1. A shared bicycle dynamic redeployment method giving consideration to both real-time determination and future uncertain information is characterized by comprising the following steps of: the method comprises the following steps:

(1) by sampling historical data, future uncertainty information is estimated. Through deep analysis of historical shared bicycle order data, the fact that order requests of the same station at the same time every day are distributed similarly during a common working day is found, and order samples of different dates are sampled from the historical data.

(2) The method comprises the steps of constructing an offline prediction and Online planning shared bicycle rescheduling model (Online _ VRBR), considering real-time information of the distribution of the shared bicycles and uncertain information of future order requests, wherein the shared bicycle rescheduling model (Online _ VRBR) can be formally represented by the following tuples:

the symbols are explained as follows:

·represents a site set, whereinRepresenting a shared single parking station s;

·representing a collection of transportation vehicles that can be used to balance a shared bicycle at each station, whereinRepresents a vehicle v;

discretizing morning and evening rush hour times into time slicesWherein each time slice Δ may be expressed as a fixed length of time in minutes, such as 10 minutes. All time slices herein are in units of Δ;

q represents a time period for which future uncertainty information is predicted, and Q × Δ represents a time length for which future uncertainty information is predicted. The short prediction time affects the long-term optimization efficiency, and the long prediction time affects the prediction accuracy. Therefore, the size of Q needs to be set reasonably;

·representing the distance travelled between different stations, wherein d_s,s'Representing the time required to travel from station s to station s', wherein for any station s the time distance traveled by the vehicle inside the station is 1, i.e. 1Since the distance between stations is short, in this context, we assume that at any time the distance traveled between stations is fixed;

·C^#representing a collection of different station capacities, i.e. the maximum number of shared vehicles that can be parked, whereinRepresenting sitesThe capacity of (a);

·C^*representing a collection of capacities (i.e. the maximum number of shared vehicles that can be loaded) of different transport vehicles, whereinIndicating transport vehicleThe capacity of (a);

at the current instant, i.e. t is 0, the initial distribution of the shared bicycle isDetermined and known. In sample k (we take the order data case of a certain day as a sample), at any time t>0, number of shared vehicles at station s is

At the current time, i.e. t equal to 0, the number of shared bicycles located on the transport vehicle v is equal toIn sample k, at an arbitrary time t>0 number of shared vehicles on transport vehicle v

·Indicating the distribution of vehicles at the station ifIndicating whether the vehicle v is at station s at time t, otherwise,

f denotes the number of samples (i.e., weekday cases) for the shared bicycle ride request and ride end distribution, where | F | ═ K denotes the sample size.Representing the number of shared bicycle offerings that are ridden by the user at site s in sample k, time slice t,representing the number of user bicycle riding requests on site s in sample k, t time slice;

·P_s,s'representing the expense of a vehicle from station s to station s', including labor costs, transportation costs, etc.

2. The method for shared-bicycle dynamic redeployment with both real-time determination and future uncertain information as defined in claim 1, wherein: according to a given modelFor the current time t, the average of a plurality of historical samples may be optimized using an Integer Programming (IP) model:

online multistage stochastic optimization (OnlinemSS)

For the current time t ═ 0, table 2 formally describes the OnlineMSS framework, where the equations are explained as follows:

equation (1) represents the goal of OnlineMSS, i.e., to satisfy as many long-term user cycling order requests as possible, Q represents the length of time in the future, K represents the number of samples, and the average of order requests for multiple sample users, i.e., the desire to satisfy order service rates in the scenario of uncertain shared bicycle requests.

Constraint (2) ensures that in sample k, at time t, if a vehicle v does not pass station s, it will not load or unload the shared bicycle at that station.

Constraint (3) serving constraints for the ride orders for the sites, ensuring that in sample k, the number of shared bicycle order services per siteNot exceeding the required quantity

Constraints (5) - (18) are respectively constrained for shared single-Vehicle redeployment (BP) and transit Vehicle Routing (Vehicle Routing).

