A route recalling method, a navigation method, related devices and systems

Abstract

The application discloses a route recalling method, a navigation method, related devices and systems. The route recalling method comprises the following steps: when receiving a route calculation request, selecting a multi-route target combination comprising at least two single-route targets, determining at least one set of single-route target weight combinations corresponding to the multi-route target combination, wherein each single-route target weight combination in the at least one set of single-route target weight combinations comprises a weight corresponding to each single-route target in the multi-route target combination; and recalling a multi-route target navigation route according to start and end point information carried by the route calculation request and the at least one set of single-route target weight combinations. The application solves the problem that the navigation route obtained by the prior art single-route target route recalling method cannot take into account different two or more route targets and cannot meet the user's demand for route planning.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of electronic map, in particular to a route recall method, a navigation method, related devices and systems. BACKGROUND

[0002] In the prior art, the navigation route generally adopts a single route calculation target recall strategy, recalling multiple navigation routes for each single route calculation target to obtain a navigation route combination. The recalled navigation route cannot simultaneously achieve the optimization of multiple targets because the navigation route that simultaneously meets multiple route calculation targets cannot be recalled in priority according to the single route calculation target recall strategy. For example, when the route calculation target is the shortest time or the shortest distance, the recalled route is the route with the shortest time or the shortest distance, and the route that is not optimal on either the shortest time or the shortest distance but is relatively optimal in combination of the two route calculation targets is difficult to be recalled in priority. However, when a user uses a travel application to plan a navigation route, the user has multiple requirements for the route, for example, the user may consider the road congestion degree, the driving distance, the driving time, the toll, and the like. However, the existing technology based on a single route calculation target cannot support such product requirements. SUMMARY

[0003] In view of the above problems, the present application is proposed to provide a route recall method, a navigation method, a device and a system that overcome the above problems or at least partially solve the above problems.

[0004] In a first aspect, an embodiment of the present application provides a route recall method, comprising the following steps:

[0005] When a route calculation request is received, a multiple route calculation target combination including at least two single route calculation targets is selected, and at least one set of single route calculation target weight combinations corresponding to the multiple route calculation target combination is determined, wherein each single route calculation target in the multiple route calculation target combination corresponds to a weight in the set of single route calculation target weight combinations.

[0006] According to the start and end point information carried by the route calculation request and the at least one set of single route calculation target weight combinations, a multiple route calculation target navigation route is recalled.

[0007] In one or some optional embodiments, the at least one set of single route calculation target weight combinations is obtained by pre-training in the following manner:

[0008] A plurality of historical route calculation requests are obtained;

[0009] For each historical route calculation request, at least one multiple route calculation target navigation route is obtained by using a multi-objective algorithm and the at least one set of single route calculation target weight combinations according to a preset multiple route calculation target combination.

[0010] According to the plurality of historical route calculation requests and the determined single route calculation target, recall at least one single route calculation target navigation route;

[0011] Input the route set composed of the obtained at least one multi-route calculation target navigation route and the at least one single route calculation target navigation route into a machine learning model for training, and adjust each set of single route calculation target weight combinations according to the training result until at least one set of optimized single route calculation target weight combinations is obtained.

[0012] In one or some optional embodiments, the at least one multi-route calculation target navigation route obtained according to the preset multi-target combination, the multi-target algorithm and the at least one set of single route calculation target weight combinations obtained through initial setting or training comprises:

[0013] According to the start and end point information carried by each historical route calculation request, determine a Pareto route set by using a selected multi-target algorithm; the Pareto route set includes a plurality of candidate routes meeting the preset multi-route calculation target combination requirement;

[0014] According to the at least one set of single route calculation target weight combinations obtained through initial setting or training, recall at least one multi-route calculation target navigation route from the Pareto route set.

[0015] In one or some optional embodiments, the inputting the route set composed of the obtained at least one multi-route calculation target navigation route and the at least one single route calculation target navigation route into the machine learning model for training, and adjusting at least one set of single route calculation target weight combinations according to the training result until at least one set of optimized single route calculation target weight combinations is obtained comprises:

[0016] In the machine learning model, determine the optimal route of the route set and the optimal route of the at least one single route calculation target navigation route;

[0017] According to the attribute value of the optimal route in the route set and the attribute value of the optimal route in the at least one single route calculation target navigation route, determine the loss value of the preset loss function, and according to the loss value of the loss function, adjust the weight of each single route calculation target in each set of single route calculation target weight combinations and the parameters of the machine learning model; the preset loss function is determined according to the preset navigation target combination by using a rectified linear unit (ReLU) function;

[0018] Repeat the above process until at least one set of single route calculation target weight combinations that minimize the loss function is obtained.

[0019] In one or some optional embodiments, the preset navigation target combination includes one optimization target and at least one target as a constraint condition, and the optimization target and the target as a constraint condition are selected from the following targets:

[0020] Expected driving time, driving distance, toll, congestion distance, distance of small roads, distance of unpaved roads, number of traffic lights, and number of steering actions.

[0021] In one or some optional embodiments, recalling the multi-algorithm route target navigation route according to the start and end point information carried in the algorithm request and the at least one set of single-algorithm route target weight combination comprises:

[0022] Determining, according to the start and end point information carried in the algorithm request and by using each set of single-algorithm route target weight combination, a route segment included in the multi-target navigation route to be recalled;

[0023] Weighting the weight of each route segment included in the multi-target navigation route to be recalled to obtain a route weight of the corresponding multi-target navigation route to be recalled;

[0024] Selecting a preset number of routes from the multi-target navigation routes to be recalled in order of route weight from small to large as the multi-algorithm route target navigation route.

[0025] In one or some optional embodiments, determining, according to the start and end point information carried in the algorithm request and by using each set of single-algorithm route target weight combination, a route segment included in the multi-target navigation route to be recalled comprises:

[0026] Dividing the road network into a plurality of grids in advance;

[0027] Determining, according to the start and end point information carried in the algorithm request, all grids passed through from the start point to the end point;

[0028] Weighting, by using the weight of each single-algorithm target in each set of single-algorithm route target weight combination, the attribute value of each algorithm target in each route segment corresponding to each grid passed through from the start point to the end point to obtain the weight of each route segment corresponding to each grid;

[0029] For each grid, selecting a preset number of route segments in order of weight from small to large as the route segment of the multi-target navigation route to be recalled.

[0030] In a second aspect, an embodiment of the present application provides a navigation method, comprising: obtaining at least one multi-algorithm route target navigation route according to the route recalling method described above and pushing the route to a navigation client.

[0031] In a third aspect, an embodiment of the present application provides a route recalling device, comprising:

[0032] determining a plurality of single-path target combination including at least two single-path targets when receiving the path calculation request, and determining at least one set of single-path target weight combination corresponding to the plurality of single-path target combination, wherein the set of single-path target weight combination includes the weight corresponding to each single-path target in the plurality of single-path target combination;

[0033] recalling a plurality of single-path target navigation route according to the start and end point information carried by the path calculation request and the at least one set of single-path target weight combination.

[0034] In a fourth aspect, an embodiment of the present application provides a navigation device, comprising:

[0035] determining a plurality of single-path target combination including at least two single-path targets when receiving the path calculation request, and determining at least one set of single-path target weight combination corresponding to the plurality of single-path target combination, wherein the set of single-path target weight combination includes the weight corresponding to each single-path target in the plurality of single-path target combination;

[0036] recalling a plurality of single-path target navigation route according to the start and end point information carried by the path calculation request and the at least one set of single-path target weight combination;

[0037] pushing the at least one plurality of single-path target navigation route to the navigation client.

