A two-stage responsive dispatching method and system for driverless buses
By using a two-stage responsive scheduling method, combining static and dynamic demand optimization of driverless bus routes, the problems of low operational efficiency and high cost in existing technologies are solved, and efficient and flexible bus system operation is achieved.
Patent Information
- Application Number
- CN202410349272.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-03-26
- Publication Date
- 2024-11-05
- Estimated Expiration
- 2044-03-26
AI Technical Summary
When faced with inaccurate predictions based on historical demand data, existing driverless bus systems struggle to effectively adjust bus routes and schedules, leading to low operational efficiency and increased operating costs. Furthermore, the complexity of existing algorithms and the high hardware requirements reduce the system's convenience and flexibility.
A two-stage response scheduling method is adopted. First, a static demand model is constructed to obtain control stations and main baseline routes. Then, dynamic demand insertion is optimized through a variable neighborhood search algorithm. The bus routes are optimized by combining static and dynamic demand. The variable neighborhood search algorithm is used for iteration and evaluation to generate the optimal driving route.
It improved the operational efficiency of the driverless bus system, reduced total passenger travel time and operating costs, enhanced the system's flexibility and responsiveness to dynamic demands, and improved passenger satisfaction and service quality.
Smart Images

Figure CN118333299B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of intelligent transportation technology, and in particular to a two-stage responsive scheduling method and system for driverless buses. Background Art
[0002] With the rapid development of science and technology and economy, intelligent networked bus technology has been gradually applied in urban transportation systems. As an innovative bus model, flexible bus is leading urban transportation to a more efficient and intelligent direction. The key to this bus model is the comprehensive use of cutting-edge unmanned driving technology, bus communication systems and intelligent traffic management systems to achieve autonomous operation and intelligent scheduling of buses (hereinafter referred to as "buses"). In this way, automatic buses not only improve operational efficiency, but also optimize urban traffic management, providing a feasible and effective solution for modern urban transportation.
[0003] Compared with traditional bus operations, automatic buses show more significant flexibility and convenience. Traditional buses operate according to fixed routes and schedules, which limits their ability to respond to real-time passenger flow changes. This rigid operation mode often leads to problems such as empty buses and poor operating efficiency, which leads to ineffective use of costs and a decline in service satisfaction. These problems, in turn, may affect the public's preference and choice of public transportation. Based on these problems, people began to actively explore flexible bus systems. Based on modern technology, flexible buses can collect passenger demand information on a dedicated app (the full name in English is "Application" and the full name in Chinese is "Application") in real time, so as to quickly and effectively respond to passenger needs, that is, change the operation route while ensuring the user's URT (User Ride Time). Obviously, the introduction of unmanned driving technology can significantly reduce labor costs, conduct collaborative scheduling between multiple vehicles, and make the entire scheduling system more intelligent.
[0004] At present, the method of dynamic and static coordinated scheduling of flexible buses is one of the methods in automatic bus technology. The core of this method is to combine historical demand data with real-time reservation demand data, and optimize the static scheduling plan of flexible buses by designing operating routes and flexible timetables, aiming to improve service quality and operational efficiency. In addition, the method also includes dynamic adjustment of routes and timetables to respond to real-time demand, with the aim of reducing passenger travel time and reducing bus operating costs. However, since this method does not take into account the need to make timely adjustments to bus routes and timetables in case of inaccurate predictions of historical demand data, otherwise the efficiency of bus scheduling system operation will be reduced and operating costs will be increased.
[0005] In the prior art, there is also a method for optimizing bus routes in a dynamic demand-responsive manner. The technical solution of this method is that passengers send passenger demand information including boarding location, alighting location, and required time. After receiving the real-time passenger demand information sent by the passengers, the system first uses a process model based on Markov decision-making to optimize the required bus routes in real time, and then uses an inserted heuristic algorithm and a simulated annealing algorithm to find the optimal route to efficiently respond to passenger needs. However, this method requires the use of a combination of reinforcement learning and heuristic algorithms to optimize real-time demand. The algorithm is relatively complex and may have low computing efficiency during peak demand periods. In addition, in order to support the operation of complex algorithms, the requirements for computing hardware equipment on the bus are high, which increases maintenance costs and reduces convenience.
[0006] There is also an existing technology for a heuristic solution method for the ODBRP (On-Demand Bus Routing Problem), in which the demand-responsive bus routing problem ODBRP is that given a set of buses and a set of passenger demands, it is necessary to assign a set of boarding and alighting stations to each passenger and design a bus route that can meet the needs of each passenger. This problem is an NP-hard problem and is large in scale, and cannot be solved by an exact solution algorithm. At present, the ODBRP problem is mostly solved by a neighborhood search algorithm. Although this algorithm is more flexible and can also be used for large-scale problems, it is suitable for the continuous changes of routes, can effectively find approximate optimal solutions, and is also highly efficient. However, as a kind of heuristic algorithm, the neighborhood search algorithm searches based on certain heuristic rules or experience when looking for the optimal solution to the problem, and cannot guarantee the global optimal solution. Moreover, it is necessary to improve the quality of the search domain as much as possible before using the heuristic algorithm. Summary of the invention
[0007] The purpose of the present invention is to provide a two-stage responsive scheduling system and method for driverless buses, which can improve the operating efficiency of the driverless bus system to a certain extent.