3. The method for shared-bicycle dynamic redeployment with both real-time determination and future uncertain information as defined in claim 1, wherein: considering the distribution of current shared vehicles, the arrival distribution of future shared vehicles, the capacity of shared vehicle stations, the capacity of transport vehicles, and the like, a shared vehicle redeployment constraint (BP) is established, and the specific model is as follows:

sharing a single-vehicle re-deployment (BP) constraint

The constraint equations are explained as follows:

constraints (5-6) number of orders to guarantee successful serviceThe number of sharing single vehicles available on the station s at the time t cannot be more than that;

constraint (7-8) to ensure a balance of the number of shared cycles per station, the number of shared cycles at station s at time t being equal to the number of shared cycles parked at station s at the last time t-1Net number of stops at the end and beginning of riding by the userNumber of shared bicycles with all transport vehicles unloaded onlyAnd (4) summing.

Constraints (9-10) satisfy the capacity constraints of the stations, number of shared cycles unloaded at station sWith the number of the existing shared bicyclesThe sum of which cannot exceed the shared bicycle capacityWhile ensuring the number of shared cycles loaded at station sCan not exceed the number of the shared bicycles on the current station

Constraints (11-12) ensuring a balance of the number of shared vehicles on the transport vehicle, at time tNumber of shared vehicles on vShared bicycle on transport vehicle v equal to last time t-1Number of shared vehicles net loaded with transport vehiclesSumming;

constraints (13-14) to meet transporter capacity constraints, i.e. the number of load sharing units on a transporterCan not exceed the capacity of the transport vehicleConstraints (14) ensure that the transporter capacity constraints are met at the initial time.

4. The method for shared-bicycle dynamic redeployment with both real-time determination and future uncertain information as defined in claim 1, wherein: establishing a Vehicle Routing (VR) constraint, and firstly updating a Routing scheme according to the current timeI.e. the distribution of the transport vehicles at the present moment, andi.e. there is a vehicle v, the current moment is in transit, there is a future moment t>0, vehicle v arrives at station s. The specific planning model is as follows:

vehicle Routing (VR) constraints

The equations are explained as follows:

constraints (15-16) ensure that each vehicle enters at one station and exits at the next time. Constraints (16) ensure that vehicles v (i.e. vehicles exiting from station s at time t) are) Must equal t-d at some time before_s，s'Inflowing from site s' (i.e.). For t equal to 0, sinceWith initial position determination, the constraint can ensure proper flow of the vehicle.

Constraint (17) guarantees that in any sample instance k, all vehicle movements reconfigure the operation and maintenance costs incurred by the single vehicle distribution in budget B^τWithin the range, at the algorithm's very beginning time B^τ＝B⁰The budget is then updated each time the dynamic rescheduling occurs.

The budget (18) guarantees the continuity of the movement of the vehicle v, if v is maintained at the two stations s and s ' at times t and t ' (> t), respectively, the time interval between t and t ' must be long enough to satisfy the transfer of the vehicle v, where M represents a large number.

Background

The goal of shared-bicycle dynamic redeployment is to expect that the number of shared bicycles at any time and any station can meet the demand of riding orders every day. If the riding order requirements of any station at any time are determined and known, the traditional operational research optimization algorithm can well solve the problem of shared bicycle dynamic redeployment. However, the uncertain shared bicycle request information at future time brings new challenges to the problem of dynamic relocation of the shared bicycle. How to take into account the short-term benefits of the real-time determined shared bicycle distribution information and the long-term benefits of future uncertain shared bicycle request information is the core of the problem of dynamic redeployment of the shared bicycles.

The existing solution and its drawbacks: the existing solution includes two ideas: the shared bicycle dynamic redeploying scheme based on the offline mode and the shared bicycle dynamic redeploying scheme based on the real-time distribution information.

(1) Sharing bicycle dynamic redeployment scheme based on offline mode

The working principle: through historical data, shared bicycle riding requirements at different stations and different moments are predicted, an off-line transport vehicle path planning and shared bicycle redeployment method is designed, and shared bicycle service satisfaction rate is optimized.

There is a problem: due to the existence of prediction deviation, the scheme cannot meet the dynamic distribution scene of the shared bicycle, so that the offline shared bicycle redeployment scheme is not feasible. For example, the offline solution requires that the transport vehicle transfer 5 shared vehicles from station a to station B at 9:00 am to meet the high order requirements of station B. However, in the real case, at 9 a.m.: at 00 hours, the number of shared vehicles on station a is less than 5, making the offline algorithm ineffective.