[0038] In a fifth aspect, an embodiment of the present application provides a route recalling system, comprising a server and at least one client;

[0039] the client is configured to send a path calculation request and receive the route pushed by the server;

[0040] the server is configured to determine a plurality of single-path target combination including at least two single-path targets when receiving the path calculation request, determine at least one set of single-path target weight combination corresponding to the plurality of single-path target combination, wherein the set of single-path target weight combination includes the weight corresponding to each single-path target in the plurality of single-path target combination, recall a plurality of single-path target navigation route according to the start and end point information carried by the path calculation request and the at least one set of single-path target weight combination, and push the route to the client.

[0041] In a sixth aspect, an embodiment of the present application provides a route recalling system, comprising a server and at least one client;

[0042] the server is configured to send at least one set of single-path target weight combination to the client according to the received path calculation request, wherein the set of single-path target weight combination includes the weight corresponding to each single-path target in the plurality of single-path target combination;

[0043] The client is configured to recall a multi-algorithm target navigation route according to the algorithm request and at least one set of single-algorithm target weight combination sent by the server.

[0044] In a seventh aspect, a computer readable storage medium is provided, and the computer readable storage medium stores computer instructions. The computer instructions are executed by a processor to implement the route recalling method or the navigation method.

[0045] In an eighth aspect, a server is provided, and the server comprises a processor and a memory storing processor-executable instructions. The processor is configured to execute the route recalling method or the navigation method.

[0046] The technical scheme provided by the embodiments of the present application has at least the following beneficial effects:

[0047] Since the user selects a route in actual route planning is a multi-target, rather than only considering a single-algorithm target such as time or distance, in the technical scheme provided by the embodiments of the present application, each set of single-algorithm target weight combination contains the weight corresponding to at least two algorithm targets, and the navigation route is recalled by using at least one set of single-algorithm target weight combination. According to the algorithm request, a multi-algorithm target navigation route in which at least two algorithm targets are dominant can be obtained, and the navigation route in which each algorithm target is dominant is recalled, which is more in line with the actual route selection needs of the user. Moreover, according to the actual route planning needs, multiple algorithm targets can be considered at the same time, the number of sets of single-algorithm target weight combination is planned in advance, and a suitable number of multi-algorithm target navigation routes are recalled, thereby increasing the number of navigation routes that can be recalled according to the algorithm request and expanding the selection range of the navigation routes recommended to the user.

[0048] Other features and advantages of the present application will be set forth in the following description, and in part will become apparent to those skilled in the art from the description, or can be learned by practice of the present application. The objects and other advantages of the present application can be achieved and obtained by means of the structures particularly pointed out in the written description, the claims, and the accompanying drawings.

[0049] The technical scheme of the present application will be further described in detail below with the aid of the accompanying drawings and embodiments. BRIEF DESCRIPTION OF DRAWINGS

[0050] The accompanying drawings are included to provide a further understanding of the present application, and constitute a part of the specification, illustrate the embodiments of the present application, and are used to explain the present application, and do not constitute a limitation on the present application. In the drawings:

[0051] Figure 1 The flowchart of the route recalling method in the embodiments of the present application;

[0052] Figure 2 A flowchart of a single-calculated route target weight combination training process in an embodiment of the present application is shown in FIG. 1;

[0053] Figure 3 A flowchart of a multi-calculated route target navigation route acquisition process in a single-calculated route target weight combination training process in an embodiment of the present application is shown in FIG. 2;

[0054] Figure 4 A flowchart of a single-calculated route target weight combination training process in an embodiment of the present application is shown in FIG. 1;

[0055] Figure 5 A comprehensive training model schematic diagram when a machine learning model is trained in an embodiment of the present application is shown in FIG. 3;

[0056] Figure 6 A flowchart of a multi-calculated route target navigation route acquisition process in a route recall method in an embodiment of the present application is shown in FIG. 4;

[0057] Figure 7 A route segment determination process flowchart of a multi-target navigation route to be recalled in a route recall method in an embodiment of the present application is shown in FIG. 5;

[0058] Figure 8 A route recall device schematic diagram in an embodiment of the present application is shown in FIG. 6;

[0059] Figure 9 A navigation device schematic diagram in an embodiment of the present application is shown in FIG. 7;

[0060] Figure 10 A route recall system schematic diagram in an embodiment of the present application is shown in FIG. 8. DETAILED DESCRIPTION

[0061] Exemplary embodiments of the present disclosure will be described below in greater detail with reference to the accompanying drawings. Although exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. Rather, these embodiments are provided so that the present disclosure can be more thoroughly understood and the scope of the present disclosure can be accurately conveyed to those skilled in the art.

[0062] Embodiment 1:

[0063] Embodiments of the present application aim at the problems existing in the prior art described above, and provide a route recall method, a flowchart of which is shown in FIG. 1, which comprises the following steps: Figure 1

[0064] S101: When a route calculation request is received, a multi-calculated route target combination comprising at least two single-calculated route targets is selected, and at least one set of single-calculated route target weight combinations corresponding to the multi-calculated route target combination is determined, wherein each single-calculated route target weight combination in the set comprises a weight corresponding to each single-calculated route target in the multi-calculated route target combination; ​

[0065] In the step S101, when receiving the route calculation request, a multi-route target combination including at least two single-route targets is selected, and at least one set of single-route target weight combinations corresponding to the multi-route target combination is determined. The at least one set of single-route target weight combinations is obtained by training a machine learning model according to a preset multi-route target combination including at least two single-route targets. According to different preset multi-route target combinations, the machine learning model is trained to obtain single-route target weight combinations corresponding to different multi-target combinations. Before the machine learning model is trained, the value K of the number of sets of single-route target weight combinations can be preset according to actual needs, where K is a positive integer greater than or equal to 1, so as to recall a proper number of multi-route target navigation routes. For example, the route preference indicators of the user in the route calculation request include one or more of the following route preferences: shortest travel time, shortest travel distance, least number of traffic lights, least number of turning actions, congestion avoidance, toll avoidance, non-highway, and highway priority. Different users have different route preference indicators. When receiving the route calculation request, the corresponding multi-route target combination including at least two route targets can be found according to the route preference indicators in the route calculation request, and the corresponding K sets of single-route target weight combinations are selected according to the multi-target combination.

[0066] S102: recalling a multi-route target navigation route according to the start and end point information carried by the route calculation request and the at least one set of single-route target weight combinations.

[0067] In the step S102, the start and end point information carried by the route calculation request is used to calculate a route for each set of single-route target weight combinations, and at least one multi-route target navigation route corresponding to each set of single-route target weight combinations is determined according to the route calculation results.

[0068] Since the user selects a route based on multiple targets in actual route planning, rather than only considering a single route target such as time or distance, in the technical solution provided by the embodiments of the present application, each set of single-route target weight combinations includes weights corresponding to at least two route targets, and at least one set of single-route target weight combinations is used to recall a navigation route. According to the route calculation request, a multi-route target navigation route in which at least two route targets are dominant can be obtained, and a navigation route in which each route target is dominant can be recalled, which meets the actual route selection needs of the user. Moreover, multiple route targets can be considered simultaneously according to the actual route planning needs, the number of sets of single-route target weight combinations is planned in advance, and a proper number of multi-route target navigation routes are recalled, thereby increasing the number of navigation routes that can be recalled according to the route calculation request and expanding the selection range of navigation routes recommended to the user.

