CN113358126A

CN113358126A - Navigation route obtaining method, device and system

Info

Publication number: CN113358126A
Application number: CN202010146998.1A
Authority: CN
Inventors: 赵睿; 刘岭岭; 徐龙飞; 程冉; 冀晨光; 刘凯奎
Original assignee: Alibaba Group Holding Ltd
Current assignee: Alibaba Group Holding Ltd
Priority date: 2020-03-05
Filing date: 2020-03-05
Publication date: 2021-09-07

Abstract

The invention discloses a navigation route obtaining method, a navigation route obtaining device and a navigation route obtaining system. The method comprises the following steps: recalling a multi-calculation road target navigation route according to the road calculation request and at least one set of single calculation road target weight combination, wherein the set of single calculation road target weight combination comprises weights respectively corresponding to at least two single calculation road targets; recalling the single route calculation target navigation route according to the route calculation request and the determined single route calculation target; and sequencing the multi-path target navigation route and the single-path target navigation route to determine a recommended navigation route pushed to the user. The method and the device solve the problem that in the prior art, the navigation route obtained by recalling the single calculated route target navigation route by using the single calculated route target does not meet the navigation requirement of the user.

Description

Navigation route obtaining method, device and system

Technical Field

The invention relates to the field of electronic map navigation, in particular to a navigation route acquisition method, device and system.

Background

In the prior art, a navigation route generally adopts a recall strategy of single route calculation targets, and a plurality of routes recalled aiming at each single route calculation target are sorted and then pushed to a user. The route recalled according to the recall strategy of the single calculation route target cannot simultaneously realize the optimization of a plurality of calculation route targets, because the navigation route simultaneously meeting the plurality of calculation route targets cannot be recalled preferentially according to the recall strategy of the single calculation route target. For example, when the route calculation target is the shortest time or the distance is the shortest, the route with the shortest time or the shortest distance is preferentially pushed to the user, and the route which is not optimal on any route calculation target with the shortest time or the shortest distance but has better comprehensive two route calculation targets is difficult to be preferentially recalled and pushed to the user. However, when the user performs navigation route planning using the travel-type application, the user may have various route requirements, for example, the user may consider the road congestion degree, the travel distance, the travel time, the charge, and the like. However, the existing technology based on the single computation way target cannot support the product requirement.

Disclosure of Invention

In view of the above, the present invention has been made to provide a navigation route acquisition method, apparatus and system that overcome or at least partially solve the above problems.

In a first aspect, an embodiment of the present invention provides a method for acquiring a navigation route, including the following steps:

recalling a multi-calculation road target navigation route according to the road calculation request and at least one set of single calculation road target weight combination, wherein the set of single calculation road target weight combination comprises weights respectively corresponding to at least two single calculation road targets;

recalling the single route calculation target navigation route according to the route calculation request and the determined single route calculation target;

and sequencing the multi-path target navigation route and the single-path target navigation route to determine a recommended navigation route pushed to the user.

In one or some alternative embodiments, the method further comprises:

determining a single route calculation target corresponding to the route preference index according to the route preference index carried in the route calculation request;

and if the number of the determined single route calculation targets exceeds two, executing the step of recalling the navigation route of the multiple route calculation targets.

In one or some alternative embodiments, the recalling the multi-route target navigation route according to the route calculation request and at least one set of single-route target weight combination includes:

determining route segments included in the multi-target navigation route to be recalled by utilizing each set of single calculation route target weight combination according to start and end point information carried in the route calculation request;

weighting the weight of the route segment included in each multi-target navigation route to be recalled to obtain the route weight of the corresponding multi-target navigation route to be recalled;

and selecting a preset number of routes as multi-calculation route target navigation routes from the multi-target navigation routes to be recalled according to the sequence of the route weights from small to large.

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

dividing a road network into a plurality of grids in advance;

determining all grids passing from the starting point to the end point according to the starting and end point information carried in the route calculation request;

weighting the attribute values of the single calculation targets in the route segments corresponding to each grid from the starting point to the end point by using the weight values of the single calculation targets in each set of single calculation target weight value combination to obtain the weight values of the route segments corresponding to each grid;

and for each grid, selecting a preset number of route segments as the route segments of the multi-target navigation route to be recalled according to the sequence of the weights from small to large.