(2) Dynamic redeployment scheme based on real-time sharing of bicycle distribution information

The working principle: the shared bicycle distribution information at the current moment is only concerned, the shared bicycles are transferred by the transport vehicle, and the shared bicycle distribution at different stations at the current moment is balanced, for example, a large shared bicycle transfer accessory with more shared bicycle stations has a station with less shared bicycles.

There is a problem: because the future order demand information is not considered, the shared bicycle is frequently transferred, and the operation and maintenance cost is increased. For example, at the present time, although site a has more shared vehicles and site B has fewer vehicles, it is not necessary to transfer shared vehicles from a to site B because the order demand at site a is much higher than the order demand at site B at the next time.

Disclosure of Invention

The technical problem is as follows: in order to solve the problems in the prior art, the method and the device give real-time shared bicycle distribution and transport vehicle distribution information at the current moment, sample a plurality of shared bicycle riding request and riding end sample cases, estimate the shared bicycle request and riding end distribution information in the future Q time period, design a transport vehicle path planning-shared bicycle redeployment solution, and maximize the number of the long-term shared bicycle requests. All model variables used in the Online _ VRBR can be retrieved from historical data, where Q and Δ can be manually set and adjusted empirically.

The technical scheme is as follows: a shared bicycle dynamic redeployment method giving consideration to both real-time determination and future uncertain information is characterized by comprising the following steps of: the method comprises the following steps:

(1) by sampling historical data, future uncertainty information is estimated. Through deep analysis of historical shared bicycle order data, the fact that order requests of the same station at the same time every day are distributed similarly during a common working day is found, and order samples of different dates are sampled from the historical data.

(2) The method comprises the steps of constructing an offline prediction and Online planning shared bicycle rescheduling model (Online _ VRBR), considering real-time information of the distribution of the shared bicycles and uncertain information of future order requests, wherein the shared bicycle rescheduling model (Online _ VRBR) can be formally represented by the following tuples:

the symbols are explained as follows:

·represents a site set, whereinRepresenting a shared single parking station s;

·representing a collection of transportation vehicles that can be used to balance a shared bicycle at each station, whereinRepresents a vehicle v;

discretizing morning and evening rush hour times into time slicesWherein each time slice Δ may be expressed as a fixed length of time in minutes, such as 10 minutes. All time slices herein are in units of Δ;

q represents a time period for which future uncertainty information is predicted, and Q × Δ represents a time length for which future uncertainty information is predicted. The short prediction time affects the long-term optimization efficiency, and the long prediction time affects the prediction accuracy. Therefore, the size of Q needs to be set reasonably;

·representing the distance travelled between different stations, wherein d_S，S′Representing the time required to travel from station s to station s', wherein for any station s the time distance traveled by the vehicle inside the station is 1, i.e. 1Since the distance between stations is short, in this context, we assume that at any time the distance traveled between stations is fixed;

·C^#representing a collection of different station capacities, i.e. the maximum number of shared vehicles that can be parked, whereinRepresenting sitesThe capacity of (a);

·C^*representing a collection of capacities (i.e. the maximum number of shared vehicles that can be loaded) of different transport vehicles, whereinIndicating transport vehicleThe capacity of (a);

at the current instant, i.e. t is 0, the initial distribution of the shared bicycle isDetermined and known. In sample k (we take the order data case of a certain day as a sample), at any time t > 0, the number of shared bicycles at site s is

At the current time, i.e. t equal to 0, the number of shared bicycles located on the transport vehicle v is equal toIn sample k, at any time t > 0, the number of shared vehicles on transport vehicle v is

·Indicating the distribution of vehicles at stations, e.g.FruitIndicating whether the vehicle v is at station s at time t, otherwise,

f denotes the number of samples (i.e., weekday cases) for the shared bicycle ride request and ride end distribution, where | F | ═ K denotes the sample size.Representing the number of shared bicycle offerings that are ridden by the user at site s in sample k, time slice t,representing the number of user bicycle riding requests on site s in sample k, t time slice;

·P_S，S′representing the expense of a vehicle from station s to station s', including labor costs, transportation costs, etc.

Further, according to a given modelFor the current time t, the average of a plurality of historical samples may be optimized using an Integer Programming (IP) model:

online multistage stochastic optimization (OnlinemSS)

For the current time t ═ 0, table 2 formally describes the OnlineMSS framework, where the equations are explained as follows:

equation (1) represents the goal of OnlineMSS, i.e., to satisfy as many long-term user cycling order requests as possible, Q represents the length of time in the future, K represents the number of samples, and the average of order requests for multiple sample users, i.e., the desire to satisfy order service rates in the scenario of uncertain shared bicycle requests.