[0008] To achieve the above object, the present invention provides a two-stage responsive scheduling method for driverless buses, which includes:
[0009] Step S1, combining the sequence of static demand and its corresponding demand time, building a first model with the goal of minimizing the sum of distances from static demand to facility points, and obtaining control sites;
[0010] Step S2, based on all the control sites obtained in step S1 in the bus service area, combined with the order of demand time in the static demand, a second model is constructed with the minimum operating cost as the goal, a main reference route is generated, and optional sites are obtained; wherein the main reference route is a driving route composed of the control sites obtained in step S1;
[0011] Step S3, starting from the demand time of the earliest static demand, traverse each static demand within the service range of each control station, search each bus in the warehouse and the control stations and optional stations within the current operation range, check all possible boarding and alighting control stations and optional stations for each static demand, add the control stations and optional stations that meet the conditions of the first preset standard to the current driving route, so as to obtain the first benchmark route; wherein the conditions of the first preset standard include: meeting the static demand, not violating the constraints of bus capacity and time window, and being able to minimize the increase in the total passenger travel time of passengers;
[0012] Step S4, dividing the daily operation time in the historical operation data into several time periods as the historical operation time periods, counting the number of dynamic demands received by each control station and optional station in each historical operation time period, inserting the control stations and optional stations corresponding to the dynamic demands that meet the conditions of the second preset standard into the current driving route to obtain a second benchmark route; wherein the conditions of the second preset standard include: exceeding the total threshold of dynamic demands, not violating the constraints of bus capacity and time window, and being able to minimize the increase in the total passenger travel time;
[0013] Step S5, when receiving dynamic demands during the operation of the bus, the entire driving route is optimized according to the main reference route and each secondary reference route, starting from the demand time of the earliest dynamic demand, each control station and optional station is traversed, and the dynamic demands that meet the conditions of the third preset standard are added to the current driving route in the order of demand time to obtain a third reference route; wherein the conditions of the third preset standard include: being able to minimize the increase in the total passenger travel time;
[0014] Step S6, combining the main reference route obtained in step S2 with the first, second and third reference routes to obtain the vehicle's driving route, using a variable neighborhood search algorithm to optimize, iterate and evaluate, obtain the optimal driving route, and output it.
[0015] Furthermore, the first model includes a first parameter, a first objective function and a first constraint:
[0016] (i) The first parameter includes: the decision variable x that determines whether the static demand b is allocated to the facility point d bdSet to 1 or 0, 1 means that the static demand b is allocated to the facility point f, and 0 means that the static demand b is not allocated to the facility point d;
[0017] Whether facility d is used as a control site decision variable y d Set to 1 or 0, 1 means that facility point d is used as a control site, and 0 means that facility point d is not used as a control site;
[0018] (ii) The first objective function, which is described as Z shown in the following equation (1), obtains the number n and location of the control sites:
[0019]
[0020] Where M is the number of static demand points b, N is the number of facility points d, d bd is the distance between demand point b and facility point d;
[0021] (III) The first constraint condition includes:
[0022] a. Each static demand point b is served by only one facility point d. This constraint is described as formula (2):
[0023]
[0024] b. Enable n facilities. The constraint is described as formula (3):
[0025]
[0026] c. Ensure that there will be no static demand point b where there is no facility point. This constraint is described as formula (4):
[0027] x bd ≤y d (4).
[0028] Furthermore, the second model includes a second parameter, a second objective function and a second constraint:
[0029] (i) The second parameter includes: the decision variable of whether bus k serves arc (i, j) Set to 1 or 0, 1 means bus k serves arc (i, j), 0 means bus k does not serve arc (i, j); arc (i, j) is the path between adjacent control stations i and j on the optimal driving route, i≠j, all arcs (i, j) form edge set A, all control stations form point set V, edge set A and point set V form graph G(V, A);
[0030] (ii) The second objective function is described as W shown in the following formula (5):
[0031]
[0032] Where K is the total number of buses k, c ij is the distance d between control site i and control site j ij Unit transportation cost;
[0033] (III) The second constraint condition includes:
[0034] a. Bus k leaves bus depot 0. This constraint is described as equation (6);
[0035]
[0036] b. Without considering the warehouse, the number of buses k entering and leaving the warehouse at each control station is the same. This constraint is described as formula (7);
[0037]
[0038] In the formula, s and t represent the starting point and end point of the driving path, that is, the bus depot;
[0039] c. Each bus k returns to the end point t of the driving path after completing the delivery. This constraint is described as formula (8);
[0040]
[0041] d. Without considering the warehouse, every static demand b is served, and this constraint is described as equation (9);
[0042]
[0043] e. The constraint on the size relationship between the service start times of two adjacent control stations being served is described as formula (10);
[0044]
[0045] In the formula, start i is the time when the service starts at control site i, start j is the time when the service starts at control site j, time ij is the time it takes for the bus to travel from control station i to control station j, serv i To control the service duration of site i, M is a preset constant;
[0046] f. Without considering the warehouse, the serviceable time window of each control site is not violated [e i ,l i ], the constraint description formula (11);
[0047]
[0048] g. Does not violate the passenger load, this constraint is described as formula (12);
[0049]
[0050] In the formula, q i For each passenger demand at control station i, Q k is the maximum transportation capacity of bus k;
[0051] h. Without considering the warehouse, the time when the service of site i starts will be controlled. i Set to non-negative, the constraint is described as equation (13);
[0052]
[0053] Furthermore, step S5 specifically includes:
[0054] Step 51, inserting the preliminary stations for boarding and alighting for dynamic demand according to the current driving route, and applying the variable neighborhood search algorithm to optimize so as to minimize the increase of the total passenger travel time generated by the driving route;
[0055] Step 52, in the actual operation process, after a certain period of time, if there are some planned reserve stations in the route that has not yet been traveled but are not used by passengers, such stations are called "empty stations" and are deleted;
[0056] Among them, the "variable neighborhood search algorithm" specifically includes:
[0057] a. Station replacement strategy: explore different alighting station locations through the station replacement operator, and select the first option that can minimize the total passenger travel time from the explored possible control stations and optional stations;
[0058] b. Sequence exchange strategy: Based on the first solution, the exchange operator is applied to adjust the order of passengers getting on and off the bus to obtain a second solution that can minimize the total passenger travel time;
[0059] c. Station reinsertion strategy: Based on the second solution, one or more control stations and / or optional stations in the current route are removed through the station reinsertion operator, and then reinserted at other locations in the removed route to find an optimal solution to obtain the optimal solution that minimizes the total passenger travel time.