In one or some optional embodiments, the route preference metrics include:

the traffic time is shortest, the driving distance is shortest, the number of traffic lights is least, the steering action is least, and congestion is avoided.

In one or some optional embodiments, the sorting the multi-route target navigation route and the single-route target navigation route to determine a recommended navigation route to be pushed to the user includes:

inputting the at least one multi-route target navigation route and the at least one single-route target navigation route into a trained ranking model, and outputting ranking scores of the at least one multi-route target navigation route and the at least one single-route target navigation route;

and sorting according to the sorting scores to obtain the recommended navigation route pushed to the user.

In one or some alternative embodiments, the ranking model is obtained by:

setting a loss function by utilizing a linear rectification ReLU function according to a preset navigation target combination;

determining a Pareto route set by using a selected multi-objective algorithm according to a plurality of historical route calculation requests, and recalling at least one multi-route target navigation route from the Pareto route set according to at least one set of single route target weight combination obtained in initial setting or training;

recalling at least one single calculation road target navigation route according to the plurality of historical road calculation requests and the determined single calculation road targets;

inputting a route set consisting of at least one multi-route target navigation route and at least one single-route target navigation route into a sequencing model, and determining an optimal route in the route set and an optimal route in at least one single-route target navigation route;

and determining a loss value of the loss function according to the attribute value of the optimal route in the route set and the attribute value of the optimal route in the single calculation route target navigation route, adjusting the weight of each single calculation route target of each set of single calculation route target weight combination and the parameters of the sequencing model according to the loss value of the loss function, and repeating the process until at least one set of single calculation route target weight combination and parameters of the sequencing model when the loss function is minimized are obtained.

In one or some alternative embodiments, the combination of navigation objectives comprises an optimization objective and at least one constraint objective, the optimization objective and the constraint objective being selected from the following:

predicted travel time, travel distance, toll, congestion distance, lane distance, no-condition road distance, traffic light count, and steering action count.

In a second aspect, an embodiment of the present invention provides a navigation route obtaining apparatus, including: the multi-objective recall module is used for retrieving the multi-calculation road target navigation route according to the calculation road request and at least one set of single calculation road target weight combination, wherein the set of single calculation road target weight combination comprises weights respectively corresponding to at least two single calculation road targets;

the single target recall module is used for recalling a single route calculation target navigation route according to the route calculation request and the determined single route calculation target;

and the sequencing module is used for sequencing the multi-route target navigation route and the single-route target navigation route and determining a recommended navigation route pushed to the user.

In a third aspect, the present invention provides a computer-readable storage medium, on which computer instructions are stored, where the instructions, when executed by a processor, implement the navigation route obtaining method as described above.

In a fourth aspect, an embodiment of the present invention provides a computer device, including: a processor, a memory for storing processor executable commands; wherein the processor is configured to execute the navigation route obtaining method

The technical scheme provided by the embodiment of the invention has the beneficial effects that at least:

the embodiment of the invention provides a navigation route obtaining method, a navigation route obtaining device and a navigation route obtaining system. The navigation route obtaining method recalls the multi-calculation road target navigation route through at least one set of single calculation road target weight combination, sorts the obtained multi-calculation road target navigation route and the single calculation road target navigation route obtained according to the determined single calculation road target, and determines the recommended navigation route pushed to the user. The number of the navigation routes is increased, the navigation routes with better comprehensive multiple route calculation targets can be pushed to the user, and the pushed recommended navigation routes are more in line with the requirements of the actual route planning of the user. Moreover, a plurality of route calculation targets can be simultaneously considered according to the requirement of actual route planning, the number of sets of weight combinations of the single route calculation targets can be planned in advance according to actual requirements, and the number of the navigation routes with the multiple route calculation targets recalled can be adjusted according to the number of the recommended navigation routes pushed to the user.

Additional features and advantages of the invention will be set forth in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. The objectives and other advantages of the invention will be realized and attained by the structure particularly pointed out in the written description and claims hereof as well as the appended drawings.

The technical solution of the present invention is further described in detail by the accompanying drawings and embodiments.