[0069] The specific embodiments of the present application are described in detail below.

[0070] Example 2

[0071] For one routing request, one single routing target is selected, and Dijkstra algorithm is used to perform routing, and the single routing target navigation route obtained cannot meet the demand of the user in route planning, because the user has multiple requirements for the navigation route when using the travel application to plan the navigation route, for example, the user may consider the road congestion degree, the driving distance, the driving time, the toll, and the like. Assuming that there are three routes p1, p2 and p3 from the starting point s to the ending point t, wherein the route p1 = {time = 20 minutes, distance = 10 kilometers}, the route p2 = {time = 12 minutes, distance = 11 kilometers}, and the route p3 = {time = 10 minutes, distance = 18 kilometers}. If the shortest distance single routing target is selected when performing route recall, and Dijkstra algorithm is used to perform routing, the route p1 can be recalled in theory; if the shortest time single routing target is selected, and Dijkstra algorithm is used to perform routing, the route p3 can be recalled in theory. However, based on the current single routing target, Dijkstra algorithm is used to perform routing, and only the routes p1 and p3 can be recalled, and the route p2 cannot be recalled. However, the route p2 is the optimal recalled route in terms of the two routing targets of the time and the distance, and therefore the route p2 is the route that the user wants to take in actual route planning. The inventors of the present application have found in practice that when performing routing, the two routing targets of the distance and the time are selected, and the multi-target shortest path algorithm in the prior art can be used to recall the route p2 described above.

[0072] For a given routing request, according to the start point and the end point of the routing request and the preset multi-routing target combination including at least two routing targets, the multi-target routing algorithm can obtain a Pareto route set corresponding to the preset multi-routing target combination. Assuming that the Pareto route set = {pl, p2,..., pm}, pl, p2,..., pm are candidate routes in the Pareto route set, and m is the number of candidate routes in the Pareto route set. The Pareto route set obtained by the multi-target routing algorithm is the optimal route set corresponding to the routing request, and any candidate route in the Pareto route set cannot dominate any other candidate route in the Pareto route set (for example, assuming that the distance of pl is shorter than that of p2, and the estimated arrival time of p2 is shorter than that of pl, pl cannot dominate p2, and p2 cannot dominate pl), and for a given routing request, other routes not belonging to the Pareto route set are dominated by at least one candidate route in the Pareto set (for example, assuming that a pn route is obtained according to the start point and the end point of the routing request, if the distance and the estimated arrival time of px are greater than those of p2 in the Pareto route set, the px route is dominated by the p2 route, so the px route does not belong to the Pareto route set).

[0073] In one embodiment, assuming that for a routing request, there are four routes: pl = {distance = 10 km, time = 20 min}, p2 = {distance = 12 km, time = 12 min}, p3 = {distance = 20 km, time = 10 min}, and p4 = {distance = 15 min, time = 15 min}, according to the distance and time attribute values of each route, it can be seen that the distance of route pl is the shortest, the time of route p3 is the shortest, the time of route p2 is shorter than that of pl, and the distance of route p2 is shorter than that of p3, that is, routes pl, p2 and p3 cannot dominate each other, and the time and distance of route p4 are greater than those of route p2, so route p4 is dominated by route p2. If the routing is performed according to the multi-target routing algorithm to obtain a Pareto route set, the Pareto route set = {pl, p2, p3}.

[0074] The inventor of the present application also found in experiments that the method of obtaining a Pareto route set according to an algorithm request and a preset multi-algorithm route target combination by using a multi-objective algorithm can obtain at least one multi-algorithm route target navigation route, but the more the number of single-algorithm route targets in the preset multi-algorithm route target combination, the slower the performance of the computer system used to recall the navigation route. Therefore, the method cannot be directly applied to a real-time interactive system and cannot be applied to real-time navigation route recall. This is because the multi-objective problem solved by using a multi-objective method is a non-deterministic polynomial problem (NP), which is a complex algorithm problem that cannot be solved in polynomial time, consumes a long time, and seriously occupies the resources of a computer system. Moreover, the more the number of single-algorithm route targets, the more the obtained alternative routes, the larger the Pareto route set, the more the occupation of the resources of the computer system, and the slower the system performance, which cannot meet the response requirements of a real-time interactive system. Therefore, the multi-objective algorithm cannot be applied to the navigation route recall of an electronic map.

[0075] To overcome the problem that direct use of a multi-objective algorithm cannot be applied to real-time route recall of an electronic map, the inventor of the present application improves the route recall method, pre-sets a multi-algorithm route target combination containing at least two algorithm route targets, and trains a machine learning model to obtain K sets of single-algorithm route target weight combinations corresponding to the multi-target combination, where K is a positive integer greater than or equal to 1. According to the start and end point information carried by the algorithm request and each set of single-algorithm route target weight combinations in the K sets of single-algorithm route target weight combinations, an algorithm is performed to obtain less than or equal to K multi-algorithm route target navigation routes. The route recall method can be applied to a real-time interactive system of map data, and the multi-target recall route meets the expectation of the user when planning a route.

[0076] Before training the K sets of single-algorithm route target weight combinations by using the machine learning model, the size of K can be determined according to the actual navigation route recall requirements. It should be noted that the larger the value of K, the more multi-algorithm route target navigation routes obtained when performing route recall, but the larger the value of K, the larger the amount of calculation of the computer system, the longer the time consumption, the more computer resources occupied, and the higher the performance requirements of the computer system for map data services. Therefore, to meet the time requirements of a real-time interactive system, the value of K cannot be set too large.

[0077] In one embodiment, as shown in FIG. 1, the at least one set of single-algorithm route target weight combinations is obtained by training a machine learning model. Figure 2

[0078] S201: Obtain a plurality of historical algorithm requests; ​

[0079] S202: For each historical routing request, at least one multi-objective routing path is obtained according to a preset multi-objective combination, by using a multi-objective algorithm and at least one set of single-objective weight combinations of routing objectives which are initially set or obtained in training;

[0080] S203: For each historical routing request and the determined single-objective, at least one single-objective routing path is recalled;

[0081] S204: A route set composed of the at least one multi-objective routing path and the at least one single-objective routing path is input into a machine learning model for training, and each set of single-objective weight combinations is adjusted according to a training result until at least one optimized set of single-objective weight combinations is obtained.

[0082] In one specific embodiment, referring to FIG. 2, the step S202 of obtaining at least one multi-objective routing path according to a preset multi-objective combination, by using a multi-objective algorithm and at least one set of single-objective weight combinations of routing objectives which are initially set or obtained in training, can be implemented by the following process: Figure 3 S301: According to the start and end point information carried by each historical routing request, a Pareto route set is determined by using a selected multi-objective algorithm; the Pareto route set includes a plurality of candidate routes meeting the preset multi-objective combination requirements;

[0083] S302: According to at least one set of single-objective weight combinations of routing objectives which are initially set or obtained in training, at least one multi-objective routing path is recalled from the Pareto route set.

[0084] In one specific embodiment, in the step S203, for each historical routing request and the determined single-objective, at least one single-objective routing path can be recalled by using a single-objective recall strategy in the prior art, for example, a single-objective routing and Dijkstra algorithm in the prior art can be used for routing, or a Cell-Based Routing (CBR) algorithm can be used for recalling a routing path.