[0060] The present invention also provides a two-stage responsive dispatching system for driverless buses, which includes:
[0061] A control site acquisition unit is used to combine the static demand and its corresponding demand time sequence, build a first model with the goal of minimizing the sum of distances from the static demand to the facility point, and acquire the control site;
[0062] A main reference route generating unit is used to construct a second model based on all control sites obtained by the control site obtaining unit in the bus service area, in combination with the order of demand time in the static demand, with the minimum operating cost as the goal, to generate a main reference route and obtain selectable sites; wherein the main reference route is a driving route composed of the control sites obtained by the control site obtaining unit;
[0063] A first benchmark route generation unit is used to traverse each static demand within the service range of each control station starting from the demand time of the earliest static demand, search each bus in the warehouse and the control stations and optional stations within the current operation range, check all possible boarding and alighting control stations and optional stations for each static demand, and add the control stations and optional stations that meet the conditions of the first preset standard to the current driving route to obtain the first benchmark route; wherein the conditions of the first preset standard include: meeting the static demand, not violating the constraints of bus capacity and time window, and being able to minimize the increase in the total passenger travel time of passengers;
[0064] A second benchmark route generation unit is used to divide the daily operation time in the historical operation data into a number of time periods as the historical operation time periods, count the number of dynamic demands received by each control station and optional station in each historical operation time period, and insert the control stations and optional stations corresponding to the dynamic demands that meet the conditions of the second preset standard into the current driving route to obtain a second benchmark route; wherein the conditions of the second preset standard include: exceeding the total threshold of dynamic demands, not violating the constraints of bus capacity and time window, and being able to minimize the increase in the total passenger travel time of passengers;
[0065] The third reference route generation unit is used to optimize the entire driving route according to the main reference route and each secondary reference route when receiving dynamic demand during the operation of the bus, starting from the demand time of the earliest dynamic demand, traverse each control station and optional station, and add the dynamic demand that meets the conditions of the third preset standard to the current driving route in the order of demand time to obtain the third reference route; wherein the conditions of the third preset standard include: being able to minimize the increase in the total passenger travel time;
[0066] The optimal driving route acquisition unit is used to combine the main reference route obtained by the main reference route generation unit with the first, second and third reference routes to obtain the vehicle's driving route, use a variable neighborhood search algorithm to optimize, iterate and evaluate, obtain the optimal driving route, and output it.
[0067] Furthermore, the first model includes a first parameter, a first objective function and a first constraint:
[0068] (i) The first parameter includes: the decision variable x that determines whether the static demand b is allocated to the facility point d bd Set to 1 or 0, 1 means that the static demand b is allocated to the facility d, and 0 means that the static demand b is not allocated to the facility d;
[0069] Whether facility d is used as a control site decision variable y d Set to 1 or 0, 1 means that facility point d is used as a control site, and 0 means that facility point d is not used as a control site;
[0070] (ii) The first objective function, which is described as Z shown in the following equation (1), obtains the number n and location of the control sites:
[0071]
[0072] Where M is the number of static demand points b, N is the number of facility points d, d bd is the distance between demand point b and facility point d;
[0073] (III) The first constraint condition includes:
[0074] a. Each static demand point b is served by only one facility point d. This constraint is described as formula (2):
[0075]
[0076] b. Enable n facilities. The constraint is described as formula (3):
[0077]
[0078] c. Ensure that there will be no static demand point b where there is no facility point. This constraint is described as formula (4):
[0079] x bd ≤y d (4).
[0080] Furthermore, the second model includes a second parameter, a second objective function and a second constraint:
[0081] (i) The second parameter includes: the decision variable of whether bus k serves arc (i, j) Set to 1 or 0, 1 means bus k serves arc (i, j), 0 means bus k does not serve arc (i, j); arc (i, j) is the path between adjacent control stations i and j on the optimal driving route, i≠j, all arcs (i, j) form edge set A, all control stations form point set V, edge set A and point set V form graph G(V, A);
[0082] (ii) The second objective function is described as W shown in the following formula (5):
[0083]
[0084] Where K is the total number of buses k, c ij is the distance d between control site i and control site j ij Unit transportation cost;
[0085] (III) The second constraint condition includes:
[0086] a. Bus k leaves bus depot 0. This constraint is described as equation (6);
[0087]
[0088] b. Without considering the warehouse, the number of buses k entering and leaving the warehouse at each control station is the same. This constraint is described as formula (7);
[0089]
[0090] In the formula, s and t represent the starting point and end point of the driving path, that is, the bus depot;
[0091] c. Each bus k returns to the end point t of the driving path after completing the delivery. This constraint is described as formula (8);
[0092]
[0093] d. Without considering the warehouse, every static demand b is served, and this constraint is described as equation (9);
[0094]
[0095] e. The constraint on the size relationship between the service start times of two adjacent control stations being served is described as formula (10);
[0096]
[0097] In the formula, start iis the time when the service starts at control site i, start j is the time when the service starts at control site j, time ij is the time it takes for the bus to travel from control station i to control station j, serv i To control the service duration of site i, M is a preset constant;
[0098] f. Without considering the warehouse, the total serviceable time window of each control site shall not be violated [e i ,l i ], the constraint description formula (11);
[0099]
[0100] g. Does not violate the passenger load, this constraint is described as formula (12);
[0101]
[0102] In the formula, q i For each passenger demand at control station i, Q k is the maximum transportation capacity of bus k;
[0103] h. Without considering the warehouse, the time when the service of site i starts will be controlled. i Set to non-negative, the constraint is described as equation (13);
[0104]
[0105] Furthermore, the third benchmark route generation unit specifically includes:
[0106] A variable neighborhood search algorithm optimization subunit is used to insert a preparatory station for boarding and alighting for dynamic demand according to the current driving route, and to optimize by applying a variable neighborhood search algorithm so as to minimize the increase in the total passenger travel time caused by the driving route;
[0107] The empty station deletion subunit is used in the actual operation process. After a certain period of time, if there are some planned reserve stations on the route that has not been traveled but are not used by passengers, such stations are called "empty stations" and deleted;
[0108] Among them, the "variable neighborhood search algorithm" specifically includes:
[0109] a. Station replacement strategy: explore different alighting station locations through the station replacement operator, and select the first option that can minimize the total passenger travel time from the explored possible control stations and optional stations;
[0110] b. Sequence exchange strategy: Based on the first solution, the exchange operator is applied to adjust the order of passengers getting on and off the bus to obtain a second solution that can minimize the total passenger travel time;
[0111] c. Station reinsertion strategy: Based on the second solution, one or more control stations and / or optional stations in the current route are removed through the station reinsertion operator, and then reinserted at other locations in the removed route to find an optimal solution to obtain the optimal solution that minimizes the total passenger travel time.
[0112] The beneficial effects of the present invention are as follows:
[0113] 1) Since the present invention combines the static demand and its corresponding demand time sequence, the first model is constructed with the goal of minimizing the sum of the distances from the static demand to the facility point, and the control site is obtained. Then, according to all the control sites in the bus service area, combined with the demand time sequence in the static demand, the second model is constructed with the goal of minimizing the operating cost, and the main reference route is generated. This method for adjusting the bus driving route can effectively improve the operating efficiency of the bus system, while reducing the total passenger ride time URT and operating costs.