Drawings

The accompanying drawings, which are included to provide a further understanding of the invention and are incorporated in and constitute a part of this specification, illustrate embodiments of the invention and together with the description serve to explain the principles of the invention and not to limit the invention. In the drawings:

FIG. 1 is a flowchart of a navigation route obtaining method according to an embodiment of the present invention;

FIG. 2 is a flow chart of a multi-way object navigation route recall method in accordance with an embodiment of the present invention;

FIG. 3 is a flow chart illustrating a process of determining route segments of multi-destination navigation routes to be recalled in the multi-destination navigation route recall method according to an embodiment of the present disclosure;

FIG. 4 is a flowchart of a ranking model training process in an embodiment of the present invention;

FIG. 5 is a schematic diagram of a ranking model provided in an embodiment of the present invention;

FIG. 6 is a schematic diagram of a navigation route acquisition device according to an embodiment of the present invention;

FIG. 7 is a diagram illustrating a navigation route obtaining system according to an embodiment of the present invention.

Detailed Description

Exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be embodied in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

The embodiment of the present invention provides a method for acquiring a navigation route, which is directed to the problems in the prior art, and the flow of the method is shown in fig. 1, and the method includes the following steps:

s101: and recalling the multi-route target navigation route according to the route calculation request and at least one set of single route target weight combination, wherein one set of single route target weight combination comprises weights respectively corresponding to at least two single route targets.

At least one set of single-calculation road target weight combination in the step S101 is obtained through machine learning model training according to a preset multi-calculation road target combination including at least two single-calculation road targets. Before the machine learning model training is started, the value K of the set number of the single-calculation road target weight combination can be preset according to actual needs, wherein K is a positive integer greater than or equal to 1, so that an appropriate number of multi-calculation road target navigation routes can be recalled. When a route calculation request is received, a process of recalling a multi-route calculation target navigation route may be that routes are calculated according to starting point and end point information in the route calculation request and each set of single-route calculation target weight combination in K sets of single-route calculation target weight combinations, a route calculation result corresponding to each set of single-route calculation target weight combination is obtained, and at least one multi-route calculation target navigation route corresponding to each set of single-route calculation target weight combination is determined according to the route calculation result.

S102: and recalling the single route calculation target navigation route according to the route calculation request and the determined single route calculation target.

In step S102, at least one single-route-target navigation route may be recalled by using a single-target recall policy in the prior art according to the route calculation request and the determined single-route target, for example, the route calculation may be performed by using a single target and dijkstra algorithm in the prior art, or a Cell-Based Routing (CBR) algorithm.

S103: and sequencing the multi-path target navigation route and the single-path target navigation route to determine a recommended navigation route pushed to the user.

In step S103, the multi-route target navigation route and the single-route target navigation route obtained in steps S101 and S102 are input into a trained ranking model, ranking is performed according to a preset ranking index, at least one navigation route is obtained according to a ranking result, the navigation client can obtain the at least one navigation route, and a user selects a proper navigation route according to actual requirements to complete route planning.

According to the navigation route obtaining method provided by the embodiment of the invention, the multi-route target navigation route is recalled through at least one set of single-route target weight combination, the obtained multi-route target navigation route and the single-route target navigation route obtained according to the determined single-route target are sequenced, and the recommended navigation route pushed to the user is determined. The number of the navigation routes is increased, the navigation routes with better single route calculation targets can be pushed to the user, and the pushed recommended navigation routes are more in line with the requirements of actual route planning of the user. Moreover, a plurality of single calculation road targets can be simultaneously considered according to the requirement of actual route planning, the number of single calculation road target weight combinations can be planned in advance according to actual requirements, and the number of the navigation routes with the multiple calculation road targets recalled can be adjusted according to the number of the recommended navigation routes pushed to a user.

The following describes the embodiments of the present invention in detail.

Example 1:

in a specific embodiment, in the navigation route obtaining method provided in the foregoing embodiment, before the step of recalling the multi-route target navigation route, the method may further include:

In the embodiment of the invention, the route preference indexes according to the user comprise one or more of the following route preference indexes: the traffic time is shortest, the driving distance is shortest, the number of traffic lights is least, the steering action is least, congestion is avoided, charging is avoided, high speed is not required, and high speed is preferred. The route preference of different users is different, when a route calculation request is received, one or more single calculation route targets corresponding to the route preference index can be determined according to the route preference index carried in the calculation record request, when the determined single calculation route targets exceed two, the more than two single calculation route targets in the group are used as a multi-calculation route target combination, and at least one set of corresponding single calculation route target weight combination is selected according to the multi-calculation route target combination.