[0085] In one specific embodiment, referring to FIG. 3, the step S204 of inputting a route set composed of the at least one multi-objective routing path and the at least one single-objective routing path into a machine learning model for training, and adjusting at least one set of single-objective weight combinations according to a training result until at least one optimized set of single-objective weight combinations is obtained, can be implemented by the following process:

[0086] Figure 4

[0087] ​​S401: In the machine learning model, determine the optimal route of the route set and the optimal route in at least one single route target navigation route;

[0088] S402: According to the attribute value of the optimal route in the route set and the attribute value of the optimal route in at least one single route target navigation route, determine the loss value of the preset loss function, and according to the loss value of the loss function, adjust the weight value of each single route target in each set of single route target weight combination and the parameter of the machine learning model; The preset loss function is determined by using the linear rectification ReLU function according to the preset navigation target combination; Repeat the above process until at least one set of single route target weight combination is obtained when the loss function is minimized.

[0089] In the embodiment of the application, in the step S202, the single route target in the preset multi-route target combination may be, for example, at least two of the following single route targets:

[0090] The conventional fastest, the shortest distance, the shortest estimated time, the least number of traffic lights and the least number of steering actions; wherein the conventional fastest is the route with the shortest travel time without considering the road congestion state.

[0091] In one specific embodiment, assuming that the preset multi-route target combination includes five single route targets of conventional fastest, shortest distance, shortest estimated time, least number of traffic lights and least number of steering actions, the attribute value of each candidate route in the obtained Pareto route set includes congestion avoidance attribute value, time attribute value, distance attribute value, traffic light number attribute value and navigation action number attribute value. When K sets of single route target weight combinations are initially set, K is a positive integer greater than or equal to 1, each set of single route target weight combinations includes the weight values of the five single route targets of conventional fastest, shortest distance, shortest estimated time, least number of traffic lights and least number of steering actions. In the training process, adjust the K sets of single route target weight combinations, that is, adjust the weight values of each single route target in each set of single route target weight combination in the K sets of single route target weight combinations, wherein the sum of the weight values of each single route target in each set of single route target weight combination is 1.

[0092] In one embodiment, from the K sets of single-objective weight combinations obtained from initial setting or training, K or less multi-objective navigation routes are selected from the Pareto route set. In a specific implementation, for example, each set of single-objective weight combinations can be used to weight each candidate route in the Pareto route set, i.e., the weight of each single objective in each set of single-objective weight combinations is multiplied by the corresponding objective attribute value of each candidate route in the Pareto route set, and then summed to obtain the comprehensive weight of each candidate route in the Pareto route set. According to the size of the comprehensive weight, all candidate routes in the Pareto route set are sorted, and the route with the smallest comprehensive weight is selected as the multi-objective navigation route filtered by the set of single-objective weight combinations. Since there may be a case where two sets of single-objective weight combinations filter the same multi-objective navigation route, when K sets of single-objective weight combinations are used to filter the multi-objective navigation route from the Pareto route set, K or less multi-objective navigation routes can be filtered.

[0093] In one specific embodiment, in step S402, when the preset loss function is determined using the linear rectifier (ReLU) function according to the preset navigation target combination, the preset navigation target combination includes an optimization target and at least one target as a constraint condition. The optimization target and the target as a constraint condition may, for example, be any one or more of the following:

[0094] Expected travel time, travel distance, toll, congestion distance, number of traffic lights in the route, and number of steering actions.

[0095] In one specific embodiment, the preset navigation target combination may, for example, be that the optimization target is the expected travel time, and the constraint condition is the travel distance and the toll, so as to control the route distance and the toll to be no greater than the single-objective navigation route obtained by the method according to the prior art, and to achieve the effect of reducing the expected travel time. The loss function described in the embodiments of the present application can be set according to the above-mentioned navigation target combination, for example, it can be set as the following formula:

[0096] Loss = feature_eta + lam_dis * tf.nn.relu(feature_dis-online_dis) + lam_toll * tf.nn.relu(feature_toll-online_toll).

[0097] In the above formula, feature_eta is the time attribute value of the optimal route in the route set composed of the multi-computation route target navigation route and the single-computation route target navigation route, feature_dis is the distance attribute value of the optimal route in the route set composed of the multi-computation route target navigation route and the single-computation route target navigation route, feature_toll is the toll attribute value of the optimal route in the route set composed of the multi-computation route target navigation route and the single-computation route target navigation route, online_dis is the distance attribute value of the optimal route in the single-computation route target navigation route, online_toll is the toll attribute value of the optimal route in the single-computation route target navigation route; lam_dis*tf.nn.relu(feature_dis-online_dis) represents a ReLU function set according to the distance, which is used to compare the sizes of feature_dis and online_dis, if feature_dis-online_dis is greater than 0, the function value is the real value of feature_dis-online_dis, if feature_dis-online_dis is less than or equal to 0, the function value is 0; lam_toll*tf.nn.relu(feature_toll-online_toll) represents a ReLU function set according to the toll, which is used to compare the sizes of feature_toll and online_toll, if feature_toll-online_toll is greater than 0, the function value is the real value of feature_toll-online_toll, if feature_toll-online_toll is less than or equal to 0, the function value is 0.

[0098] Since the K sets of single-computation route target weight combinations are obtained by training according to the preset loss function of the navigation target combination, the determination of the weight of each single-computation route target in each set of single-computation route target weight combinations conforms to the expected effect of the navigation target combination, and when the multi-computation route target navigation route is recalled through the K sets of single-computation route target weight combinations, the comprehensive multi-computation route target of the recalled navigation route is better than the single-computation route target navigation route recalled according to the single-computation route target recall strategy of the prior art.

[0099] In one embodiment, in the machine learning model training process, in the machine learning model, the process of determining the optimal route of each route set and the optimal route in the single-computation route target navigation route is as follows:

[0100] When the obtained route set composed of the multi-algorithm target navigation route and the single-algorithm target recall route is input into the machine learning model, the machine learning model determined according to the parameters of the machine learning model can output the ranking scores of all navigation routes in the route set, and all navigation routes in the route set are ranked according to the ranking scores, and the navigation route with the highest ranking score is the optimal route of the route set, and the navigation route with the highest ranking score in the single-algorithm target navigation route is the optimal route in the single-algorithm target navigation route.

[0101] In one specific embodiment, the machine learning model is a deep neural network (DNN) model, which can be, for example, a multi-dimensional neural network with intermediate connection neurons and activation functions. The specific structure of the machine learning model can refer to the description in the prior art, and is not specifically limited in the embodiment of the present application.

[0102] The training process of the K-set single weight combination and the parameters of the machine learning model is described below through a specific embodiment:

[0103] Referring to Figure 5 During training of the machine learning model, a comprehensive training model is established, including an algorithm request acquisition module 501, a Pareto set acquisition module 502, a multi-target route recall module 503, a single-target route recall module 504, and a machine learning model to be trained 505. For a plurality of historical algorithm requests, for example, a plurality of historical algorithm requests in the Beijing morning peak period, a plurality of algorithm target combinations including five single-algorithm targets of regular fastest, shortest distance, shortest estimated time, fewest traffic lights, and fewest turning actions are pre-set. And, suppose that the preset navigation target combination for setting the loss function is: to achieve the goal of shortening the estimated travel time while keeping the travel distance and toll no higher than the single-algorithm target navigation route obtained by the existing single-algorithm target recall strategy. The loss function is set as Loss = feature_eta + lam_dis * tf.nn.relu(feature_dis-online_dis) + lam_toll * tf.nn.relu(feature_toll-online_toll).