[0114] 2) Since the present invention receives dynamic demands during the operation of the bus, the entire driving route is optimized according to the main reference route and each secondary reference route, starting from the demand time of the earliest dynamic demand, each control site and optional site are traversed, and the dynamic demand that can increase the total passenger travel time the least is added to the current driving route in the order of demand time to obtain a third reference route, which can better meet the needs of real-time passenger flow, thereby improving the quality and convenience of the overall service, reducing inefficient buses, and improving the flexibility of buses.
[0115] 3) Since the present invention divides the daily operating time in the historical operating data into several time periods as the historical operating time periods, the number of dynamic demands received by each control station and optional station in each historical operating time period is counted, and the buses that meet all conditions of the preset standards give priority to inserting the control stations and optional stations corresponding to the dynamic demands into the current driving route to obtain the second benchmark route. Therefore, the present invention can effectively improve the fault tolerance and responsiveness of the operating system to dynamic demands, effectively utilize current resources, and improve passenger satisfaction. BRIEF DESCRIPTION OF THE DRAWINGS
[0116] Figure 1 It is a flow chart of a two-stage responsive scheduling method for driverless buses according to an embodiment of the present invention. DETAILED DESCRIPTION
[0117] In the drawings, the same or similar reference numerals are used to represent the same or similar elements or elements with the same or similar functions. The embodiments of the present invention are described in detail below in conjunction with the drawings.
[0118] In the description of the present invention, the terms "center", "longitudinal", "lateral", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside" and the like indicate directions or positional relationships based on the directions or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific direction, be constructed and operated in a specific direction, and therefore should not be understood as limiting the scope of protection of the present invention.
[0119] The responsive scheduling method provided in the embodiment of the present invention can be understood as dynamically adjusting the routes and stops of public transportation vehicles based on real-time data and changing passenger demand, aiming to improve the operational efficiency of the public transportation system and is applicable to unmanned public transportation systems.
[0120] Those skilled in the art will know that the reference route is a fixed route planned by the bus operator based on the passenger situation in the current service area. The embodiment of the present invention divides the reference route into a primary reference route and a secondary reference route, wherein the primary reference route is the main route during the bus travel. The secondary reference route is the nearest station and the planned route to which the passenger will go after the dynamic demand appears. When the bus is traveling on the secondary reference route, if it encounters the dynamic demand of the passenger, it deviates from the secondary reference route and provides service, and then returns to the secondary reference route. The dynamic demand is planned to obtain the secondary reference route.
[0121] In view of the different demands, the embodiment of the present invention divides the two stages for driverless buses into a static demand planning stage and a dynamic demand response planning stage. Among them, the static demand planning stage is based on the passenger demand information collected the day before the bus operation, that is, the static demand, and the control sites are planned according to the static demand, and then the main benchmark route is planned. The main benchmark route includes the order of visiting the control sites. Dynamic demand refers to the passenger demand that appears in real time during the operation of the vehicle. The system needs to respond to this demand in real time to meet the passenger demand. In the dynamic demand response planning stage, the demand-responsive bus (Demand-Responsive Transit, DRT) will travel along the benchmark route when it is driving.
[0122] like Figure 1 As shown, the two-stage responsive scheduling method for driverless buses provided by the embodiment of the present invention includes:
[0123] Step S1, combining the order of static demand and its corresponding demand time, constructing the first model with the goal of minimizing the sum of distances from static demand to facility points, and obtaining control sites. The control site is the site that makes the main reference route the shortest, and is also the site where the bus must stop during driving. The demand time is the time required to arrive at the location to serve passengers.
[0124] Step S2, based on all the control sites obtained in step S1 in the bus service area, combined with the order of demand time in the static demand, a second model is constructed with the minimum operating cost as the goal, and an exact solution algorithm is used, or a heuristic algorithm, a shortest path algorithm, etc. can be used under the condition of satisfying the constraints, to generate a main reference route, and selectable sites are also obtained. Among them, the main reference route is a driving route composed of the control sites obtained in step S1. The setting method of the selectable site, for example, a certain number of selectable sites can be set around each control site. The selectable site is a part of the facility position other than the control site, and the position is known. The number can be selected in sequence according to, for example, but not limited to, 25%, 50%, and 75% of the remaining facility position. The selectable site provides favorable conditions for the subsequent determination of the secondary reference route. In the static demand planning stage, passengers need to go to the control site. In the dynamic demand response planning stage, passengers walk to the nearest site, which can be a control site or an selectable site.
[0125] Step S3, starting from the earliest static demand demand time, traverse each static demand within the service range of each control station, use a search method such as a greedy heuristic algorithm, search each bus in the warehouse and the control stations and optional stations within the current operating range, check all possible boarding and alighting control stations and optional stations for each static demand, add the control stations and optional stations that meet the conditions of the first preset standard to the current driving route, so as to obtain the first benchmark route. The conditions of the first preset standard include: meeting the static demand, not violating the constraints of bus capacity and time window, and being able to minimize the increase in the total passenger travel time of passengers.
[0126] Step S4, divide the daily operation time in the historical operation data into several time periods as the historical operation time periods, count the number of dynamic demands received by each control station and optional station in each historical operation time period, and insert the control stations and optional stations corresponding to the dynamic demands that meet the conditions of the second preset standard into the current driving route to obtain the second benchmark route. The conditions of the second preset standard include: exceeding the total threshold of dynamic demands, not violating the constraints of bus capacity and time window, and being able to minimize the increase in the total passenger travel time of passengers.
[0127] The specific numerical setting principles of the total threshold of dynamic demand include: 1. Calculate the variance and average value of historical demand data for each station in different time periods. 2. For stations with small volatility, that is, small variance, use the average value of historical demand. 3. For stations with large volatility, that is, large variance, use 1.25 times the average value to prevent frequent route adjustments and improve system stability.
[0128] Step S5, when receiving dynamic demands during the operation of the bus, the entire driving route is optimized according to the main reference route and each secondary reference route, starting from the demand time of the earliest dynamic demand, each control station and optional station is traversed, and the dynamic demands that meet the conditions of the third preset standard are added to the current driving route in the order of demand time to obtain the third reference route. Among them, the conditions of the third preset standard include: being able to minimize the increase in the total passenger travel time.
[0129] Step S6, combining the main reference route obtained in step S2 with the secondary reference routes obtained in steps S3, S4 and S5, respectively, to obtain the vehicle's driving route, using a variable neighborhood search (VNS) algorithm to optimize, iterate and evaluate, obtain the optimal driving route, and output it.
[0130] In one embodiment, the first model includes first parameters, a first objective function, and a first constraint:
[0131] (i) The first parameter includes: the decision variable x that determines whether the static demand b is allocated to the facility point d bd Set to 1 or 0, 1 means that the static demand b is allocated to the facility point d, and 0 means that the static demand b is not allocated to the facility point d. Among them, the facility location can be understood as the collection of all sites, also refers to the location of the site. Through the decision variable x bd , it can be determined whether the static demand b is allocated to the facility d.
[0132] Whether facility d is used as a control site decision variable y d Set to 1 or 0, 1 means that facility d is used as a control site, and 0 means that facility d is not used as a control site. d , which can ensure that these facilities efficiently serve static demands b.
[0133] (ii) The first objective function is described as Z shown in the following formula (1), which obtains the number n and position of the control sites, where the position is represented by the set {X, Y} to represent the horizontal and vertical coordinates of the control site:
[0134]
[0135] Where M is the number of static demand points b, N is the number of facility points d, d bd is the distance between demand point b and facility point d.
[0136] (III) The first constraint condition includes:
[0137] a. Each static demand point b is served by only one facility point d. This constraint is described as formula (2):
[0138]
[0139] b. Enable n facilities. The constraint is described as formula (3):
[0140]
[0141] c. Ensure that there will be no static demand point b where there is no facility point. This constraint is described as formula (4):
[0142] x bd ≤y d (4).
[0143] In one embodiment, the second model includes second parameters, a second objective function, and a second constraint:
[0144] (i) The second parameter includes: the decision variable of whether bus k serves arc (i, j) Set to 1 or 0, 1 means bus k serves arc (i, j), 0 means bus k does not serve arc (i, j). Arc (i, j) is the path between adjacent control stations i and j on the optimal driving route, i≠j, all arcs (i, j) form edge set A, all control stations form point set V, edge set A and point set V form graph G(V, A).
[0145] (ii) The second objective function is described as W shown in the following formula (5):
[0146]
[0147] Where K is the total number of buses k, c ij is the distance d between control site i and control site j ij Unit transportation cost.
[0148] (III) The second constraint condition includes:
[0149] a. Bus k leaves bus depot 0. This constraint is described as equation (6);
[0150]
[0151] b. Without considering the warehouse, the number of buses k entering and leaving each control station is the same. This constraint is described as formula (7), which is used to ensure that the number of buses entering and leaving each control station is the same;
[0152]
[0153] Where s and t represent the starting point and end point of the driving path, that is, the bus depot, except {s, t} means that the depot is not considered.
[0154] c. Each bus k returns to the end point t of the driving path after completing the delivery. This constraint is described as formula (8), which is used to ensure that each bus k must return to the starting point s after completing the delivery;
[0155]
[0156] d. Without considering the warehouse, every static demand b is served, and this constraint is described as equation (9);
[0157]
[0158] e. The constraint on the relationship between the start time of service of two adjacent control stations being served is described as formula (10);
[0159]
[0160] In the formula, start i is the time to start service at control site i, which corresponds to the demand time of the static demand in step S1. j is the time when the service starts at control site j, time ij is the time it takes for the bus to travel from control station i to control station j, serv i is the service time of control station i. The time it takes for a bus to travel from control station i to control station j plus the time it starts serving at control station i plus the travel time between the two stations will not exceed the start service time of control station j. If the bus does not travel from control station i to control station j, then the M parameter will be added. M is a very large constant that is preset and used as a decision variable When the decision variable is 0, the inequality always holds. When it is 1, it does not affect the inequality, and this method is called the "big M method".
[0161] For example, there is a delivery task that includes two static demands, referred to as A and B below. The service time window of A is from 9:00 to 10:00 a.m., and the service time window of B is from 10:15 to 11:15 a.m. i The travel time from A to B is 15 minutes. AB According to the time window constraint provided by formula (10), if the delivery task is completed, the bus must start A +15+30 or earlier to reach point B. If this delivery task is not completed, then x AB If it is 0, the bus will not go from A to B.
[0162] f. Without considering the warehouse, the serviceable time window of each control site is not violated [e i ,l i ], the constraint description formula (11), that is, the arrival time of the control station i must be after the start time of the time window, and must arrive at one of its possible destination stations before the latest arrival time. The earliest departure time of a passenger at a station will vary depending on the walking time of the passenger from the departure point to the station;
[0163]
[0164] g. Does not violate the passenger load, the constraint is described as formula (12):
[0165]
[0166] In the formula, q i For each passenger demand at control station i, Q k is the maximum transportation capacity of bus k.
[0167] h. Without considering the warehouse, the time when the service of site i starts will be controlled. i Set to non-negative, the constraint is described as equation (13);
[0168]
[0169] In one embodiment, step S3 is used to plan the initial sub-reference route. Each control station and optional station has a sub-reference route within the service range. The demand time of early static demand is given priority, and the station that increases the total passenger travel time the least is given priority, effectively reducing the passenger's travel time. In view of this, to find the optimal boarding and alighting station for each static demand, step S3 preferably adopts a greedy heuristic algorithm, because the greedy heuristic algorithm is more efficient and is conducive to improving the flexibility and response speed of scheduling. Step S3 specifically includes:
[0170] Step 31, sorting the static demands received within the service range of the control site and the optional site. According to the earliest boarding time and the latest alighting time of the passenger included in each static demand, the static demands of the passengers are sorted in order from the earliest boarding time to the largest.
[0171] Step 32, traverse each static demand in the order of the earliest demand time. Use a greedy heuristic algorithm to search each bus and the control station and optional station within the current operating range, and check each possible boarding and alighting control station and optional station of each dynamic demand.
[0172] Step 33: If there is a condition that satisfies the dynamic demand without violating the bus capacity and time window constraints and increases the total passenger travel time of the user to the minimum, the station will be added to the current driving route.
[0173] Step 34, if no feasible solution is found, a new bus is dispatched to serve the dynamic demand.
[0174] In one embodiment, step S5 specifically includes:
[0175] Step 51, inserting the preliminary stations for boarding and alighting for dynamic demand according to the current driving route, and applying the variable neighborhood search algorithm for optimization, so as to minimize the increase in the total passenger travel time of the users generated by the driving route.
[0176] Step 52, in the actual operation process, after a certain period of time, if there are some planned stations in the untraveled route that are not used by passengers, for example, when the vehicle reaches 10:45, there are still 5 minutes to the next station, but there is no dynamic demand from passengers, then the station can be abandoned at this time, and this type of station is called an "empty station" and deleted.
[0177] Among them, the "variable neighborhood search algorithm" specifically includes:
[0178] a. Station replacement strategy: explore different alighting station locations through the station replacement operator to select the first solution that can minimize the total passenger travel time from the possible control stations and optional stations explored. This embodiment evaluates various possible station configurations to determine one or more alternative solutions, each of which aims to minimize the total passenger travel time, and can accurately identify which alternative stations can effectively reduce travel time, so as to select the best solution.
[0179] b. Sequence exchange strategy: Based on the first solution, the exchange operator is applied to adjust the order of passengers getting on and off the bus to obtain a second solution that can minimize the total passenger travel time. This embodiment not only involves the selection of stations, but also includes the optimization of the passenger boarding order, with the aim of further reducing the total passenger travel time. By systematically analyzing different combinations of boarding orders, a solution that can maximize efficiency is determined.
[0180] c. Station reinsertion strategy: Based on the second solution, one or more control stations and / or optional stations in the current route are removed through the station reinsertion operator, and then reinserted at other locations in the removed route to find an optimal solution to obtain the optimal solution that minimizes the total passenger travel time. This embodiment flexibly adjusts the station location to adapt to changes in passenger demand, thereby determining the final solution.
[0181] In the above embodiment, in the variable neighborhood search algorithm, on the one hand, the bus route is optimized by introducing alternative alighting operators, etc., to improve the flexibility and adaptability of the route. On the other hand, the "empty station" processing mechanism is introduced to delete the stations that are planned to exist in the route but are never used, thereby reducing unnecessary travel. On the third hand, the stations for boarding and alighting are carefully selected with the consideration of minimizing the total passenger travel time of users.
[0182] For the variable neighborhood search algorithm in step 5, a neighborhood search algorithm such as a large neighborhood search algorithm may also be used to obtain the optimal driving route.
[0183] In the above embodiment, the variable neighborhood search algorithm will evaluate the current driving route plan after multiple iterations and local searches, mainly considering the total passenger travel time of users corresponding to the calculated route plan. This evaluation process includes all static and dynamic requirements and attempts to find a balance point, that is, to minimize the total passenger travel time while meeting all requirements. After finding this optimal solution, the algorithm can determine which route can best reduce the total passenger travel time of users, thereby improving the efficiency of the entire public transportation system. For example: the output route and timetable are shown in the following table:
[0184]
[0185] The table shows the control sites and optional sites that make up a route f. There are multiple optional sites between two adjacent control sites. If a dynamic demand is generated, the demand can be inserted into the route according to step 5.
[0186] The embodiment of the present invention further provides a two-stage responsive dispatching system for driverless buses, which includes a control site acquisition unit, a main reference route generation unit, a first reference route generation unit, a second reference route generation unit, a third reference route generation unit and an optimal driving route acquisition unit, wherein:
[0187] The control site acquisition unit is used to combine the static demand and its corresponding demand time sequence, build a first model with the goal of minimizing the sum of distances from the static demand to the facility point, and acquire the control site.
[0188] The main reference route generation unit is used to construct a second model based on all control sites obtained by the control site acquisition unit in the bus service area, combined with the order of demand time in the static demand, with the minimum operating cost as the goal, to generate the main reference route and also obtain the optional sites. Among them, the main reference route is a driving route composed of the control sites obtained by the control site acquisition unit.
[0189] The first reference route generation unit is used to traverse each static demand within the service range of each control station starting from the demand time of the earliest static demand, search each bus in the warehouse and the control stations and optional stations within the current operation range, check all possible boarding and alighting control stations and optional stations for each static demand, and add the control stations and optional stations that meet the conditions of the first preset standard to the current driving route to obtain the first reference route. The conditions of the first preset standard include: meeting the static demand, not violating the constraints of bus capacity and time window, and being able to minimize the increase in the total passenger travel time of passengers.
[0190] The second reference route generation unit is used to divide the daily operation time in the historical operation data into several time periods as the historical operation time periods, count the number of dynamic demands received by each control station and optional station in each historical operation time period, and insert the control stations and optional stations corresponding to the dynamic demands that meet the conditions of the second preset standard into the current driving route to obtain the second reference route. The conditions of the second preset standard include: exceeding the total threshold of dynamic demands, not violating the constraints of bus capacity and time window, and being able to minimize the increase in the total passenger travel time of passengers.
[0191] The third reference route generation unit is used to optimize the entire driving route according to the main reference route and each secondary reference route when receiving dynamic demand during the operation of the bus, starting from the demand time of the earliest dynamic demand, traversing each control station and optional station, and adding the dynamic demand that meets the conditions of the third preset standard to the current driving route in the order of demand time to obtain the third reference route. Among them, the conditions of the third preset standard include: being able to minimize the increase in the total passenger travel time.
[0192] The optimal driving route acquisition unit is used to combine the main reference route obtained by the main reference route generation unit with the first, second and third reference routes to obtain the vehicle's driving route, use a variable neighborhood search algorithm to optimize, iterate and evaluate, obtain the optimal driving route, and output it.
[0193] In one embodiment, the third reference route generation unit specifically includes a variable neighborhood search algorithm optimization subunit and an empty station deletion subunit, wherein:
[0194] The variable neighborhood search algorithm optimization subunit is used to insert the preliminary stations for boarding and alighting for dynamic demand according to the current driving route, and to optimize by applying the variable neighborhood search algorithm so as to minimize the increase in the total passenger travel time caused by the driving route.
[0195] The empty station deletion subunit is used in the actual operation process. After a certain period of time, if there are some planned reserve stations on the route that has not yet been traveled but are not used by passengers, such stations are called "empty stations" and deleted.
[0196] Finally, it should be pointed out that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit them. Those skilled in the art should understand that the technical solutions described in the above embodiments may be modified, or some of the technical features thereof may be replaced by equivalents; these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.
Claims
1. A two-stage responsive scheduling method for driverless buses, characterized in that: include: Step S1, combining the sequence of static demand and its corresponding demand time, building a first model with the goal of minimizing the sum of distances from static demand to facility points, and obtaining control sites; Step S2, based on all the control sites obtained in step S1 in the bus service area, combined with the order of demand time in the static demand, a second model is constructed with the minimum operating cost as the goal, a main reference route is generated, and optional sites are obtained; wherein the main reference route is a driving route composed of the control sites obtained in step S1; Step S3, starting from the demand time of the earliest static demand, traverse each static demand within the service range of each control station, search each bus in the warehouse and the control stations and optional stations within the current operation range, check all possible boarding and alighting control stations and optional stations for each static demand, add the control stations and optional stations that meet the conditions of the first preset standard to the current driving route, so as to obtain the first benchmark route; wherein the conditions of the first preset standard include: meeting the static demand, not violating the constraints of bus capacity and time window, and being able to minimize the increase in the total passenger travel time of passengers; Step S4, dividing the daily operation time in the historical operation data into several time periods as the historical operation time periods, counting the number of dynamic demands received by each control station and optional station in each historical operation time period, inserting the control stations and optional stations corresponding to the dynamic demands that meet the conditions of the second preset standard into the current driving route to obtain a second benchmark route; wherein the conditions of the second preset standard include: exceeding the total threshold of dynamic demands, not violating the constraints of bus capacity and time window, and being able to minimize the increase in the total passenger travel time; Step S5, when receiving dynamic demands during the operation of the bus, the entire driving route is optimized according to the main reference route and each secondary reference route, starting from the demand time of the earliest dynamic demand, each control station and optional station is traversed, and the dynamic demands that meet the conditions of the third preset standard are added to the current driving route in the order of demand time to obtain a third reference route; wherein the conditions of the third preset standard include: being able to minimize the increase in the total passenger travel time; Step S6, combining the main reference route obtained in step S2 with the first, second and third reference routes to obtain the vehicle's driving route, using a variable neighborhood search algorithm to optimize, iterate and evaluate, obtain the optimal driving route, and output it.
2. The two-stage responsive dispatching method for driverless buses according to claim 1, characterized in that: The first model includes the first parameters, the first objective function and the first constraints: (i) The first parameter includes: the decision variable x that determines whether the static demand b is allocated to the facility point d bd Set to 1 or 0, 1 means that the static demand b is allocated to the facility d, and 0 means that the static demand b is not allocated to the facility d; Whether facility d is used as a control site decision variable y d Set to 1 or 0, 1 means that facility point d is used as a control site, and 0 means that facility point d is not used as a control site; (ii) The first objective function, which is described as Z shown in the following equation (1), obtains the number n and location of the control sites: Where M is the number of static demand points b, N is the number of facility points d, d bd is the distance between demand point b and facility point d; (III) The first constraint condition includes: a. Each static demand point b is served by only one facility point d. This constraint is described as formula (2): b. Enable n facilities. The constraint is described as formula (3): c. Ensure that there will be no static demand point b where there is no facility point. This constraint is described as formula (4): x bd ≤y d (4)。 3. The two-stage responsive scheduling method for driverless buses according to claim 1, characterized in that: The second model includes second parameters, a second objective function and a second constraint: (i) The second parameter includes: the decision variable of whether bus k serves arc (i, j) Set to 1 or 0, 1 means bus k serves arc (i, j), 0 means bus k does not serve arc (i, j); arc (i, j) is the path between adjacent control stations i and j on the optimal driving route, i≠j, all arcs (i, j) form edge set A, all control stations form point set V, edge set A and point set V form graph G(V, A); (ii) The second objective function is described as W shown in the following formula (5): Where K is the total number of buses k, c ij is the distance d between control site i and control site j ij Unit transportation cost; (III) The second constraint condition includes: a. Bus k leaves bus depot 0. This constraint is described as equation (6); b. Without considering the warehouse, the number of buses k entering and leaving the warehouse at each control station is the same. This constraint is described as formula (7); In the formula, s and t represent the starting point and end point of the driving path, that is, the bus depot; c. Each bus k returns to the end point t of the driving path after completing the delivery. This constraint is described as formula (8); d. Without considering the warehouse, every static demand b is served, and this constraint is described as equation (9); e. The constraint on the relationship between the start time of service of two adjacent control stations being served is described as formula (10); In the formula, start i is the time when the service starts at control site i, start j is the time when the service starts at control site j, time ij is the time it takes for the bus to travel from control station i to control station j, serv i To control the service duration of site i, M is a preset constant; f. Without considering the warehouse, the serviceable time window of each control site is not violated [e i , l i ], the constraint description formula (11); g. Does not violate the passenger load, this constraint is described as formula (12); In the formula, q i For each passenger demand at control station i, Q k is the maximum transportation capacity of bus k; h. Without considering the warehouse, the time when the service of site i starts will be controlled. i Set to non-negative, the constraint is described as equation (13); 4. The two-stage responsive dispatching method for driverless buses according to claim 1, 2 or 3, characterized in that: Step S5 specifically includes: Step 51, inserting the preliminary stations for boarding and alighting for dynamic demand according to the current driving route, and applying the variable neighborhood search algorithm to optimize so as to minimize the increase of the total passenger travel time generated by the driving route; Step 52, in the actual operation process, after a certain period of time, if there are some planned reserve stations in the route that has not yet been traveled but are not used by passengers, such stations are called "empty stations" and are deleted; Among them, the "variable neighborhood search algorithm" specifically includes: a. Station replacement strategy: explore different alighting station locations through the station replacement operator, and select the first option that can minimize the total passenger travel time from the explored possible control stations and optional stations; b. Sequence exchange strategy: Based on the first solution, the exchange operator is applied to adjust the order of passengers getting on and off the bus to obtain a second solution that can minimize the total passenger travel time; c. Station reinsertion strategy: Based on the second solution, one or more control stations and / or optional stations in the current route are removed through the station reinsertion operator, and then reinserted at other locations in the removed route to find an optimal solution to obtain the optimal solution that minimizes the total passenger travel time.
5. A two-stage responsive dispatching system for driverless buses, characterized in that: include: A control site acquisition unit is used to combine the static demand and its corresponding demand time sequence, build a first model with the goal of minimizing the sum of distances from the static demand to the facility point, and acquire the control site; A main reference route generating unit is used to construct a second model based on all control sites obtained by the control site obtaining unit in the bus service area, in combination with the order of demand time in the static demand, with the minimum operating cost as the goal, to generate a main reference route and obtain selectable sites; wherein the main reference route is a driving route composed of the control sites obtained by the control site obtaining unit; A first benchmark route generation unit is used to traverse each static demand within the service range of each control station starting from the demand time of the earliest static demand, search each bus in the warehouse and the control stations and optional stations within the current operation range, check all possible boarding and alighting control stations and optional stations for each static demand, and add the control stations and optional stations that meet the conditions of the first preset standard to the current driving route to obtain the first benchmark route; wherein the conditions of the first preset standard include: meeting the static demand, not violating the constraints of bus capacity and time window, and being able to minimize the increase in the total passenger travel time of passengers; A second benchmark route generation unit is used to divide the daily operation time in the historical operation data into a number of time periods as the historical operation time periods, count the number of dynamic demands received by each control station and optional station in each historical operation time period, and insert the control stations and optional stations corresponding to the dynamic demands that meet the conditions of the second preset standard into the current driving route to obtain a second benchmark route; wherein the conditions of the second preset standard include: exceeding the total threshold of dynamic demands, not violating the constraints of bus capacity and time window, and being able to minimize the increase in the total passenger travel time of passengers; The third reference route generation unit is used to optimize the entire driving route according to the main reference route and each secondary reference route when receiving dynamic demand during the operation of the bus, starting from the demand time of the earliest dynamic demand, traverse each control station and optional station, and add the dynamic demand that meets the conditions of the third preset standard to the current driving route in the order of demand time to obtain the third reference route; wherein the conditions of the third preset standard include: being able to minimize the increase in the total passenger travel time; The optimal driving route acquisition unit is used to combine the main reference route obtained by the main reference route generation unit with the first, second and third reference routes to obtain the vehicle's driving route, use a variable neighborhood search algorithm to optimize, iterate and evaluate, obtain the optimal driving route, and output it.
6. The two-stage responsive dispatching system for driverless buses according to claim 5, characterized in that: The first model includes the first parameters, the first objective function and the first constraints: (i) The first parameter includes: the decision variable x that determines whether the static demand b is allocated to the facility point d bd Set to 1 or 0, 1 means that the static demand b is allocated to the facility d, and 0 means that the static demand b is not allocated to the facility d; Whether facility d is used as a control site decision variable y d Set to 1 or 0, 1 means that facility point d is used as a control site, and 0 means that facility point d is not used as a control site; (ii) The first objective function, which is described as Z shown in the following equation (1), obtains the number n and location of the control sites: Where M is the number of static demand points b, N is the number of facility points d, d bd is the distance between demand point b and facility point d; (III) The first constraint condition includes: a. Each static demand point b is served by only one facility point d. This constraint is described as formula (2): b. Enable n facilities. The constraint is described as formula (3): c. Ensure that there will be no static demand point b where there is no facility point. This constraint is described as formula (4): x bd ≤y d (4)。 7. The two-stage responsive dispatching system for driverless buses according to claim 5, characterized in that: The second model includes second parameters, a second objective function and a second constraint: (i) The second parameter includes: the decision variable of whether bus k serves arc (i, j) Set to 1 or 0, 1 means bus k serves arc (i, j), 0 means bus k does not serve arc (i, j); arc (i, j) is the path between adjacent control stations i and j on the optimal driving route, i≠j, all arcs (i, j) form edge set A, all control stations form point set V, edge set A and point set V form graph G(V, A); (ii) The second objective function is described as W shown in the following formula (5): Where K is the total number of buses k, c ij is the distance d between control site i and control site j ij Unit transportation cost; (III) The second constraint condition includes: a. Bus k leaves bus depot 0. This constraint is described as equation (6); b. Without considering the warehouse, the number of buses k entering and leaving the warehouse at each control station is the same. This constraint is described as formula (7); In the formula, s and t represent the starting point and end point of the driving path, that is, the bus depot; c. Each bus k returns to the end point t of the driving path after completing the delivery. This constraint is described as formula (8); d. Without considering the warehouse, every static demand b is served, and this constraint is described as equation (9); e. The constraint on the relationship between the start time of service of two adjacent control stations being served is described as formula (10); In the formula, start i is the time when the service starts at control site i, start j is the time when the service starts at control site j, time ij is the time it takes for the bus to travel from control station i to control station j, serv i To control the service duration of site i, M is a preset constant; f. Without considering the warehouse, the total serviceable time window of each control site shall not be violated [e i , l i ], the constraint description formula (11); g. Does not violate the passenger load, this constraint is described as formula (12); In the formula, q i For each passenger demand at control station i, Q k is the maximum transportation capacity of bus k; h. Without considering the warehouse, the time when the service of site i starts will be controlled. i Set to non-negative, the constraint is described as equation (13); 8. The two-stage responsive dispatching system for driverless buses according to claim 5, 6 or 7, characterized in that: The third benchmark route generation unit specifically includes: A variable neighborhood search algorithm optimization subunit is used to insert a preparatory station for boarding and alighting for dynamic demand according to the current driving route, and to optimize by applying a variable neighborhood search algorithm so as to minimize the increase in the total passenger travel time caused by the driving route; Empty station deletion subunit, which is used in the actual operation process, if after a certain period of time, there are some planned reserve stations on the route that has not been traveled but are not used by passengers, such stations are called "empty stations" and deleted; Among them, the "variable neighborhood search algorithm" specifically includes: a. Station replacement strategy: explore different alighting station locations through the station replacement operator, and select the first option that can minimize the total passenger travel time from the explored possible control stations and optional stations; b. Sequence exchange strategy: Based on the first solution, the exchange operator is applied to adjust the order of passengers getting on and off the bus to obtain a second solution that can minimize the total passenger travel time; c. Station reinsertion strategy: Based on the second solution, one or more control stations and / or optional stations in the current route are removed through the station reinsertion operator, and then reinserted at other locations in the removed route to find an optimal solution to obtain the optimal solution that minimizes the total passenger travel time.