In one embodiment, referring to fig. 2, the specific process of recalling the multi-way target navigation route is as follows:

s201: determining route segments included in the multi-target navigation route to be recalled by utilizing each set of single calculation route target weight combination according to start and end point information carried in the route calculation request;

s202: weighting the weight of the route segment included in each multi-target navigation route to be recalled to obtain the route weight of the corresponding multi-target navigation route to be recalled;

s203: and selecting a preset number of routes as multi-calculation route target navigation routes from the multi-target navigation routes to be recalled according to the sequence of the route weights from small to large.

In a specific embodiment, referring to fig. 3, in step S201, according to the start-end point information carried in the route calculation request, determining route segments included in the multi-target navigation route to be recalled by using each set of single route calculation target weight combination may be implemented through the following processes:

s301: dividing a road network into a plurality of grids in advance;

s302: determining all grids passing from the starting point to the end point according to the starting and end point information carried in the route calculation request;

s303: weighting the attribute values of the single calculation targets in the route segments corresponding to each grid from the starting point to the end point by using the weight values of the single calculation targets in each set of single calculation target weight value combination to obtain the weight values of the route segments corresponding to each grid;

s304: and for each grid, selecting a preset number of route segments as the route segments of the multi-target navigation route to be recalled according to the sequence of the weights from small to large.

In the embodiment of the invention, in order to determine the route from the starting point to the end point of the route calculation request, the road network is divided into a plurality of grids in advance. When the route is recalled, firstly determining grids where a starting point and an end point of the route calculation request are located, and then determining all grids from the starting point to the end point; weighting attribute values of corresponding targets in route segments corresponding to grids passing from a starting point to an end point by utilizing weights corresponding to the single calculation targets in each set of single calculation route target weight combination to obtain weights of multiple calculation route targets of the route segments corresponding to the grids, sequencing each grid according to the sequence from small to large of the weights of the multiple calculation route targets, and selecting route segments with preset number as route segments of the multiple target navigation route corresponding to the grid to be recalled; and summing the weights of the route segments of the multi-target navigation routes to be recalled corresponding to all grids from the starting point to the end point to obtain all possible multi-target navigation routes to be recalled from the starting point to the end point, sequencing the multi-target navigation routes to be recalled from small to large according to the sequence of the weights of the multi-target navigation routes to be recalled, and selecting a preset number of routes from all the possible multi-target navigation routes to be recalled as the multi-calculation target navigation routes.

In a specific embodiment, after a road network is divided into a plurality of grids in advance, when K sets of single calculation road target weight combinations are used for recalling a route, for each calculation road request, firstly determining grids where a starting point and an end point of the calculation road request are located, and then determining all grids passing from the starting point to the end point; for each set of single calculation route target weight combination in the K sets of single calculation route target weight combinations, respectively weighting the attribute values of corresponding targets in each route segment corresponding to each grid passing from the starting point to the end point by using the weight of each single calculation route target in the single calculation route target weight combination to obtain the multiple calculation route target weight of each route segment corresponding to each grid, and screening out a shortest route segment for each grid by using a Dijkstra algorithm; summing the weights of the shortest-path line sections corresponding to all grids from the starting point to the end point to obtain all multi-objective navigation routes to be recalled from the starting point to the end point, selecting one route with the smallest weight from all the multi-objective navigation routes to be recalled as a multi-objective navigation route corresponding to the single-objective weight combination, wherein the multi-objective navigation route can be recalled by K sets of single-objective weight combinations, but considering the situation that more than two sets of single-objective weight combinations recall the same route, the final result is that the number of the multi-objective navigation routes is less than or equal to K, and K is a positive integer greater than or equal to 1.