[0104] The training process of the comprehensive training model is specifically described as follows:

[0105] The algorithm request acquisition module 501 acquires a plurality of historical algorithm requests in the Beijing morning peak period from historical algorithm request data;

[0106] The Pareto set obtaining module 502 obtains, for each historical route calculation request, a Pareto route set meeting a preset multi-route target combination by using a multi-target algorithm, each candidate route in the Pareto route set being represented by attribute values corresponding to five route calculation targets;

[0107] The multi-target route recalling module 503 multiplies and sums the weights of each single route calculation target in each single route calculation target weight combination in the initialized K sets of single route calculation target weight combinations and attribute values of corresponding targets of all candidate routes in the Pareto set to obtain a comprehensive weight of each candidate route, and sorts comprehensive weights of all candidate routes from low to high, so that a route with the lowest comprehensive weight, i.e., a multi-route target navigation route, is obtained for each single route calculation target weight combination. Because several single route calculation target weight combinations may recall the same multi-route target navigation route, K sets of single route calculation target weight combinations are used to obtain less than or equal to K multi-route target navigation routes.

[0108] While the multi-route target navigation routes are determined by using the multi-target algorithm and the K sets of single route calculation target weight combinations, the single-target route recalling module 504 recalls at least one single-route target navigation route for each historical route calculation request and determined single route calculation target by using a single-route target recalling strategy.

[0109] The route set composed of the less than or equal to K multi-route target navigation routes and the at least one single-route target navigation route is input into the machine learning model 505 determined according to the parameters of the initialized machine learning model, and sorting scores of all navigation routes in the route set are output, the navigation routes are sorted in descending order of the sorting scores, the optimal route of the route set and the optimal route in the single-route target navigation routes are determined, time attribute values, distance attribute values and toll attribute values of the optimal route of the route set and distance attribute values and toll attribute values of the optimal route in the single-route target navigation routes are obtained, a loss value of a loss function is calculated, and the weights of each single route calculation target in the K sets of single route calculation target weight combinations and the parameters of the machine learning model are adjusted according to the loss value of the loss function.

[0110] In the same way as in the above process, the multi-target route recall module 303 obtains again less than or equal to K multi-calculation target navigation routes; the route set composed of the obtained less than or equal to K multi-calculation target navigation routes and at least one single-calculation target navigation route is input into the machine learning model 505 determined according to the parameters of the machine learning model obtained in the training, and then the optimal route of the corresponding route set and the optimal route in the single-calculation target navigation route are obtained, the loss value of the loss function is calculated, and according to the loss value of the loss function, the weight of each single-calculation target in each single-calculation target weight combination in the K single-calculation target weight combinations and the parameters of the machine learning model are continuously adjusted. The above process is continuously continued until the K single-calculation target weight combinations and the machine learning model parameters at the time when the loss function is minimized are obtained, and the training process of the machine learning model 505 is ended.

[0111] In the embodiment of the application, the training index of the machine learning model is to reduce the loss value of the loss function by adjusting the K single-calculation target weight combinations and the parameters of the machine learning model, and to find the optimal K single-calculation target weight combinations and the parameters of the machine learning model at the time when the loss function is minimized. That is, in the machine learning model training process, the K single-calculation target weight combinations and the parameters of the machine learning model are adjusted to optimize the loss value of the loss function, and the K single-calculation target weight combinations and the parameters of the machine learning model at the time when the loss function is minimized are obtained. When the route calculation request is received in real-time route recall, the route calculation is performed according to the start and end point information carried in the route calculation request and the K single-calculation target weight combinations obtained by training, and at least one multi-calculation target navigation route corresponding to each single-calculation target weight combination can be determined according to the route calculation result, and the multi-calculation target navigation route recall corresponding to the route calculation request is completed.

[0112] In one embodiment, the at least one multi-calculation target navigation route obtained according to the above route recall method can be input into the trained machine learning model 505, the ranking scores of all multi-calculation target navigation routes are output, all multi-calculation target navigation routes are ranked according to the ranking scores from high to low, and a preset number of multi-calculation target navigation routes are selected and pushed to the navigation client.

[0113] In one specific embodiment, referring to FIG. 8, the multi-calculation target navigation route is recalled according to the start and end point information carried in the route calculation request and the at least one single-calculation target weight combination, and the specific process is as follows: Figure 6

[0114] S601: According to the start and end point information carried in the route calculation request, the route segments included in the multi-target navigation route to be recalled are determined by using each single-calculation target weight combination;

[0115] ​S602: weight the weight values of the route segments included in each of the to-be-recalled multi-objective navigation route, to obtain a route weight value of the corresponding to-be-recalled multi-objective navigation route;

[0116] S603: select a preset number of routes from the to-be-recalled multi-objective navigation routes in the order of the route weight values from small to large as the multi-calculation route target navigation routes.

[0117] In one specific embodiment, referring to FIG. 1, in step S601, according to the start and end point information carried in the route calculation request, the route segments included in the to-be-recalled multi-objective navigation route are determined by using each set of single-calculation route target weight combination, which can be implemented by the following process: Figure 7

[0118] S701: divide the road network into a plurality of grids in advance;

[0119] S702: determine all the grids passed through from the start point to the end point according to the start and end point information carried in the route calculation request;

[0120] S703: weight the attribute values of each single-calculation route target in each route segment corresponding to each grid passed through from the start point to the end point by using the weight value of each single-calculation route target in each set of single-calculation route target weight combination, to obtain the weight value of each route segment corresponding to each grid;

[0121] S704: for each grid, select a preset number of route segments in the order of the weight values from small to large as the route segments of the to-be-recalled multi-objective navigation route.

[0122] In the embodiment of the application, in order to facilitate the determination of the route passed through from the start point to the end point of the route calculation request, the road network is divided into a plurality of grids in advance. When recalling the route, the grid where the start point and the end point of the route calculation request are located is determined first, and then all the grids passed through from the start point to the end point are determined. The attribute values of the corresponding target in each route segment corresponding to each grid passed through from the start point to the end point are weighted by using the weight value of each single-calculation route target corresponding to each single-calculation route target in each set of single-calculation route target weight combination, to obtain the weight value of each multi-calculation route target of each route segment corresponding to each grid. For each grid, the multi-calculation route targets are sorted in the order of the weight values from small to large, and a preset number of route segments are selected as the route segments included in the to-be-recalled multi-objective navigation route corresponding to the grid. The weight values of the route segments of the to-be-recalled multi-objective navigation route corresponding to all the grids from the start point to the end point are summed, to obtain all the possible to-be-recalled multi-objective navigation routes from the start point to the end point. The to-be-recalled multi-objective navigation routes are sorted in the order of the weight values from small to large, and a preset number of routes are selected from all the possible to-be-recalled multi-objective navigation routes as the multi-calculation route target navigation routes.

[0123] ​In one specific embodiment, after the road network is divided into multiple grids in advance, when recalling routes by using K sets of single route calculation target weight combinations, for each route calculation request, the grid where the starting point and the ending point of the route calculation request are located is determined, and then all the grids passed through from the starting point to the ending point are determined; for each set of single route calculation target weight combinations in the K sets of single route calculation target weight combinations, the weight of each single route calculation target in the single route calculation target weight combination is used to weight the attribute value of the corresponding target in each route segment corresponding to each grid passed through from the starting point to the ending point, so as to obtain the multi-route calculation target weight of each route segment corresponding to each grid; for each grid, the Dijkstra algorithm is used to filter out a shortest route segment; the weights of the shortest route segments corresponding to all the grids from the starting point to the ending point are summed up, so as to obtain all the multi-target navigation routes to be recalled from the starting point to the ending point; one route with the smallest weight is selected from all the multi-target navigation routes to be recalled as the multi-route calculation target navigation route corresponding to the set of single route calculation target weight combinations, and then K sets of single route calculation target weight combinations can recall K multi-route calculation target navigation routes, but considering that there may be a situation that more than two sets of single route calculation target weight combinations recall the same route, finally, less than or equal to K multi-route calculation target navigation routes are obtained, wherein K is a positive integer greater than or equal to 1.

[0124] The route recalling method provided by the embodiment of the application can obtain at least one set of single route calculation target weight combinations corresponding to a preset multi-route calculation target combination, calculate the weights of route segments corresponding to all the grids passed through from the starting point to the ending point of a route calculation request by using the at least one set of single route calculation target weight combinations, obtain route segments of at least one multi-target navigation route to be recalled corresponding to all the grids, and finally obtain less than or equal to K multi-route calculation target navigation routes. The method can be applied to a real-time interactive system, and the multi-route calculation target navigation route obtained by the method is superior to a single-route calculation target navigation route recalled according to a single-route calculation target recalling strategy in the prior art.

[0125] Based on the same inventive concept, the embodiment of the application further provides a navigation method, a route recalling device, a navigation device, a route recalling system, a related storage medium and equipment. Since the principles of the problems solved by the method, the device, the system, the related storage medium and the equipment are similar to those of the foregoing route recalling method, the implementation of the method, the device, the system, the related storage medium and the equipment can be referred to the implementation of the foregoing route recalling method, and the repeated parts will not be described herein.

[0126] Embodiment 3

[0127] Based on the same inventive concept, the embodiment provides a navigation method. At least one multi-route calculation target navigation route is obtained by using the route recalling method provided in the foregoing embodiment 1 or embodiment 2, and is pushed to a navigation client.

[0128] The specific process of pushing the multi-algorithm target navigation route in the navigation method provided by the embodiments of the present application can refer to the manner described in the above embodiment 2, or the obtained at least one multi-algorithm target navigation route is sorted by using other manners in the prior art, and a preset number of multi-algorithm target navigation routes in the front of the sorting are selected and pushed to the navigation client. In the embodiments of the present application, the specific pushing manner is not limited.

[0129] Embodiment 4

[0130] Based on the same inventive concept, the present embodiment provides a route recall device, which can be installed in a local server or a cloud server, and at least one multi-target recall route is obtained by executing the above route recall method. Referring to FIG. 8, the route recall device includes: Figure 8

[0131] The determining module 801 is configured to, when receiving an algorithm request, select a multi-algorithm target combination including at least two single-algorithm targets, and determine at least one set of single-algorithm target weight combinations corresponding to the multi-algorithm target combination, wherein each single-algorithm target weight combination includes the weight corresponding to each single-algorithm target in the multi-algorithm target combination.

[0132] The recalling module 802 is configured to recall a multi-algorithm target navigation route according to the start and end point information carried by the algorithm request and the at least one set of single-algorithm target weight combinations.

[0133] In one embodiment, the route recall device can further include a training module 800 configured to pre-train the at least one set of single-algorithm target weight combinations by the following manner:

[0134] Obtain a plurality of historical algorithm requests.

[0135] For each historical algorithm request, at least one multi-algorithm target navigation route is obtained according to a preset multi-algorithm target combination, a multi-target algorithm, and at least one set of single-algorithm target weight combinations obtained by initial setting or training.

[0136] Recall at least one single-algorithm target navigation route according to the plurality of historical algorithm requests and the determined single-algorithm target.

[0137] Input a route set composed of the obtained at least one multi-algorithm target navigation route and at least one single-algorithm target navigation route into a machine learning model for training, adjust each set of single-algorithm target weight combinations according to the training result, and obtain an optimized at least one set of single-algorithm target weight combinations.

[0138] ​In one embodiment, the training module 800 is specifically configured to determine a Pareto route set by using a selected multi-objective algorithm according to start and end point information carried by each historical route request;

[0139] According to at least one set of single-objective route target weight combination obtained in the initial setting or in the training, at least one multi-objective navigation route is recalled from the Pareto route set.

[0140] In one embodiment, the training module 800 is specifically configured to determine an optimal route of the route set and an optimal route of at least one single-objective navigation route in a machine learning model;

[0141] According to attribute values of the optimal route of the route set and attribute values of the optimal route of at least one single-objective navigation route, a loss value of a preset loss function is determined, and according to the loss value of the loss function, weights of each single-objective route target in each set of single-objective route target weight combination and parameters of the machine learning model are adjusted; the preset loss function is determined according to a preset navigation target combination by using a rectified linear unit (ReLU) function;

[0142] The above process is repeated until at least one set of single-objective route target weight combination is obtained when the loss function is minimized.

[0143] In one embodiment, the preset navigation target combination includes one optimization target and at least one target as a constraint condition, and the optimization target and the target as the constraint condition are selected from the following targets:

[0144] Expected driving time, driving distance, toll, congestion distance, distance of small roads, distance of roads without road conditions, number of traffic lights, and number of steering actions.

[0145] In one embodiment, the recalling module 802 is specifically configured to determine route segments included in a multi-objective navigation route to be recalled by using each set of single-objective route target weight combination according to start and end point information carried in a route request;

[0146] Each route segment included in each multi-objective navigation route to be recalled is weighted to obtain a route weight of a corresponding multi-objective navigation route to be recalled;

[0147] From the multi-objective navigation routes to be recalled, a preset number of routes are selected as multi-objective navigation routes in order of route weights from small to large.

[0148] In one embodiment, the recalling module 802 is specifically configured to pre-divide a road network into a plurality of grids;

[0149] According to the start and end point information carried in the algorithm request, all grids passed from the start point to the end point are determined;

[0150] The attribute values of each algorithm target in each route segment corresponding to each grid are weighted by using the weight values of each single algorithm target in each set of single algorithm target weight value combinations, to obtain the weight values of each route segment corresponding to each grid.

[0151] For each grid, a preset number of route segments are selected in the order of weight values from small to large, as the route segments of the multi-target navigation route to be recalled.

[0152] Embodiment 5

[0153] Based on the same inventive concept, the embodiment also provides a navigation device which can be installed in a local server or a cloud server, and obtains at least one multi-target recalled route by executing the above navigation method and pushes the route to a navigation client.

[0154] Referring to Figure 9 The navigation device comprises:

[0155] A determination module 901 is configured to, when receiving an algorithm request, select a multi-algorithm target combination comprising at least two single-algorithm targets, and determine at least one set of single-algorithm target weight value combinations corresponding to the multi-algorithm target combination, wherein each single-algorithm target weight value combination comprises weight values corresponding to each single-algorithm target in the multi-algorithm target combination.

[0156] A recall module 902 is configured to recall a multi-algorithm target navigation route according to the start and end point information carried in the algorithm request and the at least one set of single-algorithm target weight value combinations.

[0157] A push module 903 is configured to push the obtained at least one multi-algorithm target navigation route to a navigation client.

[0158] The navigation device provided by the embodiment of the present application, wherein the determination module 901 and the recall module 902 are similar to the determination module 901 and the recall module 902 in the route recall device in Embodiment 4, and the specific implementation can refer to the description of Embodiment 4.