Constraint (2) ensures that in sample k, at time t, if a vehicle v does not pass station s, it will not load or unload the shared bicycle at that station.

Constraint (3) serving constraints for the ride orders for the sites, ensuring that in sample k, the number of shared bicycle order services per siteNot exceeding the required quantity

Constraints (5) - (18) are respectively constrained for shared single-Vehicle redeployment (BP) and transit Vehicle Routing (Vehicle Routing).

Further, a shared-bicycle redeployment constraint (BP) needs to be established in consideration of the distribution of current shared bicycles, the arrival distribution of future shared bicycles, the capacity of a shared-bicycle station, the capacity of a transport vehicle, and the like, and the specific model is as follows:

sharing a single-vehicle re-deployment (BP) constraint

The constraint equations are explained as follows:

constraints (5-6) number of orders to guarantee successful serviceThe number of sharing single vehicles available on the station s at the time t cannot be more than that;

constraint (7-8) to ensure a balance of the number of shared cycles per station, the number of shared cycles at station s at time t being equal to the number of shared cycles parked at station s at the last time t-1Net number of stops at the end and beginning of riding by the userNumber of shared bicycles with all transport vehicles unloaded onlyAnd (4) summing.

Constraints (9-10) satisfy the capacity constraints of the stations, number of shared cycles unloaded at station sWith the number of the existing shared bicyclesThe sum of which cannot exceed the shared bicycle capacityWhile ensuring the number of shared cycles loaded at station sCan not exceed the number of the shared bicycles on the current station

Constraints (11-12) ensure that the number of shared vehicles on the transport vehicle is balanced, the number of shared vehicles on transport vehicle v at time tShared bicycle on transport vehicle v equal to last time t-1Number of shared vehicles net loaded with transport vehiclesSumming;

constraint (13-14) to meet the capacity of the transport vehicleQuantity constraints, i.e. number of load sharing cycles on transport vehiclesCan not exceed the capacity of the transport vehicleConstraints (14) ensure that the transporter capacity constraints are met at the initial time.

Further, a transport Vehicle Routing (VR) constraint is established, and firstly, according to a Routing scheme at the current moment, the VR constraint is updatedI.e. the distribution of the transport vehicles at the present moment, andnamely, the vehicle v exists, the current time is in the transportation process, the future time t is larger than 0, and the vehicle v arrives at the station s. The specific planning model is as follows:

vehicle Routing (VR) constraints

The equations are explained as follows:

constraints (15-16) ensure that each vehicle enters at one station and exits at the next time. Constraints (16) ensure that vehicles v (i.e. vehicles exiting from station s at time t) are) Must equal t-d at some time before_S，S′Inflowing from site s' (i.e.). For t equal to 0, sinceWith initial position determination, the constraint can ensure proper flow of the vehicle.

Constraint (17) guarantees that in any sample instance k, all vehicle movements reconfigure the operation and maintenance costs incurred by the single vehicle distribution in budget B^τWithin the range, at the algorithm's very beginning time B^τ＝B⁰The budget is then updated each time the dynamic rescheduling occurs.

The budget (18) guarantees the continuity of the movement of the vehicle v, if v is maintained at two stations s and s ' at times t and t ' (> t), respectively, the time interval between t and t ' must be long enough to satisfy the transfer of the vehicle v, where M represents a large number.

Has the advantages that:

(1) and the united government related departments perform application demonstration in the urban bicycle sharing infrastructure service, so that the life experience of residents is improved. Meanwhile, a new idea is provided for relevant government departments in reducing motor vehicle congestion and reducing urban exhaust emission;

(2) from the perspective of government departments, the operation of a good bicycle sharing system can improve the reports of new capital investment of the government departments; from the perspective of commercial companies, the good shared bicycle system can not only improve the operation income, but also reduce the extra cost caused by the non-standard use of the shared bicycle and the labor cost required by operation and maintenance.

Drawings

FIG. 1 is a block diagram of an implementation of a shared bicycle dynamic redeployment method that takes into account both real-time determinations and future uncertain information.

Detailed Description

The invention will be described in further detail with reference to the following detailed description and accompanying drawings:

the invention relates to a shared bicycle dynamic redeployment method which gives consideration to real-time determination and future uncertain information. As shown in FIG. 1, the input of the present invention is the current time modelThe output is a transporter routing policyAnd loading of each vehicle (i.e. loading of each vehicle)) And unloading (i.e.) The number of shared vehicles. Therefore, it is desirable to develop a shared single-vehicle rescheduling system for implementing the output of the solution. The specific system framework comprises rear-end input, a shared single-vehicle heavy-duty scheduling algorithm and front-end strategy visual output.

According to a given modelWe propose a mixed Integer Programming (IP) model to describe the Online VRBR problem. The variables involved in MILP are shown in Table 1.

Table 1 description of variables

Table 2: online multistage stochastic optimization (OnlinemSS)

For the current time t ═ 0, table 2 formally describes the OnlineMSS framework, where the equations are explained as follows:

equation (1) represents the goal of OnlineMSS, i.e., to satisfy as many long-term user cycling order requests as possible, Q represents the length of time in the future, K represents the number of samples, and the average of order requests for multiple sample users, i.e., the desire to satisfy order service rates in the scenario of uncertain shared bicycle requests.

Constraint (2) ensures that in sample k, at time t, if a vehicle v does not pass station s, it will not load or unload the shared bicycle at that station.

Constraint (3) serving constraints for the ride orders for the sites, ensuring that in sample k, the number of shared bicycle order services per siteNot exceeding the required quantity

Constraints (5) - (18) are respectively constrained for shared single-Vehicle redeployment (BP) and transit Vehicle Routing (Vehicle Routing).

Shared-bicycle dynamic redeployment sub-problem

Table 3 formally describes the shared bicycle redeployment constraint (BP), where the equations are explained below;

constraints (5-6) number of orders to guarantee successful serviceThe number of sharing single vehicles available on the station s at the time t cannot be more than that;

constraint (7-8) to ensure a balance of the number of shared cycles per station, the number of shared cycles at station s at time t being equal to the number of shared cycles parked at station s at the last time t-1Net number of stops at the end and beginning of riding by the userNumber of shared bicycles with all transport vehicles unloaded onlyAnd (4) summing.

Constraints (9-10) satisfy the capacity constraints of the stations, number of shared cycles unloaded at station sWith the number of the existing shared bicyclesThe sum of which cannot exceed the shared bicycle capacityWhile ensuring the number of shared cycles loaded at station sCan not exceed the number of the shared bicycles on the current station

Constraints (11-12) ensure that the number of shared vehicles on the transport vehicle is balanced, the number of shared vehicles on transport vehicle v at time tShared bicycle on transport vehicle v equal to last time t-1Number of shared vehicles net loaded with transport vehiclesAnd (4) summing.

Constraints (13-14) to meet transporter capacity constraints, i.e. the number of load sharing units on a transporterCan not exceed the capacity of the transport vehicleConstraints (14) ensure that the transporter capacity constraints are met at the initial time.

Transport vehicle routing problem

Table 4 describes the Vehicle Routing (VR) sub-problem. Firstly, updating according to the routing scheme at the current momentI.e. the distribution of the transport vehicles at the present moment, andnamely, the vehicle v exists, the current time is in the transportation process, the future time t is larger than 0, and the vehicle v arrives at the station s. The specific formulas are explained as follows;

constraints (15-16) ensure that each vehicle enters at one station and exits at the next time. Constraints (16) ensure that vehicles v (i.e. vehicles exiting from station s at time t) areMust equal some time before t-d_S，S′Inflowing from site s' (i.e.). For t equal to 0, sinceWith initial position determination, the constraint can ensure proper flow of the vehicle.

Constraint (17) guarantees that in any sample instance k, all vehicle movements reconfigure the operation and maintenance costs incurred by the single vehicle distribution in budget B^τWithin the range, at the algorithm's very beginning time B^τ＝B⁰The budget is then updated each time the dynamic rescheduling occurs.

The budget (18) guarantees the continuity of the movement of the vehicle v, if v is maintained at the two stations s and s at times t and t '(> t), respectively, the time interval between t and t' must be long enough to satisfy the transfer of the vehicle v, where M represents a significant number.