈ I)_shortage) And the site needing to be called out (j ∈ I)_excess) OD pairs in between (the dispatcher is on the road to drive to the next stop to be called);

return links (Staff returning links): connecting the site to which the call is to be made and the dispatch center (the dispatcher returns to the dispatch center).

And solving the network flow model by adopting an integer programming model. The generated scheduling scheme trip chain is shown in fig. 3.

In order to make the model responsive to real-time changes in the system state, a dynamically optimized scrolling method is employed. The events that need to trigger re-optimization are indicated below. In addition to those listed below, the scheduling algorithm may be recalled immediately if there are actually other reasons to re-optimize.

i) The station state is abnormal, such as full load and empty of the station;

ii) the set optimization period ends;

iii) the total change in system status (as measured by total change in vehicle inventory level at the station) reaches a specified value;

iv) employee status changes (e.g., employee shift, new employee addition).

When re-optimizing, the new optimization cycle threshold calculation can have two methods: firstly, performing on-line (on-line) calculation, calling historical order data according to a re-optimization starting time point, and performing model calculation; the method is long in time consumption and not recommended without the support of a high-performance computer; another method is off-line (off-line) calculation; and calculating the threshold corresponding to the T length optimization period started at each time point at intervals of about 30 min.

The meanings of the symbols referred to in the above brief description of the principles are shown in table 1.

TABLE 1 symbology Table

3) Parameter calibration

The setting of the optimization period T should be such that the parameter p is stable during this period, i.e. the assumption of the scheduling requirement generation model holds. The length of T is chosen such that the multi-day sample data for parameter p obeys a normal distribution, and can be estimated using the sample mean. For practical systems, T should generally not be set to less than 2 hours. It is recommended herein that T be set between 3 and 6 hours.

Prob_uAnd Prob_lThe lower the setting, the lower the probability of a site failure is indicated to be allowed, the higher the scheduling requirement. If the setting is too low, scheduling requirements are generated that are not actually possible to accomplish. Too high a setting may result in wasted labor. Analysis gives Prob_uAnd Prob_lThe reasonable value range is 0.3-0.7.

Claims

1. A method for dispatching vehicles of an automobile sharing system is characterized by comprising the following steps:

2. The method as claimed in claim 1, wherein in the step 2), the objective function of the scheduling model is:

<mrow> <mi>min</mi> <mo>{</mo> <munder> <munder> <mo>&Sigma;</mo> <mrow> <mi>i</mi> <mo>&Element;</mo> <msub> <mi>I</mi> <mrow> <mi>e</mi> <mi>x</mi> <mi>c</mi> <mi>e</mi> <mi>s</mi> <mi>s</mi> </mrow> </msub> </mrow> </munder> <mrow> <mi>j</mi> <mo>&Element;</mo> <msub> <mi>I</mi> <mrow> <mi>s</mi> <mi>h</mi> <mi>o</mi> <mi>r</mi> <mi>t</mi> <mi>a</mi> <mi>g</mi> <mi>e</mi> </mrow> </msub> </mrow> </munder> <mrow> <mo>(</mo> <msub> <mi>&alpha;</mi> <mi>i</mi> </msub> <mo>+</mo> <msub> <mi>&alpha;</mi> <mi>j</mi> </msub> <mo>)</mo> </mrow> <msub> <mi>x</mi> <mrow> <mi>i</mi> <mi>j</mi> </mrow> </msub> <mo>}</mo> </mrow>

wherein x is_ijα for the number of dispatched vehicles between station i requiring to dispatch a vehicle to station j requiring to dispatch a vehicle within the optimization period T_i、α_jTo add coefficients to the priority, I_excessTo call out a set of stations for vehicles, I_shortageTo call in a collection of stations for a vehicle.

3. The dispatching method of vehicle sharing system as claimed in claim 2, wherein in step 2), the priority setting includes a station priority and a status priority, wherein the station priority indicates the priority of the station by setting a priority addition coefficient, and wherein the status priority sets the priority of the station with a full station higher than the priority of the station with an empty station.

4. The method according to claim 2, wherein in the step 2), the constraint conditions include a scheduling amount upper limit constraint, an employee trip chain length constraint, a cruising constraint, a node conservation constraint, a single scheduling distance limit constraint and a feasibility constraint.

5. The method as claimed in claim 4, wherein the upper limit of the scheduling amount is constrained to ensure that the number of scheduled vehicles at each station is not more than the number of vehicles actually required to be scheduled, that is, not more than the scheduling requirement, and the expression is:

<mrow> <msub> <mi>Veh</mi> <mi>i</mi> </msub> <mo>-</mo> <munder> <mo>&Sigma;</mo> <mi>j</mi> </munder> <msub> <mi>x</mi> <mrow> <mi>i</mi> <mi>j</mi> </mrow> </msub> <mo>&GreaterEqual;</mo> <msubsup> <mi>Thrd</mi> <mi>i</mi> <mi>U</mi> </msubsup> <mo>,</mo> <mi>f</mi> <mi>o</mi> <mi>r</mi> <mi> </mi> <mi>e</mi> <mi>a</mi> <mi>c</mi> <mi>h</mi> <mi> </mi> <mi>i</mi> <mo>&Element;</mo> <msub> <mi>I</mi> <mrow> <mi>e</mi> <mi>x</mi> <mi>c</mi> <mi>e</mi> <mi>s</mi> <mi>s</mi> </mrow> </msub> </mrow>

<mrow> <msub> <mi>Veh</mi> <mi>i</mi> </msub> <mo>+</mo> <munder> <mo>&Sigma;</mo> <mi>i</mi> </munder> <msub> <mi>x</mi> <mrow> <mi>i</mi> <mi>j</mi> </mrow> </msub> <mo>&le;</mo> <msubsup> <mi>Thrd</mi> <mi>i</mi> <mi>L</mi> </msubsup> <mo>,</mo> <mi>f</mi> <mi>o</mi> <mi>r</mi> <mi> </mi> <mi>e</mi> <mi>a</mi> <mi>c</mi> <mi>h</mi> <mi> </mi> <mi>i</mi> <mo>&Element;</mo> <msub> <mi>I</mi> <mrow> <mi>s</mi> <mi>h</mi> <mi>o</mi> <mi>r</mi> <mi>t</mi> <mi>a</mi> <mi>g</mi> <mi>e</mi> </mrow> </msub> </mrow>

6. The method according to claim 4, wherein the employee travel chain length constraint is:

wherein,for the scheduling task vector of stations i to j,is Euclidean space RⁿBase vector of, Dist_ijIs the distance, V, between station i and station jFor speed of travel, TL_kIs the practical travel length limit value of the employee K, K is the employee number set,the kth dimension of the scheduling task vector for sites i through j, i.e., the number of times the kth worker goes from site i to site j.

7. The method as claimed in claim 6, wherein the endurance constraint is:

wherein,andrespectively the 1 st maximum driving range and the 2 nd maximum driving range in the vehicle at the station i,is composed ofKth of (1)₁The number of the dimensions is one,is composed ofKth of (1)₂The number of the dimensions is one,for site i to site j₁The distance between the two or more of the two or more,respectively site i to site j₂The distance between them.

8. The method according to claim 6, wherein the node conservation constraint is:

9. The method as claimed in claim 8, wherein the single-dispatch distance limit constraint is:

wherein Dist_maxThe maximum distance limit for a single scheduling task.

10. The method of claim 8, wherein the feasibility constraint is: