JP7057471B2 - Transportation systems and methods for passenger and route scheduling - Google Patents

Description

This application relates to the application of transportation and routing. Specifically, the application relates to transport management systems and transport management methods. More specifically, the present application relates to passenger scheduling and routing, especially in multi-stop scenarios.

Recent personal transportation services are shifting from human driving to mobility as a service as a service. Here, the person being transported no longer owns the vehicle, but uses the mobility provided on demand. That is, the service transported from point A to point B at the time of selection is used. Such a process can now be of a private or semi-private nature. A private journey is one in which the transport vehicle is scheduled exclusively for passengers or a group of passengers with whom they belong, and in most cases all passengers want to travel between the same departure and destination. be. The semi-private journey is characterized by the fact that certain, most often small numbers of people are pooled together and share a vehicle. Such a semi-private journey can be between the same points A and B of all passengers currently sharing the vehicle. Alternatively, the vehicle may schedule the itinerary at multiple points that allow passengers to get on and off along the itinerary route.

In order to provide such transportation services, multiple users need to be planned and assigned to multiple vehicles serving the service to the destination desired by the user. Therefore, it is necessary to provide a solution to the problem of routing and schedule allocation. In particular, in this specification, it is necessary to consider the so-called vehicle routing problem (VRP).

VRP can be described as a combinatorial optimization problem that involves finding the optimal route and number of vehicles to visit a series of locations. Several variations of VRP are known and they differ from each other due to constraints. For example, one variant known in the art as the "Traveling Salesman Problem" is related to only one vehicle, but there are no restrictions on the time and order of visiting locations. Another variant known in the art as the "pick-up and delivery problem" is that the boarding location must be in front of the corresponding disembarking location, thus limiting the order of site visits. Another variant known in the art as "Capacitated Vehicle Routing Problem with Time Windows, CVRPTW" is that each vehicle has a capacity attribute and each stroke requirement or Consider the reservation as having a capacity requirement and consider the preferred time window for boarding and disembarking. Most traditional VRP variations limit the vehicle to the start and end of the route at the depot.

The VRP problem is usually formulated as a graph using the aforementioned nodes representing depot, boarding, and getting off locations, and connections representing route segments between contiguous nodes. Solution optimality is described or measured using cost functions. For example, the cost function can be the total distance of all routes, and the best solution can result in the minimum cost or the minimum total route distance. In some application scenarios, the cost can be total uptime. In other scenarios, it may be designed to prioritize a small number of vehicles over total distance. The cost function may also be related to costs such as fuel consumption, or the environmental impact of the use of vehicles and fleets, and is the practice for manipulating a particular vehicle for each distance traveled and / or passengers transported. Cost may also be included. Of course, the cost function can consider not only one that can be appropriately weighted, but also multiple individual costs.

Due to the huge combination search space of the VRP problem, it is computationally difficult to find the best solution globally due to the real computing resources and time constraints. Therefore, VRPs are regularly solved by finding suboptimal solutions with realistic time constraints using efficient combinatorial optimization techniques and meta-heuristics. Metaheuristics, metaheuristic functions, or metaheuristic guide functions (these terms may be used interchangeably) are high-level procedures designed to guide, search, generate, or select heuristics. That is, a partial search algorithm, which can provide a sufficiently good solution to optimization problems, especially if the information is incomplete or incomplete or the computing power is limited. Metaheuristics do not always find the best solution globally for a class of problems. Many metaheuristics practice some form of stochastic optimization, so the solution found depends on the set of random variables generated. In combinatorial optimization, by searching a large set of feasible solutions, metaheuristics can find better solutions with less computational effort than optimization algorithms, iterative methods, or simple heuristics.

For example, a metaheuristic method may guide a neighborhood search method through multiple iterations to derive a solution. At each iteration, the neighborhood search method currently finds a new solution in the search space around the solution, and the metaheuristic method determines if the new solution is acceptable. If acceptable and all constraints are met, the new solution will be the current solution in the next iteration.

The most frequently adopted algorithm for solving combinatorial optimization problems is neighborhood search.

This algorithm repeatedly searches for the best solution locally or globally in the solution space by creating a sequence of current solutions according to specific rules and generating candidate solutions. Candidate solutions are accepted if their cost is lower than the cost of the best current solution and all constraints are met. In general, candidate generation strategies are highly dependent on problem details and constraints. When properly designed, the strategy can greatly simplify constraint compliance checks during the sequential process.

The search space and solution space can be understood in particular as a set of all possible points of an optimization problem (a set of values of selection variables) that satisfy the constraints of the problem.

The search space is the initial set of candidate solutions for a particular optimization problem before the set of candidates is narrowed down. The narrowed set of points that meet the constraints of the problem is sometimes referred to as the solution space. Constraints are conditions of optimization problems that the solution space must meet.

An example of such a neighborhood search is shown in FIG. The search begins with an initial solution that is generated and stored in a memory element such as the solution storage element 1100. At each iteration, the current solution within the solution storage element 1100 is passed to a different local search operator element 1120. All valid local search operators are executed in a loop. The operator uses the current solution to generate a new adjacency solution. All operators use their own algorithms to modify the current solution to generate one or more adjacent solutions. The algorithm is usually simple and does not consider all the constraints defined for the problem, so the generated candidate (adjacent) solution may or may not meet the constraints. All solutions are passed to a filter element 1110 containing one or more filter instructions, which satisfy the constraints and have better objective function values, i.e. provide a better solution than the current solution. Select only the solution you want. If such a solution is found, it is stored in solution storage element 1100 and local search continues using the newly discovered solution as a reference to operator element 1120. The operator elements 1120 are applied sequentially, and the order in which the operators are applied is controlled by the operator sequence 1130. When filter element 1110 finds a new solution, the operator sequence can start from the first operator. If no new solution is found after a complete cycle of all operators, local search will stop.

Illustratively, three types of operator elements 1120 in a local search step, namely "2-Opt", "cross" and "exchange", are described herein.

Using "2-Opt", the operator tries to find an adjacent solution with a better objective value by reversing all possible combinations of subsequences of all routes.

FIG. 12 shows an exemplary embodiment of a neighborhood search method using an operator element that uses the neighborhood search “2-OPT”.

For example, every route contains N + 1 nodes. Where n ₀ is the starting node or location, eg depot, and n _N is the ending node or location, eg another (same or different) depot. The subsequence [n _F , n _L ] of this route includes all the nodes from node F to node L. Here, F indicates the first node of the subsequence, and L indicates the last node of the subsequence. The goal is to thoroughly generate all possible combinations of subsequences that meet the following three conditions:

Condition 1: The first node is before the last node. (F <L: step 1210)
Condition 2: The first node is on the route, but the depot is excluded. (0 <F <N: step 1250)
Condition 3: The last node is on the route, but the depot is excluded. (0 <L <N: step 1230)

The algorithm starts at 1200 and initializes F and L of the current route to 1. This is the node after the start depot of the current route. If condition 1 is met (determination: Yes), the method proceeds to step 1220 to create a new candidate solution by reversing the route subsequence [n _F , n _L ] and then the relevant objective value. evaluate. Steps 1230 to 1260 systematically generate new subsequences, ensuring that the other conditions 2 and 3 are met.

After all subsequences of the current route have been evaluated, step 1270 checks if the current route is the last route. If so, the algorithm ends. Otherwise, the method returns to step 1200 and the steps are repeated for the next route.

Using "cross", operators aim to exchange the starting subsequences of a pair of routes to find adjacent solutions with more objective values.

FIG. 13 shows an exemplary embodiment of a neighborhood search method using an operator element that uses a “cross” for neighborhood search.

For example, for every pair of routes, the first route contains N + 1 nodes. Where n ₀ is the starting location, eg, a depot, and n _N is the ending location, eg, another (same or different) depot. The starting subsequence [n ₁ , n _L ] of this route includes all the nodes from node 1 to node L. Here, L indicates the last node of the subsequence. Similarly, the second route includes an M + 1 node. Where m ₀ is the start location, eg depot, and _mM is the end location, eg, another (same or different) depot. The starting subsequence [m ₁ , m _K ] of this route includes all nodes from node 1 to node K. Here, K indicates the last node of the subsequence. The goal is to thoroughly generate the starting subsequences of all possible pairs of routes and exchange them to evaluate the resulting solution.

This method is started in step 1300 by selecting a new route pair and initializing L and K to 1. Step 1310 creates a candidate solution by exchanging [n ₁ , n _L ] of the first route and [m ₁ , m _K ] of the second route, and evaluates its target value. Step 1320 checks if both L and K have reached the end of their respective routes. If not reached (determination: No), L and K are simultaneously incremented in step 1330 to generate a new pair of starting subsequences and continue from step 1330 for solution generation and evaluation. When reached (judgment: Yes), a new route pair is selected and the entire process is repeated. If step 1340 determines that all possible pairs of routes have been processed, the process terminates.

When using "exchange", the operator attempts to find an adjacent solution with a more objective value by exchanging all possible pairs of nodes on all routes.

FIG. 14 shows an exemplary embodiment of a neighborhood search method using an operator element that uses a neighborhood search “exchange”.

For example, every route contains N + 1 nodes. Where n ₀ is the starting location, eg, a depot, and n _N is the ending location, eg, another (same or different) depot. One pair of nodes on this route is represented by (n _F , n _S ). Here, F indicates the first node of the pair, and S indicates the second node of the pair. The goal is to thoroughly generate all possible pairs of nodes that meet the three conditions.

Condition 1: The first node is in front of the second node. (F <S: step 1410)
Condition 2: The first node is on the route, but the depot is excluded. (0 <F <N: step 1450)
Condition 3: The second node is on the route, but the depot is excluded. (0 <S <N: step 1430)

This method starts at 1400 and initializes F and S of the current route to 1. This is the node after the start depot of the current route. If condition 1410 is met, the method proceeds to step 1420 to create a new candidate solution by exchanging n _F and n _S for routes, and then evaluate the relevant objective values. Steps 1430 to 1460 systematically generate a new pair of nodes, ensuring that the other conditions 2 and 3 are met.

After all pairs of nodes on the current route have been evaluated, step 1470 determines if the current route is the last route. If so, the method ends. Otherwise, the method returns to step 1400 and the steps are repeated for the next route.

One of the classical metaheuristic methods is "simulated annealing". "Pseudo-annealing" is an iterative algorithm based on neighborhood search. Each iteration of the neighborhood search uses a stochastic mechanism that uses parameters (eg, temperature parameters to determine the probability of accepting a candidate solution). Parameters such as temperature start with a high initial value and decrease in the process of iterative neighborhood search according to the cooling scheme chosen. As the parameters decrease, the probability that a candidate solution with higher energy (ie, a worse solution) will be accepted, for example towards a "zero" temperature, decreases. This property of accepting worse solutions in the first iteration allows "pseudo-annealing" to search a wider range of globally optimal solutions.

For routing issues, solution costs are usually defined to be directly proportional to energy and an exponential cooling scheme is used. As the iteration progresses, the probability of accepting a higher cost solution gradually decreases. Better solutions with lower cost values are always acceptable in every iteration.

Another neighborhood search method is "Ruin and Create". "Ruin and Rebirth" destroys some of the roots of the solution in the "Ruin" phase and recreates them in the "Recreate" phase. One "ruin" strategy is called "adjacent-based selection", which selects a random node on the graph until the number of nodes defined in this iteration is removed. Delete a node or node subsequence in descending order of distance from. Subsequent "recreate" phases adapt existing routes or create new routes to consolidate the most recently isolated nodes.

Another "ruin" strategy that is particularly effective in the early stages of iteration is the "full-route selection". This removes all the nodes of one route, thereby reducing the total number of routes. Normally, routes with less node constraints are selected for "ruin". The "recreate" phase then attempts to insert the newly deleted node into the rest of the route.

7 (a)-(c) show an exemplary embodiment of the "ruin and recreate" neighborhood search method.

FIG. 7 (a) shows an intermediate solution that connects a set of nodes in a suboptimal way with three routes intertwined. By adopting the "adjacent-based selection" strategy, a circular region 7010 is defined around the random node 7000.

FIG. 7 (b) shows that the "ruin" process destroyed all connections that were completely contained in the circular region 7010, resulting in some isolated nodes and / or disconnected nodes 7020. It is shown that.

FIG. 7 (c) shows that a "recreate" process was performed to generate a new solution. In this process, the more optimal set of routes is replaced by the ruined routes.

There are known solutions to the problem of dispatch routes, but there is not yet an optimal solution to the problem in all situations. Therefore, there is a need for an improved solution to the vehicle allocation route problem. The present disclosure has been devised in the light of the above considerations.

Most generally, the present disclosure relates to novel solutions for solving vehicle allocation route problems and their application in transportation systems for boarding scheduling.

According to the first aspect of the present disclosure, a computer-implemented method for route calculation in a multi-vehicle environment is provided, wherein the method receives the following steps: a) Reservation data corresponding to the user's desired journey. B) At least one first route data is determined as the current route data corresponding to the reserved data, where each of the at least one first route data is at least one first route. Includes characteristics, c) Corresponds to the reserved data and determines at least one subsequent route data based on the current route data, where each of the at least one subsequent route data has at least one subsequent route data. Including the route characteristics of d) the at least one subsequent route characteristic is compared with the at least one route characteristic of the current route data, and e) based on the comparison, of the at least one subsequent route data. Determine if the quality is improved over the at least one current route data, f) the current route as a subsequent route data used as a route data solution or for the next iteration. A higher quality one is selected from the data and the subsequent route data, g) the route data solution comprises transmitting to at least one of the vehicles as a route solution and to the user as a booking solution. The route characteristics indicate at least the quality of each route data, steps c) to f) are repeated until the defined termination conditions are met, and at least one of the at least one subsequent route data is Determined by using a meta-heuristic function that guides the neighborhood search function, the neighborhood search function is adapted to determine subsequent route data that operates from the current route data, and the search scope of the neighborhood search function. Is set by the parameters of the meta-heuristic function.

According to a second aspect of the present disclosure, a route calculation system in a multi-vehicle environment is provided, said system is a communication element adapted to receive reservation data corresponding to a user's desired journey, said reservation data. A determinant adapted to determine at least one first route data as the current route data corresponding to, each of the at least one first route data having at least one first route. A determinant, including characteristics, a determinant that corresponds to the reserved data and is adapted to determine at least one subsequent route data based on the current route data, the at least one subsequent route data. Each of the determinants, including at least one subsequent route characteristic, a comparison element adapted to compare the at least one subsequent route characteristic with at least one route characteristic of the current route data, said comparison. Based on, as a determinant, as a route data solution, or as a determinant adapted to determine whether the quality of the at least one subsequent route data is improved relative to the at least one current route data. Subsequent route data used for the iteration of the above includes a selection element adapted to select one of the higher quality from the current route data and the subsequent route data, said route characteristic. At least indicating the quality of each route data, determining the subsequent route data, comparing the subsequent route characteristics, determining the improvement in quality, and selecting the improved quality route data is a defined termination condition. Is repeated until is satisfied, and at least one of the at least one subsequent route data is determined by using a meta-heuristic function that guides the neighborhood search function by using the method described above. The search function is adapted to determine the subsequent route data that operates from the current route data, and the search range of the neighborhood search function is set by the meta-heuristic function.

According to a third aspect of the present disclosure, a transport routing system for route scheduling in a multi-vehicle environment is provided, said transport system being adapted to receive booking data from at least one user, at least. A communication element adapted to receive vehicle availability data for one vehicle, at least one processing element for determining at least one route data solution for the vehicle according to the reservation data of the at least one user. The root data solution is determined by performing the methods described in the present disclosure.

According to a fourth aspect of the present disclosure, a computer program product comprising instructions is provided that, when the program is executed by a computer, causes the method described in the present disclosure to be performed.

According to a fifth aspect of the present disclosure, a computer-readable medium containing the computer program product of the present disclosure is provided.

In general, the present disclosure employs a novel form of vehicle routing problem, in particular a combination of a particular metaheuristic function and a particular neighborhood search method. Metaheuristic functions can be used globally on a search space to guide the search function to one or more areas of the search space that may give favorable results with respect to other areas of the search space. Subsequent optimizations can be performed using the neighborhood search function. Such a neighborhood search function can be embodied as, for example, a ruin and recreate search function.

In one embodiment of the invention, the metaheuristic function iterates over two neighborhood search functions using pseudo-annealing. In the early iterations, ruin and reconstruction are used as neighborhood search functions, while in later iterations alternative neighborhood search functions (such as simple local search, for example hill climbing) are used instead. To.

In another embodiment of the invention, the metaheuristic function uses pseudo-annealing to iterate through dynamically constructed ruin and recreate methods. For example, the number of ruined nodes is configured to be directly proportional to the global parameter (eg, the temperature parameter of pseudo-annealing). In other words, the number of nodes to be destroyed, or the radius of destruction, or the area of destruction is large at first, but gradually decreases in the process of iteration.

One of the important characteristics of the ruin and recreate neighborhood search function is the ruin area. In known applications, the catastrophic zone is statically constructed, but the catastrophic zone of the present disclosure is dynamically assigned or modified, particularly depending on parameters such as the temperature parameter of the metaheuristic function. Large areas of ruin favor early iterations designed to facilitate global searches for optimal solutions. However, as the solution improves and approaches the plateau and becomes more or less optimal in the process of iteration, large areas of ruin are ineffective in producing small improvements in the solution. On the other hand, constructing a small ruin zone first can confine the solution to local minima very early in the iterations. You can randomly select an initial node or a seed node to start the ruin and rebuild search function. In addition, multiple ruin and recreate search functions centered around different seed nodes may be executed in parallel and the best results may be accepted.

It should be noted that the pseudo-annealing metaheuristic functions and the ruin and recreate neighborhood search functions described above serve only to illustrate possible embodiments of the invention. It is understood that other metaheuristic functions such as "taboo" metaheuristic functions and neighborhood search functions exhibiting similar properties are also within the scope of the invention. Similar properties can be configured, for example, by a known neighborhood search method or to achieve a certain magnitude of "jump" in the solution space. For example, a metaheuristic function characterized by a "big jump" can be complementarily combined with a neighborhood search function characterized by a "small jump". Configurable methods can be dynamically configured to exhibit different jump characteristics at different stages of the iteration.

The present invention also provides a transportation system for boarding scheduling and user pooling. In the scenarios of the present disclosure involving multiple vehicles and dynamic passenger itinerary assignments, individual routes of vehicles are assigned to passenger itinerary assignments (ie, assigning specific passengers to vehicles according to desired boarding / alighting and disembarking locations). It is changing accordingly. Optimizing a particular mobility as a service offering requires a favorable allocation of passengers to multiple vehicles at a particular point in time. Such a preferred allocation can result in a minimum overall travel time for a particular passenger, or an average minimum overall travel time for all specific passengers. Similarly, the reduction in average travel time is closely linked to the reduction of fuel consumption, especially the fleet consumption of all related vehicles participating in mobility services (such as vehicles currently in operation). This reduces vehicle emissions. Similarly, the total amount of travel distance can be reduced, thereby extending the operating life of the vehicle and reducing maintenance costs.

The systems and methods of the invention relate specifically to transportation services for scheduling passenger pools and itineraries in multiple vehicle scenarios. The user can inquire about the itinerary in advance, and for example, the itinerary from point A to point B can be registered at 8:00 am the next day. The present invention most efficiently handles all or at least most of the requests from users of the transportation service, so that the requested departure or boarding place, the requested destination or disembarkation place, the requested boarding time or time. Schedule user and car allocations, taking into account all such travel inquiries or advance bookings (arriving in advance), depending on slots, requested disembarkation times or time slots, capacity, etc.

The system and method schedule the process, notify the user if the inquiry has been accepted, and / or provide the details of the process that may deviate to some extent from the initial request. For example, boarding locations and times may be slightly modified to accommodate the surrounding journeys. This is, for example, at 8:15 am when the user is not greeted at his / her favorite place at 8:05 am and maintains the requested disembarkation location (area) and the requested disembarkation time (frame). It means that you may be offered to be greeted at a distance of 300m. With such a small change in time and place, it may be possible to further optimize the pool service over a simple itinerary sharing method. This allows more users to be pooled than is possible if the service kept the details of the requested journey more tightly.

The system can optionally generate a plurality of steps presented to the user, from which the user can select the preferred steps. All the steps presented may correspond to the requested steps to some extent and may differ from the requested steps. For example, the first option may present the user with a time that matches the requested boarding location but is slightly earlier than the request, and the second option is within the requested time frame but further away. The itinerary may be provided at the boarding location. The user can then decide whether to accept one of the options offered or reject all options. Further optimization may be achievable depending on the multiple steps previously accepted by the user of the system. For example, vehicle assignments can be changed, i.e., vehicles may be selected for a particular trip, depending on the capacity currently required.

Scheduling by the transportation system can operate in the pre-booking phase where multiple reservations are collected and preprocessed. The data is sent to the system to perform the first VRP calculation to generate an initial set of routes to process the reservation. The system may operate in the on-demand phase if it is in operation, i.e. if the reservation is currently being processed. In this phase, the system continuously receives additional reservations and inserts them into the existing route by repeating the VRP calculation. It is believed that multiple events can trigger updates and therefore change the calculated route. Such events serve users with new reservations, reservation cancellations, groups of new reservations accumulated over time intervals, scheduled calculation events, road closures, road accidents, and / or currently in operation. It can be a breakdown of the vehicle.

However, the present invention can be applied not only to be used in such pre-scheduling environments, but also to schedule incoming strokes to allocate to only one vehicle fleet or currently in operation. Can be done. The present invention can enable vehicle passengers of a bus service, for example, to request boarding and alighting at any bus stop in a particular area of the city, for example by using a dedicated mobile app. Time may not be limited, but during off-peak hours, such as between 11:00 am and 3:00 pm, and after 8:00 pm, when demand is low and fixed-schedule public transport services may lose profitability. In some cases, it may be restricted. Off-peak buses may have lower passenger numbers, so the goal may be to reduce costs by replacing public transport services with on-demand services. Regular buses can continue to operate on the selected route, but may be less frequent. The present invention may allow individual vehicles to be ordered immediately or for vehicles to be pre-booked for a defined time, eg, up to 2 weeks in advance. The fare can be calculated based on a distance-based fare model and may reflect the actual itinerary as a result.

Preferred embodiments of the present invention can be seen as follows, which can be applied to all the technical details described above and below. The methods and systems of the present disclosure can use metaheuristic functions to guide search functions and iterate over candidate solutions with adaptive ruin and reconstruction (ie, pseudo-annealing temperatures as metaheuristic functions, etc.). Use the parameters in to reduce the adjacency range). When the parameter reaches the threshold Ts, the search function switches from adaptive ruin and rebuild to, for example, local neighborhood search, which is probably a 2-opt, cross, and exchange operation, as described here. Use at least one of the children.

The characteristic of the route data, or the route characteristic, can be the quality of the route data. A characteristic or quality indicator can represent or represent that one route data contains a route feature (eg, a node of a route that is more suitable for meeting a particular demand than poor quality route data). Can be understood as any indicator or information element used in. Higher quality routes may be, for example, shorter or faster routes, consume less fuel, or require fewer vehicles to provide a particular service. The route quality indicator may be the number of people serviced by the total number of singles / vehicles, and the higher quality is represented by the increase in the number of passengers relative to the total number of singles / vehicles, thereby providing a specific capacity. Since it can be used more effectively, the amount used will increase.

Reservation data can be understood as any indicator or information element associated with the reservation, in particular to allow the determination of route data. Reservation data may be directly related to the environment (eg, real-life scenarios such as around the city where the service is provided), or the meta used to calculate route data in the context of current methods and systems. It may be data. Therefore, the metadata may be converted data, for example, extracted from the context of the environment while being acquired from the actual reservation data.

The termination condition can be conditional on the quality of the route data being maximized, or at least when the defined threshold is exceeded or met (ie, the route becomes the defined quality or the required quality). Alternatively, the condition is a threshold related to the number of iterations or the time used for the calculation, which must not be exceeded, or the iterations may be stopped thereafter. Similarly, the termination condition can be a condition that takes into account the guide parameters of the metaheuristic function (eg, temperature by pseudo-annealing, or taboo list size for taboo search, and associated thresholds).

According to a further embodiment of the present disclosure, the metaheuristic function may include one of a "pseudo-annealing" search function or a "taboo" search function.

Using one of the metaheuristics may provide one or more preferred local solutions for a global search of the solution space.

According to a further embodiment of the present disclosure, the neighborhood search function includes at least one of a "ruin and rebuild" search function, an adaptive "ruin and rebuild" search function, and a "local neighborhood search" search function. obtain.

One of the neighborhood search functions described above can be used to find preferred solutions in the solution space, guided by metaheuristic functions. In particular, the search function can start with a metaheuristic function that leads to an adaptive "ruin and rebuild" search function.

According to a further embodiment of the present disclosure, the meta-heuristic function may include one of a "pseudo-annealing" search function or a "taboo" search function, and the neighborhood search function is i) adaptive "ruin and recreate". It may include one of a search function, a "local neighborhood search" search function, and an ii) adaptive "ruin and rebuild" search function.

The combination described above provides a preferred solution to the optimization problem in the solution space rather than executing each search function individually. Moreover, the combination described above provides a better solution to the optimization problem within the content of the present disclosure in the solution space than in known forms.

According to a further embodiment of the present disclosure, in i) or iii), each "ruin and rebuild" search function may be used for the first iteration, and the "local neighborhood search" search function may be the first. It can be used to repeat 2 times.

In other words, the method can switch the neighborhood search function during the iteration. The first neighborhood search function (eg, "ruin and rebuild" or adaptive "ruin and rebuild") can be used until certain conditions are met. In the context of pseudo-annealing, the temperature parameter can reach the defined initial temperature value, but in the context of taboo, the parameter can be taboo list size. After reaching the parameter value, the method can then continue the neighborhood search using another, eg, a second neighborhood search function, eg, "local neighborhood search". Alternatively, the order of the neighborhood search functions can be reversed.

According to a further embodiment of the present disclosure, setting a neighborhood search range with a metaheuristic function may include reducing the adjacency range of an adaptive "ruin and recreate" search function. In other words, the ruin zone may depend on the parameter values of the metaheuristic function.

According to a further embodiment of the present disclosure, the metaheuristic function may include parameters that are reduced within the iteration of at least steps c) to e), in particular a temperature parameter or taboo list size. Here, the adjacency range can be reduced depending on the parameters.

In other words, global parameters, such as the temperature parameter or taboo list size of a metaheuristic function, set the adjacency range of the catastrophic region. As the optimization progresses, the parameters gradually decrease, for example, the temperature drops, resulting in a smaller size of the catastrophic region. This provides the ability to reduce the search radius or adjacency range of the neighborhood search function as the optimization progresses, especially reducing the distance between the previous / current solution and the new possible solution being evaluated. Avoid big jumps in the solution space as you approach the end of the metaheuristic function.

According to a further embodiment of the present disclosure, the communication element can be further adapted to transmit the root data solution to at least one of the vehicles as a route solution and to the user as a reservation solution.

According to a further embodiment of the present disclosure, the communication element further provides the route data solution to the user as the route data of the determined route data and / or provides the route data solution to at least one vehicle. Can be adapted.

In other words, one or more discovered solutions may be offered to the user, for example, as a user-selectable offer. Thus, the user can consider different journeys or routes of the journey and choose the one that best matches the user's particular preferences. Two journeys may be provided to the user that are slightly different in terms of boarding location and boarding time, while satisfying the user's overall constraints such as departure and arrival timeframes and time locations. Similarly, when the itinerary is accepted by the user, the assigned vehicle or the driver of the assigned vehicle receives information about the itinerary. The information can be individualized or combined with information such as "carrying three passengers at place A at time B".

According to a further embodiment of the present disclosure, a route data solution comprises a requested boarding place, an assigned boarding place, a requested disembarking place, an assigned disembarking place, a boarding time, a boarding time frame. , Disembarkation time, disembarkation time frame, required capacity, service sequence, assigned vehicle, assigned driver at least one, and vehicle availability data includes driver schedule, average vehicle speed, vehicle capacity, And at least one of the vehicle positions, in particular vehicle capacity is the number of human passengers, the number of animal passengers, the size requirements of the goods to be transported, the wheelchair requirements, the bed requirements, the container size, and the requirements. Can include the size of the storage location.

According to a further embodiment of the present disclosure, the communication element can be adapted to receive road data and use the road data to determine at least one route, in particular the road data. It may include at least one of road network information, traffic service information, road closure information, and traffic information.

In other words, the transportation system may receive information about current road conditions or conditions when scheduling routes / itineraries. For example, roads that are closed for road maintenance may require significant detours, resulting in longer travel times. Similarly, the traffic system may receive information about traffic congestion (eg, due to an accident), reflecting current road conditions. Therefore, the transportation system can utilize the information when determining the route / itinerary. Such information also assigns another vehicle that does not have to pass through the affected traffic area, for example to serve the user, when servicing the current schedule or user, thereby delaying. Can be used by avoiding.

According to a further embodiment of the present disclosure, the transportation system may further include a reservation system, the reservation system may include a reservation application, and the reservation application may be a mobile computing platform, a smartphone, a smart (digital) clock, a tablet, a lap. It may be adapted to receive at least one booking data from a user via a top, short messaging service (SMS), mobile messaging application, centralized booking portal, telephone, and voice calling system.

In other words, the transportation system can use the booking front end to communicate with the users of the system to provide a defined interface for registering itinerary requests. The front end provides, for example, a user interface for display on the device screen, where the user configures the itinerary in a standardized, especially machine-readable manner to facilitate receipt of itinerary requests. Enter.

According to a further embodiment of the present disclosure, the transport system is adapted to at least one of a driver operated vehicle, an autonomous vehicle, a ground vehicle, an aerial vehicle, a surface vehicle, an amphibious vehicle and an amphibious aerial vehicle. Can be made to.

According to a further embodiment of the present disclosure, the determination of route data may further include preprocessing the boarding place and the disembarking place to the permissible boarding place and the permissible disembarking place by the preprocessing element, in particular, in particular. At least some of the permissible boarding and disembarking locations can be associated with a weighting function, from the requested location to locations close to and / or statistically popular with the requested location. Higher weight parameters can be assigned than in distant locations, and the weighting function can be used in determining permissible boarding and disembarking locations.

Such preprocessing may shrink the initial search space or solution space to start with only the variables allowed in the solution space. For example, the desired boarding time of 08:48 may be assigned to a boarding time of 08:50, or a time frame of 08:50 to 08:55, or 08:45 to 09:00. Similarly, any requested boarding location is some preset near the requested boarding location to reduce complexity and avoid multiple short journeys around a particular boarding area. Can be assigned to a boarding location. The requested user parameters (eg, boarding time and boarding location may be weighted according to how well the parameters match the assigned parameters, eg, potential further away from the requested boarding location. A good boarding location may be less weighted than a potential boarding location near the requested boarding location.

According to a further embodiment of the present disclosure, the determination of route data may further include changing the assigned boarding and / or disembarking location of at least one route data.

In other words, the transportation system can adapt the generated route of a potentially already assigned process, eg, a user-accepted process. The system may change routes based on certain parameters, such as current traffic data or external inputs such as new itinerary requests. These itinerary requests can be better handled by another solution in the solution space. At least one modified route may be communicated to the user, accepting or rejecting the modification, and in some cases rejecting the entire journey. Depending on the service contract, the user may be required to accept such changes in itinerary assignments. When a new booking is inserted into an existing route, the route changes or is updated on a regular basis, but retains certain parameters of the existing offer to the currently scheduled user (eg, boarding and alighting locations). In order to do so, additional restrictions may be imposed during processing. Boarding and disembarking times are subject to change, but may need to be within the time frame promised to the offered user.

There are two possible scenarios for updating or changing the assigned location. In the first pre-booking scenario, a plurality of received reservations are processed at about the same time. The offer, that is, the itinerary data determined to be acceptable or serviceable, is sent to the user after all routes have been finalized. If a new solution to a routing problem requires a change in location assignments, the location assignment change is irrelevant to the user, as in the pre-booking scenario this happens before the offer is sent to the user. Is.

In on-demand scenarios where new reservations may be inserted into existing offers (ie, offered / accepted reservations), pre-processing, routing engine, and post-processing need to be filled by routing methods. It imposes additional restrictions, for example, the location to which an offer is assigned may be required to remain unchanged. As a result of offering a new reservation (in addition), updated boarding and boarding times for existing offers may be requested within the promised time frame.

According to a further embodiment of the present disclosure, changing the assigned boarding and / or disembarking location allows the system to identify multiple permissible locations around each assigned location and said permissible. All acceptable routes between locations are identified, the shortest route is determined from said acceptable routes, and the assigned boarding and / or disembarking location is the acceptable route for the shortest route obtained. It may further include being adapted to change location.

In other words, the transportation system can determine the weighting factor by itself when considering changes in the assigned boarding and / or disembarking location. Here, exemplary, the shortest route of the permissible route is used as the weighting factor, but additional considerations can be taken into account and, in general, of the permissible route, such as the time required for the route. Determine the weighting factor.

According to a further embodiment of the present disclosure, the transportation system collects commuting settings including preferred boarding place, preferred disembarking place, preferred boarding time, and preferred disembarking time, and at least one popular place. To include, route patterns may be determined from the commuting settings and assigned boarding and / or disembarking locations may be further adapted to prioritize popular locations.

In other words, the user can provide a set of commuting settings that can be stored by the system in the appropriate database, and this preference is used when determining the user's route. The system may prioritize particularly popular locations, i.e., locations that are frequently used by multiple users or easily accessible by users, and thus particularly have a beneficial impact on the operation of transportation services.

An embodiment illustrating the principles of the invention will be described in more detail with reference to the accompanying figures so that the invention can be understood and further aspects and features thereof can be understood.

FIG. 1 shows an exemplary embodiment of a method that combines a metaheuristic function and a neighborhood search function. FIG. 2 shows an exemplary embodiment of a method according to the present disclosure that uses pseudo-annealing and two neighborhood search functions as metaheuristic functions. FIG. 3 illustrates an exemplary embodiment of a method according to the present disclosure that uses pseudo-annealing and adaptive neighborhood search functions as metaheuristic functions. FIG. 4 shows an exemplary embodiment of the transportation system according to the present disclosure. FIG. 5 shows a modification of an exemplary embodiment of the transport system of FIG. FIG. 6 shows an exemplary embodiment for finding the minimum length route in the transportation system according to the present disclosure. 7a-c show exemplary embodiments of the "ruin and recreate" neighborhood search method. FIG. 8 shows an exemplary embodiment of a method according to the present disclosure that uses a "taboo" search and a neighborhood search function as metaheuristic functions. FIG. 9 illustrates an exemplary embodiment of a method according to the present disclosure that uses a dynamic "taboo" search and neighborhood search function as metaheuristic functions. FIG. 10 illustrates an exemplary embodiment of a method of using adaptive "ruin and recreate" neighborhood search functions and metaheuristic functions that guide local neighborhood search functions according to the present disclosure. FIG. 11 shows an exemplary embodiment of the neighborhood search method. FIG. 12 shows an exemplary embodiment of a neighborhood search method using an operator element that uses the neighborhood search “2-OPT”. FIG. 13 shows an exemplary embodiment of a neighborhood search method using an operator element that uses a “cross” for neighborhood search. FIG. 14 shows an exemplary embodiment of a neighborhood search method using an operator element that uses a neighborhood search “exchange”. FIG. 15 illustrates an exemplary embodiment of the ruined portion of the "ruin and recreate" neighborhood search method according to the present disclosure. FIG. 16 shows an exemplary embodiment of the recreated portion of the "ruin and rebuild" neighborhood search method according to the present disclosure.

FIG. 1 shows an exemplary embodiment of a method that combines a metaheuristic function and a neighborhood search function. In particular, FIG. 1 shows a combination of "pseudo-annealing" and "ruin and reconstruction".

As mentioned above, the "solution" of the vehicle allocation route problem includes a series of routes, one route for each vehicle. Each route further contains a sequence of nodes ordered according to the time they are serviced by the vehicle.

Step 100 sets the initial temperature of the "pseudo-annealing". In addition, step 100 also creates a simple initial solution for the iterative neighborhood search function and records the initial solution as the best solution. One of the first possible solutions is, for example, to arrange all the nodes in chronological order by ride time and assign them to the vehicle fleet with a "greedy insertion" approach. The initial solution must comply with the constraints of the problem. Step 110 then performs the "ruin and recreate" method by ruining some of the last accepted solutions and then recreating a new candidate solution. Step 120 checks whether the candidate solution (either the initial solution or the solution created by the ruin and recreate algorithm) is acceptable. If some constraints are not met, for example, if the boarding time required to meet the planned schedule is not met, this candidate solution will be rejected and a new candidate solution will be based on the last approved solution. Generated. If the initial solution fails the constraint check, the algorithm may try or fail another initial solution algorithm.

If the candidate solution meets all the constraints, step 130 uses a randomly and evenly distributed value within the range [0, 1] to determine the acceptance probability P (T, S, S _best ) of the candidate solution. By calculating, it is subject to further testing. The acceptance probability is an exponential function of the energy difference between the candidate solution energy (energy (S)) and the best solution energy (energy (S _best )) divided by the current temperature T.

If the random value is lower than the acceptance probability, the candidate solution is accepted in step 140, otherwise the candidate solution is rejected and the algorithm continues to step 170. Probabilistically, step 130 is to allow a solution that is worse than the best solution to be accepted as the current solution as part of the pseudo-annealing mechanism for performing a global search during the initial iteration. Please note. As the temperature drops in subsequent iterations, it becomes more difficult to accept a worse solution.

For an accepted solution, step 150 calculates the cost (S), which is the cost of the candidate solution, and compares it to the cost (S _best ), which is the cost of the best solution. If the cost of the candidate solution is lower, the candidate solution replaces the best solution in step 160.

To improve readability, the solution processing steps 120-160 are subsequently encapsulated in the "Temperature Dependent Solution Selection" step 115 in the subsequent figure.

Step 170 lowers the temperature of the system according to the selected cooling method. Step 180 checks for a new temperature in the system. Below the preset minimum T _min , the algorithm stops and presents the last accepted solution as the result / final solution in step 185. Otherwise, the method repeats the steps of ruin and rebuild in step 110.

FIG. 2 illustrates an exemplary embodiment of a method according to the present disclosure that uses pseudo-annealing as a metaheuristic function to guide two neighborhood search functions. In particular, embodiments of the invention are shown that use pseudo-annealing as a metaheuristic function to guide two neighborhood search methods that operate in tandem, namely the ruin and rebuild methods and the local search method.

The functions of steps 200, 210, 215, 220, 230, and 235 reflect steps 100, 110, 115, 170, 180, and 185 of FIG. 1, respectively. Similar to step 130, step 230 checks whether the temperature T is below the threshold T _min . If the condition is true, the method ends and step 235 returns the last solution as the final solution. Otherwise, the process compares the temperature to another threshold T _S in step 240. If the temperature is above _TS and further ruin and re-creation is possible, the process performs the next “ruin and rebuild” iteration in step 210. This can lead to large jumps in the solution space. Otherwise, if the temperature is below _TS , the process switches to the "local search" search function in step 250. This introduces a further improved small jump into the solution space. Such a "local search" search function can be, for example, a hill climbing method or a steepest descent method. Therefore, step 240 satisfies the condition of switching from the first search function (here, exemplified here, ruin and rebuild or adaptive ruin and rebuild) to the second search function (here, exemplary, "local search"). Can be regarded as a step. The solution generated by "local search" is also subjected to temperature dependent processing in step 260, similar to the solution by "ruin and reconstruction" in step 215.

FIG. 3 illustrates an exemplary embodiment of a method according to the present disclosure that uses pseudo-annealing and adaptive neighborhood search functions as metaheuristic functions. In particular, embodiments of the present invention have been demonstrated that use pseudo-annealing as a metaheuristic to guide adaptive ruin and re-creation methods. Adaptive ruin and recreate search functions can be, in particular, ruin and recreate search functions based on the reduction of adjacency ranges. In other words, the adjacency range of adaptive ruin and recreate search functions can decrease over time, depending on the respective parameters of the metaheuristic function, such as the temperature of the pseudo-annealing or the tabu list size of the tabu search. ..

The functions of steps 300, 315, 320, 330, and 335 reflect steps 100, 115, 170, 180, and 185 of FIG. 1, respectively. In addition, the ruin and recreate search function in step 310 is adaptable to global parameters of metaheuristic functions, such as temperature parameters. For example, if "ruin and rebuild" is a neighborhood search function and, as shown in FIGS. 7a-c, the "adjacent-based selection" strategy is adopted, the adaptive "ruin and rebuild" neighborhood search function is Depending on the temperature of the parameter, the number of nodes is reduced to the ruin 7010 in FIG. 7a, and the area of ruin and thus the magnitude of the jump between the old and new solutions in the solution space is gradually reduced.

FIG. 15 shows an exemplary embodiment of the ruined portion of the "ruin and recreate" neighborhood search method according to the present disclosure.

The ruin phase uses an adjacency-based string ruin algorithm, while the recreate phase employs a greedy insertion algorithm. Here, "string" refers to a contiguous set of nodes along the root.

For each iteration of the inclusive method, step 1500 calculates the target number of ruined strings, represented by N, based on the average length of the roots of the current solution. In general, a long average route length means that fewer vehicles are needed, but long strings need to be destroyed.

Step 1510 randomly selects a starting node, also known as a seed node, and a route to which the seed node belongs. Step 1520 calculates the number of consecutive nodes (also known as the ruin string length) along the route to ruin for the selected route. The ruin string length is randomly selected from a predetermined range.

The ruin string is then processed by one of the two possible ruin algorithms set forth in steps 1530 and 1540. The choice is randomly determined for each iteration. Step 1530 performs string destruction that destroys the entire string of nodes. Step 1540 performs split string ruin. This algorithm randomly selects two separate substrings along the selected route. The total length of the substrings is the length of the ruin string determined in step 1520. One of the substrings contains the selected seed node. Usually, two substrings are chosen so that they are closely spaced and one of the substrings is near the end of the route.

Step 1550 checks if the total number of currently ruined strings, indicated by n, has reached the maximum number of target ruined strings. If so, the algorithm ends. If not, a new seed node is randomly selected to continue the ruin process in step 1520. As a selection criterion, it requires that the new seed node is adjacent to the previous seed node, but does not belong to the route processed by the previous ruin algorithm.

FIG. 16 shows an exemplary embodiment of the recreated portion of the "ruin and rebuild" neighborhood search method according to the present disclosure.

Step 1600 sorts the ruined nodes according to their start and end positions, such as depots, and their demand, and requires that nodes far from the depot and nodes in high demand be inserted first. Step 1610 thoroughly attempts all possible insertion points and compares the resulting route lengths to the ruined node to all available routes at the best possible insertion position. Try to insert one by one. In this phase, a random probability factor is introduced to skip some insertion points. Step 1620 repeats the insertion for all the ruined nodes before terminating the method.

FIG. 8 shows an exemplary embodiment of a method according to the present disclosure that uses a "taboo" search and a neighborhood search function as metaheuristic functions.

The principle of "taboo" search as a metaheuristic function is based on inputting the best solutions in the search space recently found in the solution space into a pre-determined size first-in, first-out memory structure called a taboo list. The rules for entering the taboo list, and the size of the taboo list, determine the range of diversification or global reach, i.e. the distance of a "big jump" to a new area of search space. In other words, large memory sizes prevent the selection of the latest optimal solution and facilitate a more global search of the search space. The reverse is also true.

At the start, step 800 prepares the initial solution S and an empty taboo list of predetermined size to facilitate global exploration in the solution space. Step 800 also sets the global best solution, S _best , to this initial solution S. Step 810 searches around the vicinity of S for a solution with a better "cost value" than S, i.e., a better or better solution. The neighborhood search function should exclude all solutions in the taboo list for this particular iteration. If found, this new solution will replace the current solution S.

Step 820 checks whether the new solution S is superior to the global best solution S _best by comparing their cost values. If S has a better cost value, S becomes the new S _best in step 830. Regardless of whether S becomes the new global best solution, S will be inserted into the taboo list at step 840 and will not be considered again in the next few iterations.

Step 850 checks if the conditions for ending the iteration are met. When satisfied, the best global solution is output. Otherwise, the neighborhood search function is restarted in step 810, but this time around the new solution S. A possible termination condition is, for example, whether the cost difference between the best solutions in a row is below the threshold or the maximum number of iterations is reached, whichever comes first.

In one embodiment of the invention, every few search iterations, the taboo list is clipped, the oldest solution in the list is purged, and the neighborhood search direction is changed from global search to local search.

FIG. 9 shows an embodiment of the present invention based on a dynamic taboo list. The functions of steps 900, 910, 920, 930, 940, and 950 reflect steps 800, 810, 820, 830, 840, and 850, respectively, of FIG. At step 960, every few iterations, or at the end of every iteration, the taboo list is deleted and the oldest solution is destroyed. This brings about the same search space reduction effect as the embodiment shown in FIG.

FIG. 10 illustrates an exemplary embodiment of a method of using adaptive "ruin and recreate" neighborhood search functions and metaheuristic functions that guide local neighborhood search functions according to the present disclosure. In particular, embodiments of the invention are shown that use pseudo-annealing as a metaheuristic function to guide two neighborhood search methods that work in tandem: adaptive ruin and recreate methods and local search methods.

The functions of 1000, 1010, 1015, 1020, 1030, 1035, 1040, 1050, and 1060 reflect steps 200, 210, 215, 220, 230, 235, 240, 250, 260, respectively, of FIG. However, instead of the ruin and rebuild function, which is based solely on the current best solution "S", an adaptive ruin and rebuild function is adopted, which is a global parameter of the metaheuristic function, such as the temperature parameter ". Take further into account "T".

Similar to step 230, step 1030 checks whether the temperature T is below the threshold T _min . If the condition is true, the method ends and step 1035 returns the last solution as the final solution. Otherwise, the process compares the temperature to another threshold T _S at step 1040. If the temperature is above _TS and the adaptive ruin and re-creation iterations are possible, the process performs the next "adaptive ruin and re-creation" iteration in step 1010. This can lead to large jumps in the solution space. Depending on the (current) value of the temperature parameter "T", these jumps can be affected by scale or magnitude. In other words, relatively higher temperatures than relatively lower temperatures that limit the jump distance or adjacency range may allow for larger jumps in the solution space.

Otherwise, if the temperature is below _TS , the process switches to the "local search" search function at step 1050. This introduces a further improved small jump into the solution space. Such a "local search" search function is, for example, a hill climbing method or a steepest descent method. Therefore, step 1040 should be regarded as a step that satisfies the condition for switching from the first search function (here, exemplary, adaptive ruin and reconstruction) to the second search function (here, exemplary, "local search"). Can be done. The solution generated by "local search" is also subjected to temperature dependent processing in step 1060, similar to the solution by "adaptive ruin and reconstruction" in step 1015.

It should be noted that the scope of the invention extends to all technically plausible combinations of these embodiments, even if the above embodiments are disclosed individually. For example, in the case of the tabu search-based embodiment shown in FIG. 9, the neighborhood search function in step 910 can be replaced by the adaptive ruin and recreate neighborhood search functions, for example, according to FIG. See FIG. Alternatively, the neighborhood search function of FIG. 9 can be started with a ruin and rebuild neighborhood search function and switched to a simple local search after multiple iterations or when certain conditions are met.

FIG. 4 shows an exemplary embodiment of the transportation system according to the present disclosure. In particular, FIG. 4 shows an exemplary embodiment of ride sharing of the present invention.

The user 4000 submits a reservation to the reservation system 4010 individually or collectively using an electronic device. Bookings are wirelessly accessible via mobile computing platforms (eg smartphones, smart (digital) watches, tablets, laptops, short messaging services (SMS), mobile messaging applications (eg WhatsApp, WeChat)). It can be done via an application, a centralized booking portal, or a manual or automated voice call system.

The reservation system 4000 processes the reservation to generate structured reservation data 4020 for the transportation system 4140. Structured booking data 4020 may include the requested boarding location, the requested disembarking location, the boarding time or disembarking time, and at least a portion of the required capacity.

In the case of a manned vehicle, the vehicle data 4060 and the driver data are registered in the vehicle control system 4070 prior to service driving. When vehicles are servicing users, they also communicate real-time information with the vehicle control system 4070. The vehicle control system 4070 processes vehicle and driver information into the structured vehicle data 4080 of the transport system 4140. Structured vehicle data 4080 may include at least a portion of the driver schedule, average vehicle speed, vehicle capacity, and current position.

Before and during service operation, the routing system 4140 periodically collects at least a portion of road network data, road closure schedules, and traffic data (collectively referred to as road data 4130), or the road network and It may be provided by the Transportation Service Portal 4120 or others that provide such data.

To provide the ride-sharing service of the present invention, the transportation system 4140 is used and at least a portion of the data provided to solve the vehicle routing problem as described above is used. Specifically, an initial solution is generated from structured reservation data 4020, structured vehicle data 4080, and optionally road data 4090. The computer implementation method for route calculation and optimization shown in FIG. 2 or 3 is then used to improve the initial solution over multiple iterations to generate the optimal set of routes. Each route data includes at least a linked list of nodes, vehicle information, and / or driver information for each route. Each node contains at least the offered boarding time and the offered disembarking time.

The subset of optimal route information 4030 described above is delivered to the reservation system 4010 and then sent to the user with the current reservation 4050. Illustratively, the transmitted information 4050 includes at least the offer / rejected status and, if offered, the offered boarding time, the offered disembarking time, the vehicle registration number, and optionally driver information. ..

A further subset of the optimal route information 4030 described above is delivered to the vehicle control system 4070 and then transmitted to the vehicle or driver 4110. Illustratively, the transmitted information 4110 includes at least all the nodes of the optimized route associated with the vehicle.

The above embodiment presupposes that any requested boarding and alighting place 4020 is a legally permissible service place. In practice, traffic laws and regulations may limit the location of services to, for example, existing bus stops and taxi stands. Given the requested location, there may be multiple acceptable locations within the acceptable walking distance.

Alternative locations can be selected based on a simple "shortest walking distance" criterion. Based on the selected + alternative acceptable location, the routing engine may generate a set of routes / offers optimized according to cost value. Each route / offer consists of a combination of selected locations and alternative acceptable locations. For example, one route / offer may have a selected permissible ride and an alternative permissible disembarkation. At this stage, the user can select one of the routes (or offers) or reject all of them.

FIG. 5 shows a modification of an exemplary embodiment of the transport system of FIG. To solve the above real problem, the modification of the last embodiment further includes the node pre-processing element 5150 and the node post-processing element 5180 to calculate the optimal route for approximation.

For each requested location 5020, the node preprocessing element 5150 selects a set of locations 5160 allowed for the transport system 5140. The permissible set of locations may consist of one or more locations. As a first variation of this embodiment, the choice can be based on a location that includes the shortest walking distance. As a second variation, the choice can be based on popular locations according to the transportation pattern. As a third variation, both locations can be selected.

For every 5170 assigned location, the node post-processing element 5180 allocates an alternative assigned location, for example, based on other criteria. As a first variation, alternative positions for each node of the optimized route are simultaneously detected to meet the minimum route length goal. As a second variation of this embodiment, the alternative assigned location can be a popular location according to the transport pattern.

FIG. 6 shows an exemplary embodiment for finding the minimum length route in the transportation system according to the present disclosure.

Areas A, B, and C, represented by circles in FIG. 6, are locations initially requested by the user of transport system 4140, 5140. The permissible places are represented by triangles: A ₁ to A ₃ are permissible places around A, B ₁ to B ₂ are permissible places around B, and C ₁ to C ₃ are around C. Is an acceptable place. If the node preprocessing element 5150 determines the boarding position according to the shortest walking distance reference, A ₁ , B ₁ and C ₁ become the permissible nodes 5160 output by the transport system 5140. This is because these boarding locations are closest to their respective requested locations 5000. Therefore, route A ₁ → B ₁ → C ₁ constitutes the initial optimum route output by the routing system of 5170.

Node post-processing element 5180 identifies all allowed routes between the requested locations. Based on the example in Figure 6, there are six permissible routes from A to B: A ₁ B ₁ , A ₁ B ₂ , A ₂ B ₁ , A ₂ B ₂ , A ₃ B ₁ , and A ₃ B ₂ . There are also six permissible routes from B to C: B ₁ C ₁ , B ₁ C ₂ , B ₁ C ₃ , B ₂ C ₁ , B ₂ C ₂ and B ₂ C ₃ . Theoretically, there are 6 × 6 = 36 possible routes from A to C. However, some routes can be deleted in advance, for example, based on time frame constraints. Dynamic programming can be called to find the shortest route from A to C. In this example, the shortest total route of the vehicle can be visually recognized as A ₂ → B ₂ → C ₂ . As a result, A ₂ , B ₂ , and C ₃ become acceptable alternatives to replace A ₁ , B ₁ , and C ₁ , respectively.

It is understood that all system components described herein can be hosted on multiple computers. In addition, they can be centralized or distributed across multiple geographic locations. The functionality of the component can also be handled by different computers or different parties. For example, in the case of vehicle control system 4070, vehicle and driver registration can be processed by one set of computers, while real-time communication can be processed by another set of computers. As another example, in the case of the road network and the traffic service portal 4120, the road network may be cached in a local database, but the traffic information may be obtained from the portal of an outsider.

Similarly, transport systems 4140, 5140 are implemented in a single computing unit further comprising reservation systems 4010, 5010, vehicle control systems 4070, 5070, and node pre-processing elements 4150, 5150 and node post-processing elements 4180, 5180. Can be done.

In the following, the technical concepts of the present disclosure will be described in different ways, using different languages. The present invention relates to a method for calculating optimal routes for multiple vehicles to serve multiple reservations, which method generates an initial route for processing reservations and is currently in solution space. To improve a route with multiple iterations by searching for an improved route near the route, to specify the improved route as the current route, and to adaptively change the neighborhood search range for each iteration. including. "Multiple iterations" specifically refers to the iteration loop of a meta-heuristic function, and "modification" refers to the adjacency range of a neighborhood search function that is embodied as a switch to a local search function or a neighborhood search function, or as ruin and reconstruction. Can be understood as a reduction of. If you exceed the time allotted to search for more solutions, if you cannot find a better solution than the previous solution with a defined number of iterations, or if the solution has, for example, a satisfactory / acceptable value. When reached, i.e., corresponding to a particular threshold, the iteration may end. The search is repeated unless the termination condition is met. Iterative methods include, but are not limited to, pseudo-annealing. Neighborhood searches, on the other hand, include, but are not limited to, ruin and recreate.

Further, the step of changing the neighborhood search range may include reducing the neighborhood of the ruin and recreating the search function. In addition, the step of changing the neighborhood search range may use the ruin and rebuild search functions for some iterations and the local neighborhood search methods for other different iterations. In addition, the adjacency range can be reduced according to the temperature parameters of simulated annealing. In addition, the step of switching to the local neighborhood search method is performed when the temperature parameter of pseudo-annealing falls below the threshold or when pseudo-annealing no longer provides a solution, whichever comes first. obtain.

Further, the present disclosure may provide a method for ride sharing. This method involves receiving multiple boarding reservations from the booking system (reservations include at least the requested boarding location, the requested disembarking location, either the boarding time or the disembarking time, and the required capacity. Obtaining, but not limited to, the booking system can receive data for multiple vehicles from the vehicle control system, including, but not limited to, at least the driver's schedule, average vehicle speed, vehicle capacity, and location. However, but not limited to these), road data can be received from the road network and traffic services (road data includes, but is not limited to, the road network, road closure schedules, and some or all of the traffic data) and provides services to the reservation. Therefore, the optimum route of the vehicle can be calculated. The booking system can improve the route in multiple iterations by generating an initial route to process the reservation and searching for the improved route near the current route in the solution space, the improved route. Is specified as the current route, the neighborhood search range is adaptively changed for each iteration, a subset of optimal route information is delivered to the user directly or through the reservation system, and a subset of optimal route information is delivered to the vehicle. It may further include delivery directly or via a vehicle control system. In addition, the booking system optionally includes application programming interfaces (APIs), smartphones, smart (digital) clocks, tablets or laptops, short messaging services (SMS), mobile messaging applications, centralized booking portals, and manual. Alternatively, it may include, but is not limited to, a reservation application accessible from an automated voice call system. In addition, the vehicle can be manned or self-driving and can be ground, air, water, amphibious and amphibious. In addition, the required capacity may accommodate humans, animals, or cargo, and may include special features such as wheelchairs, beds, containers, or storage areas. In addition, optimal route information includes assigned boarding locations, assigned disembarking locations, boarding and disembarking time slots for each reservation, reservation service order, assigned vehicles, and, in the case of manned vehicles, the driver. Can be included, but is not limited to these.

Further, the step of calculating the optimum route may further include a step of pre-treating the boarding place to an acceptable boarding place and a step of pre-treating the getting-off place to an allowed getting-off place. Further, the step of calculating the optimal route may further include the step of changing the assigned boarding or alighting place of a plurality of reservations of the optimized route. In addition, the step of selecting an acceptable boarding or disembarking location may be configured to prioritize the location closest to the requested boarding or disembarking location. In addition, the step of selecting an acceptable boarding or disembarking location can be configured to prioritize statistically popular locations. In addition, the step of changing the assigned boarding or alighting location identifies multiple permissible locations around each assigned location, identifies all permissible routes between said permissible locations, and identifies. It may further include deriving the shortest route from the permissible route and changing the assigned boarding or alighting location to the permissible location of the permissible route of the shortest route obtained. Further, this method collects commuting settings including preferred boarding place, preferred disembarking place, preferred boarding time, and preferred disembarking time, and from the commuting setting, a route pattern including at least one popular place. And may include changing the assigned boarding or alighting location to prioritize popular locations.

Means for performing the functions disclosed in the above description, claims below, or in the accompanying drawings and in their particular form or disclosed, or methods or processes for obtaining the disclosed results. The features appropriately expressed in view of the above can be utilized to realize the present invention in various forms thereof, either separately or in any combination of such features.

The invention has been described in conjunction with the exemplary embodiments described above, but given the present disclosure, many equivalent modifications and variations will be apparent to those of skill in the art. Accordingly, the above exemplary embodiments of the invention are considered exemplary and not limiting. Various modifications to the described embodiments can be made without departing from the scope of the invention.

To avoid misunderstanding, the theoretical explanations provided here are provided for the purpose of improving the reader's understanding. We do not want to be bound by any of these theoretical explanations.

The headings of all sections used herein are for structural purposes only and should not be construed as limiting the subject matter being described.

Throughout the specification, including the claims below, the terms "have," "include," "provide," and variations thereof, unless otherwise specified in the context, are specified integers or steps or or variations thereof. It is understood to mean that it includes a group of integers or steps, but does not exclude other integers or steps or groups of integers or steps.

As used herein and in the appended claims, the singular forms "a", "an", and "the" include multiple referents unless the context clearly dictates otherwise. Please note that. The range may be expressed herein as "about" from one particular value and / or "about" another particular value. When such a range is represented, another embodiment includes from that particular value and / or to that other particular value. Similarly, when values are expressed as approximations, it will be understood that by using the antecedent "about", a particular value forms another embodiment. The term "about" associated with numbers is optional and means, for example, +/- 10%.

As used herein, the terms "favorable" and "favored" refer to embodiments of the invention that may provide a particular advantage under some circumstances. However, it should be understood that other embodiments may also be preferred under the same or different circumstances. Accordingly, the enumeration of one or more preferred embodiments does not imply or imply that the other embodiments are not useful and exclude the other embodiments from the scope of the present disclosure or the claims. Is not intended.

Claims (16)

A computer-implemented method for route calculation in a multi-vehicle environment, with the following steps:
a) Receive the reservation data corresponding to the user's desired process,
b) At least one first route data is determined as the current route data corresponding to the reserved data, where each of the at least one first route data has at least one first route characteristic. Including,
c) Corresponding to the reserved data, at least one subsequent route data based on the current route data is determined, wherein each of the at least one subsequent route data has at least one subsequent route characteristic. Including,
d) The at least one subsequent route characteristic is compared with the at least one route characteristic of the current route data.
e) Based on the comparison, it is determined whether the quality of the at least one subsequent route data is improved relative to the at least one current route data.
f) Select one of the higher quality of the current route data and the subsequent route data as the route data solution or as the subsequent route data used for the next iteration.
g) Including transmitting the route data solution to at least one of the vehicles as a route solution and transmitting to the user as a booking solution.
The route characteristics indicate at least the quality of each route data.
Steps c) to f) are repeated until the defined termination conditions are met.
At least one of the at least one subsequent route data is determined by using a metaheuristic function that guides the neighborhood search function.
The neighborhood search function is adapted to determine a subsequent route data that operates from the current route data.
The search range of the neighborhood search function is set by the parameters of the metaheuristic function, and the metaheuristic function is composed of either a "pseudo-annealing" function or a "taboo" function.
The neighborhood search function includes an adaptive "ruin and rebuild" search function.
The setting of the neighborhood search range by the metaheuristic function includes the reduction of the adjacency range of the adaptive "ruin and rebuild" search function.
The parameter of the metaheuristic function is reduced within the iteration of at least steps c) to e), the parameter being a temperature parameter or taboo list size.
A method in which the adjacency range of the adaptive "ruin and recreate" search function is reduced according to the parameters. The method according to claim 1, wherein the neighborhood search function is any of the following.
i) Adaptive "ruin and rebuild" search function; or ii) Adaptive "ruin and rebuild" search function and "local neighborhood search" search function. In ii), the adaptive "ruin and rebuild" search function is used for the first number of iterations, and the "local neighborhood search" search function is used for the second number of iterations. The method according to claim 2. A route calculation system in a multi-vehicle environment
Communication elements adapted to receive reservation data corresponding to the user's desired journey,
A determinant adapted to determine at least one first route data as the current route data corresponding to the reserved data, each of the at least one first route data having at least one first route data. Determinants, including 1 root characteristic,
A determinant that corresponds to the reserved data and is adapted to determine at least one subsequent route data based on the current route data, each of the at least one subsequent route data being at least one. Determinants, including subsequent route characteristics,
A comparison element adapted to compare the at least one subsequent route characteristic with at least one root characteristic of the current route data.
A determinant adapted to determine whether the quality of the at least one subsequent route data is improved relative to the at least one current route data, based on the comparison.
A selection element adapted to select one of the higher quality from the current route data and the subsequent route data, either as a route data solution or as subsequent route data to be used for the next iteration. Including,
The route characteristics indicate at least the quality of each route data.
Determining subsequent route data, comparing subsequent route characteristics, determining quality improvement, and selecting improved quality route data is repeated until the defined termination conditions are met.
At least one of the at least one subsequent route data is determined by using a metaheuristic function that guides the neighborhood search function.
The neighborhood search function is adapted to determine a subsequent route data that operates from the current route data.
The search range of the neighborhood search function is set by the parameters of the metaheuristic function , and the metaheuristic function is composed of either a "pseudo-annealing" function or a "taboo" function.
The neighborhood search function includes an adaptive "ruin and rebuild" search function.
The setting of the neighborhood search range by the metaheuristic function includes the reduction of the adjacency range of the adaptive "ruin and rebuild" search function.
The parameters of the metaheuristic function are reduced at least within the iterative process of determining subsequent route data, comparing subsequent route characteristics, and determining quality improvements, and the parameters are temperature parameters. Or it is a taboo list size,
The adjacency range of the adaptive "ruin and recreate" search function is reduced according to the parameters.
system. The system of claim 4, wherein the communication element is further adapted to transmit the route data solution to at least one of the vehicles as a route solution and to the user as a booking solution. A transportation system for route scheduling in a multi-vehicle environment.
A communication element adapted to receive booking data from at least one user and to receive vehicle availability data for at least one vehicle,
Containing at least one processing element for determining at least one route data solution for the vehicle according to the reservation data of the at least one user.
The route data solution is a transportation system determined by performing the method according to any one of claims 1 to 3. The communication element is further adapted to provide the route data solution to the user and / or provide the route data solution to the at least one vehicle as itinerary data of the determined route data. , The transportation system according to claim 6. The route data solution provides the required boarding location, assigned boarding location, required disembarking location, assigned disembarking location, boarding time, boarding time frame, disembarking time, disembarking time frame, required capacity, service. It comprises at least one of a sequence, an assigned vehicle, an assigned driver, and / or said vehicle availability data is composed of at least one of a driver schedule, average vehicle speed, vehicle capacity, and vehicle position. In particular, the vehicle capacity includes the number of human passengers, the number of animal passengers, the size requirements of the goods to be transported, the wheelchair requirements, the bed requirements, the container size, and the required storage space size.
The transportation system according to claim 6 or 7. The communication element is further
Adapted to receive road data and use said road data to determine said at least one route.
In particular, the road data includes at least one of road network information, traffic service information, road closure information, and traffic information.
The transportation system according to any one of claims 6 to 8. Including reservation system
The reservation system includes a reservation application.
The booking application is via a mobile computing platform, smartphone, smart (digital) clock, tablet, laptop, short messaging service (SMS), mobile messaging application, centralized booking portal, telephone, and voice calling system. Adapted to receive at least one booking data from the user,
The transportation system according to any one of claims 6 to 9. Determining route data further includes preprocessing the boarding and disembarking locations to the permissible boarding and disembarking locations by means of preprocessing elements.
In particular, at least some of the permissible boarding locations and the permissible disembarking locations are associated with weight functions and are required for locations close to the requested location and / or statistically popular locations. A higher weight parameter is assigned than the location far from the location, and the weighting function is used to determine the permissible boarding and disembarking locations.
The transportation system according to any one of claims 6 to 10. The transportation system according to any one of claims 6 to 11, further comprising changing the boarding and / or disembarking location to which at least one route data is assigned. Changing the assigned boarding and / or disembarking location can be done by the transportation system.
Identify multiple permissible locations around each assigned location and
Identify all allowed routes between the allowed locations and
Determine the shortest route from the allowed routes and
12. The transportation system of claim 12, further comprising adapting the assigned boarding and / or disembarking location to an acceptable location on the permissible route of the shortest route obtained. Collect commuting settings, including preferred boarding locations, preferred disembarking locations, preferred boarding times, and preferred disembarking times.
Route patterns are determined from the commuting settings to include at least one popular location.
The transportation system according to any one of claims 6 to 13, further adapted to change the assigned boarding and / or disembarking place to give priority to a popular place. A computer program including instructions, wherein when the computer program is executed by the computer, the computer causes the computer to execute the method according to any one of claims 1 to 3. A computer-readable medium containing the computer program of claim 15.

Images