[0159] Embodiment 6

[0160] Based on the same inventive concept, the embodiment also provides a route recall system, referring to Figure 10 The route recall system comprises a server 1 and at least one client 2.

[0161] The client 2 is configured to send an algorithm request and receive a route pushed by the server.

[0162] The server 1 is configured to, when receiving an algorithm request, select a multi-algorithm target combination including at least two single-algorithm targets, determine at least one set of single-algorithm target weight combinations corresponding to the multi-algorithm target combination, wherein each single-algorithm target weight combination includes weights corresponding to each single-algorithm target in the multi-algorithm target combination, recall a multi-algorithm target navigation route according to the start and end point information carried by the algorithm request and the at least one set of single-algorithm target weight combinations, and push the multi-algorithm target navigation route to the client 2.

[0163] Embodiment 7

[0164] Based on the same inventive concept, the embodiment also provides another route recalling system, which refers to Figure 10 as shown, including a server 1 and at least one client 2.

[0165] The server 1 is configured to send at least one set of single-algorithm target weight combinations to the client according to the received algorithm request, wherein each single-algorithm target weight combination includes weights corresponding to each single-algorithm target in the multi-algorithm target combination.

[0166] The client 2 is configured to recall a multi-algorithm target navigation route according to the algorithm request and the at least one set of single-algorithm target weight combinations sent by the server.

[0167] Embodiment 8

[0168] Based on the same inventive concept, the embodiment also provides a computer readable storage medium having computer instructions stored thereon, wherein the instructions are executed by a processor to implement the route recalling method of the above-mentioned embodiment 1 or 2.

[0169] Embodiment 9

[0170] Based on the same inventive concept, the embodiment also provides a computer readable storage medium having computer instructions stored thereon, wherein the instructions are executed by a processor to implement the navigation method of the above-mentioned embodiment 3.

[0171] Embodiment 10

[0172] Based on the same inventive concept, the embodiment also provides a server, including a processor and a memory for storing processor-executable instructions; wherein the processor is configured to execute the route recalling method of the above-mentioned embodiment 1 or 2.

[0173] Embodiment 11

[0174] Based on the same inventive concept, the embodiment also provides a server, including a processor and a memory for storing processor-executable instructions; wherein the processor is configured to execute the navigation method of the above-mentioned embodiment 3.

[0175] Those skilled in the art will appreciate that embodiments of the present application can be readily used as software, hardware, or a combination of software and hardware. In a software embodiment, the methods can be tangibly embodied in a machine-readable storage medium having stored thereon instructions that can be used to program a computer to perform any of the methods. The software implementation can be for example, in a modulated data signal such as a carrier wave or other transport mechanism, or a computer readable storage medium.

[0176] The present application is described in reference to the drawings using a flowchart and / or a block diagram of methods, apparatus (systems) and computer program products according to embodiments of the application. It will be understood that each block of the flowchart and / or block diagram, and combinations of blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart and / or block diagram block or blocks. Figure 1 one or more functions specified in the flowchart and / or block diagram block or blocks. Figure 1 means for performing each of the functions specified in the flowchart and / or block diagram block or blocks.

[0177] These computer program instructions can also be stored in a computer- readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instructions which implement the flowchart and / or block diagram block or blocks. Figure 1 one or more functions specified in the flowchart and / or block diagram block or blocks. Figure 1 means for performing each of the functions specified in the flowchart and / or block diagram block or blocks.

[0178] These computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the flowchart and / or block diagram block or blocks. Figure 1 one or more functions specified in the flowchart and / or block diagram block or blocks. Figure 1 means for performing each of the functions specified in the flowchart and / or block diagram block or blocks.

[0179] Obviously, numerous modifications and variations of the present application are possible in light of the above teachings. It is therefore to be understood that within the scope of the appended claims and their equivalents, the application can be practiced otherwise than as specifically described.

Claims

1. A route recall method, wherein, The method comprises the following steps: When receiving an algorithm request, a multi-algorithm target combination including at least two single-algorithm targets is selected, and at least one set of single-algorithm target weight combinations corresponding to the multi-algorithm target combination is determined, wherein each single-algorithm target weight combination includes the weight corresponding to each single-algorithm target in the multi-algorithm target combination; According to the start and end point information carried by the algorithm request and the at least one set of single-algorithm target weight combinations, a multi-algorithm target navigation route is recalled; The at least one set of single-algorithm target weight combinations is obtained through the following method: A plurality of historical algorithm requests are obtained; For each historical algorithm request, at least one multi-algorithm target navigation route is obtained by using a multi-objective algorithm and at least one set of single-algorithm target weight combinations which are initially set or obtained during training according to a preset multi-algorithm target combination; At least one single-algorithm target navigation route is recalled according to the plurality of historical algorithm requests and the determined single-algorithm target; The route set composed of the at least one multi-algorithm target navigation route and the at least one single-algorithm target navigation route is input into a machine learning model for training, and each set of single-algorithm target weight combinations is adjusted according to the training result until the optimized at least one set of single-algorithm target weight combinations is obtained.

2. The route call-back method of claim 1, wherein, The at least one multi-algorithm target navigation route is obtained by using a multi-objective algorithm and at least one set of single-algorithm target weight combinations which are initially set or obtained during training according to a preset multi-algorithm target combination, and the method comprises the following steps: According to the start and end point information carried by each historical algorithm request, a Pareto route set is determined by using a selected multi-objective algorithm; the Pareto route set includes a plurality of candidate routes meeting the preset multi-algorithm target combination requirements; At least one multi-algorithm target navigation route is recalled from the Pareto route set according to at least one set of single-algorithm target weight combinations which are initially set or obtained during training.

3. The route call-back method of claim 1, wherein, The route set composed of the at least one multi-algorithm target navigation route and the at least one single-algorithm target navigation route is input into a machine learning model for training, and each set of single-algorithm target weight combinations is adjusted according to the training result until the optimized at least one set of single-algorithm target weight combinations is obtained, and the method comprises the following steps: In the machine learning model, the optimal route in the route set and the optimal route in the at least one single-algorithm target navigation route are determined; According to the attribute values of the optimal route in the route set and the optimal route in the at least one single-algorithm target navigation route, the loss value of a preset loss function is determined, and the weight of each single-algorithm target in each set of single-algorithm target weight combinations and the parameters of the machine learning model are adjusted according to the loss value of the loss function; the preset loss function is determined by using a rectified linear unit (ReLU) function according to a preset navigation target combination; The above process is repeated until the at least one set of single-algorithm target weight combinations which minimizes the loss function is obtained.

4. The route call-back method of claim 3, wherein, The preset navigation target combination includes an optimization target and at least one target as a constraint condition, and the optimization target and the target as a constraint condition are selected from the following targets: The estimated driving time, driving distance, toll, congestion distance, distance on a small road, distance on an unpaved road, number of traffic lights, and number of steering actions.

5. The route call-back method of claim 1, wherein, The recalling of the multi-algorithm route target navigation route is based on the start and end point information carried by the algorithm request and the at least one set of single-algorithm route target weight combination. The route segments included in the multi-route target navigation route to be recalled are determined based on the start and end point information carried by the algorithm request and each set of single-algorithm route target weight combination. The weights of the route segments included in each multi-route target navigation route to be recalled are weighted to obtain the route weight of the corresponding multi-route target navigation route to be recalled. A preset number of routes are selected from the multi-route target navigation routes to be recalled in order of the route weight from small to large as the multi-algorithm route target navigation route.

6. The route recall method of claim 5, wherein, The route segments included in the multi-route target navigation route to be recalled are determined based on the start and end point information carried by the algorithm request and each set of single-algorithm route target weight combination. The road network is divided into a plurality of grids in advance. All the grids passed through from the start point to the end point are determined based on the start and end point information carried by the algorithm request. The attribute values of each algorithm target in each route segment corresponding to each grid passed through from the start point to the end point are weighted by using the weights of each single-algorithm target in each set of single-algorithm route target weight combination to obtain the weight of each route segment corresponding to each grid. For each grid, a preset number of route segments are selected in order of the weight from small to large as the route segments of the multi-route target navigation route to be recalled.

7. A method of navigation, wherein, The route recalling method according to any one of claims 1-6 obtains at least one multi-algorithm route target navigation route and pushes the route to a navigation client. The determining module is configured to, when receiving an algorithm request, select a multi-algorithm route target combination including at least two single-algorithm route targets, and determine at least one set of single-algorithm route target weight combination corresponding to the multi-algorithm route target combination, wherein each set of single-algorithm route target weight combination includes weights corresponding to each single-algorithm route target in the multi-algorithm route target combination; 8. A route recall apparatus wherein, The recalling module is configured to recall a multi-algorithm route target navigation route based on start and end point information carried by the algorithm request and the at least one set of single-algorithm route target weight combination. The at least one set of single-algorithm route target weight combination is obtained by the following method: A plurality of historical algorithm requests are obtained. For each historical algorithm request, at least one multi-algorithm route target navigation route is obtained by using a multi-target algorithm and at least one set of single-algorithm route target weight combination according to a preset multi-algorithm route target combination and initial setting or training. At least one single-algorithm route target navigation route is recalled based on the plurality of historical algorithm requests and the determined single-algorithm route target. The route set composed of the obtained at least one multi-algorithm route target navigation route and at least one single-algorithm route target navigation route is input into a machine learning model for training, and each set of single-algorithm route target weight combination is adjusted according to the training result until the optimized at least one set of single-algorithm route target weight combination is obtained. The route recalling method according to any one of claims 1-6 obtains at least one multi-algorithm route target navigation route and pushes the route to a navigation client. The determining module is configured to, when receiving an algorithm request, select a multi-algorithm route target combination including at least two single-algorithm route targets, and determine at least one set of single-algorithm route target weight combination corresponding to the multi-algorithm route target combination, wherein each set of single-algorithm route target weight combination includes weights corresponding to each single-algorithm route target in the multi-algorithm route target combination; 9. A device for navigation, wherein, The recalling module is configured to recall a multi-algorithm route target navigation route based on start and end point information carried by the algorithm request and the at least one set of single-algorithm route target weight combination. The at least one set of single-algorithm route target weight combination is obtained by the following method: A plurality of historical algorithm requests are obtained. For each historical algorithm request, at least one multi-algorithm route target navigation route is obtained by using a multi-target algorithm and at least one set of single-algorithm route target weight combination according to a preset multi-algorithm route target combination and initial setting or training. At least one single-algorithm route target navigation route is recalled based on the plurality of historical algorithm requests and the determined single-algorithm route target. The route set composed of the obtained at least one multi-algorithm route target navigation route and at least one single-algorithm route target navigation route is input into a machine learning model for training, and each set of single-algorithm route target weight combination is adjusted according to the training result until the optimized at least one set of single-algorithm route target weight combination is obtained. determining a multi-algorithm target combination including at least two single-algorithm targets when receiving the algorithm request, and determining at least one set of single-algorithm target weight combinations corresponding to the multi-algorithm target combination, wherein each single-algorithm target weight combination includes weights corresponding to each single-algorithm target in the multi-algorithm target combination; recalling a multi-algorithm target navigation route according to the start and end point information carried by the algorithm request and the at least one set of single-algorithm target weight combinations; pushing the obtained at least one multi-algorithm target navigation route to the navigation client; wherein the at least one set of single-algorithm target weight combinations is obtained through the following method: obtaining a plurality of historical algorithm requests; for each historical algorithm request, obtaining at least one multi-algorithm target navigation route according to a preset multi-algorithm target combination, a multi-target algorithm, and at least one set of single-algorithm target weight combinations obtained through initial setting or training; recalling at least one single-algorithm target navigation route according to the plurality of historical algorithm requests and determined single-algorithm targets; inputting a route set composed of the obtained at least one multi-algorithm target navigation route and at least one single-algorithm target navigation route into a machine learning model for training, adjusting each set of single-algorithm target weight combinations according to a training result, and obtaining an optimized at least one set of single-algorithm target weight combinations.

10. A route recall system wherein, comprise: a server and at least one client; the client, configured to send an algorithm request and receive a route pushed by the server; the server, configured to determine a multi-algorithm target combination including at least two single-algorithm targets when receiving the algorithm request, determine at least one set of single-algorithm target weight combinations corresponding to the multi-algorithm target combination, wherein each single-algorithm target weight combination includes weights corresponding to each single-algorithm target in the multi-algorithm target combination, recall a multi-algorithm target navigation route according to the start and end point information carried by the algorithm request and the at least one set of single-algorithm target weight combinations, and push the multi-algorithm target navigation route to the client; wherein the at least one set of single-algorithm target weight combinations is obtained through the following method: obtaining a plurality of historical algorithm requests; for each historical algorithm request, obtaining at least one multi-algorithm target navigation route according to a preset multi-algorithm target combination, a multi-target algorithm, and at least one set of single-algorithm target weight combinations obtained through initial setting or training; recalling at least one single-algorithm target navigation route according to the plurality of historical algorithm requests and determined single-algorithm targets; inputting a route set composed of the obtained at least one multi-algorithm target navigation route and at least one single-algorithm target navigation route into a machine learning model for training, adjusting each set of single-algorithm target weight combinations according to a training result, and obtaining an optimized at least one set of single-algorithm target weight combinations.

11. A route recall system wherein, comprise a server and at least one client; the server, configured to send at least one set of single-algorithm target weight combinations to the client according to a received algorithm request, wherein each single-algorithm target weight combination includes weights corresponding to each single-algorithm target in a multi-algorithm target combination; The client is configured to recall a multi-algorithm target navigation route according to the algorithm request and at least one set of single-algorithm target weight combination sent by the server. The at least one set of single-algorithm target weight combination is obtained by the following method: Obtain a plurality of historical algorithm requests; For each historical algorithm request, at least one multi-algorithm target navigation route is obtained according to a preset multi-algorithm target combination, a multi-objective algorithm, and at least one set of single-algorithm target weight combination obtained by initial setting or during training. Recall at least one single-algorithm target navigation route according to the plurality of historical algorithm requests and the determined single-algorithm target. Input a route set composed of the obtained at least one multi-algorithm target navigation route and at least one single-algorithm target navigation route into a machine learning model for training, adjust each set of single-algorithm target weight combination according to a training result, and obtain an optimized at least one set of single-algorithm target weight combination.

12. A computer readable storage medium having stored thereon computer instructions, wherein, The instructions are executed by the processor to implement the route recall method of any one of claims 1-6 or the navigation method of claim 7.

13. A server, wherein, Comprise: A processor and a memory for storing processor-executable commands, wherein the processor is configured to execute the route recall method of any one of claims 1-6 or the navigation method of claim 7.

Images