Example 2

In a specific embodiment, the sorting the multi-route target navigation route and the single-route target navigation route described in step S103 to determine a recommended navigation route to be pushed to the user includes:

In the embodiment of the present invention, the described ranking score may be a ranking score calculated by the ranking model for the at least one multi-route target navigation route and the at least one single-route target navigation route, and the ranking score of each route is output. And in the sequencing model, sequencing according to the sequencing scores from high to low, and selecting a preset number of navigation routes with the sequencing scores at the top from the sequencing results as recommended navigation routes pushed to the user.

In a specific embodiment, the ranking model described in the above embodiment is obtained by pre-training in the following manner:

determining a Pareto route set by using a selected multi-objective algorithm according to a plurality of historical route calculation requests, and screening and obtaining K multi-route calculation target navigation routes which are less than or equal to K from the Pareto route set according to K sets of single route calculation target weight combinations obtained in initial setting or training;

recalling at least one single calculation road target navigation route according to the single calculation road targets determined by the plurality of historical calculation road requests and determined by the plurality of historical calculation road requests;

inputting a route set composed of K multi-route target navigation routes and at least one single-route target navigation route into a sequencing model, and determining an optimal route in the route set and an optimal route in the at least one single-route target navigation route;

and determining the loss value of the loss function according to the attribute value of the optimal route in the route set and the attribute value of the optimal route in the single calculation route target navigation route, adjusting the weight of each single calculation route target of each set of single calculation route target weight combination and the parameters of the machine learning model according to the loss value of the loss function, and repeating the process until K sets of single calculation route target weight combinations and the parameters of the machine learning model when the loss function is minimized are obtained.

In a specific embodiment, the combination of navigation objectives comprises an optimization objective and at least one objective as a constraint, the optimization objective and the objective as a constraint being selected from the following objectives:

Example 3:

for a route calculation request, a single route calculation target is selected, route calculation is performed by utilizing a dijkstra algorithm, and the obtained single route calculation target navigation route cannot meet the requirement of a user on the navigation route when the user performs route planning. It is assumed that there are three routes p1, p2, and p3 from the start point s to the end point t, where the route p1 is 20 minutes and 10 km, the route p2 is 12 minutes and 11 km, and the route p3 is 10 minutes and 18 km. If the shortest distance single calculation path target is selected when the path is recalled, calculation is carried out by utilizing Dijkstra algorithm, and the p1 path can be recalled theoretically; and selecting a single calculation path target with the shortest time, calculating a path by utilizing a Dijkstra algorithm, and theoretically recalling the p3 path. But based on the current single-calculation route target, the route recalling is carried out by utilizing Dijkstra algorithm, only the p1 and p3 routes can be recalled, and the p2 route cannot be recalled. However, the p2 route integrates the time and distance to calculate the optimal navigation route, so the p2 route is the route the user most wants to go when planning the actual route. The inventor of the present invention finds in practice that when performing route calculation, two route calculation targets of distance and time are selected, and the above-described p2 route can be recalled by using the multi-target shortest path algorithm in the prior art.

Specifically, for a given route calculation request, a Pareto route set corresponding to at least two route calculation targets can be obtained by using an existing multi-target algorithm according to a starting point and an end point of the route calculation request and a multi-route calculation target combination including at least two route calculation targets. Assuming that the Pareto route set is { p1, p2, … pm }, then p1, p2, … pm are the alternative routes in the Pareto route set, and m is the number of alternative routes in the Pareto route set. The Pareto route set obtained by the multi-objective algorithm is the optimal route set corresponding to the routing request, and any alternative route in the Pareto route set cannot take precedence (cancel) on any other alternative route in the same Pareto route set (for example, if the distance of the p1 route is shorter than that of the p2 route, and the predicted arrival time of the p2 route is shorter than that of the p1 route, the p1 cannot cancel 2, and the p2 cannot cancel the p1), and for a given routing request, all other routes that are not in the Pareto route set are domined by at least one alternative route in the Pareto set (e.g., if a pn route is obtained from the start and end points of the routing request, the px route is domined by the p2 route if the distance and the expected arrival time of the px route are both greater than the distance and the expected arrival time of the p2 route in the Pareto route set, so the px route does not belong to the Pareto route set).

In one embodiment, if there are 4 routes for a way calculation request: p1 ═ distance 10 km, time 20 min }, p2 ═ distance 12 km, time 12 min }, p3 ═ distance 20 km, time 10 min }, p4 ═ distance 15 min, time 15 min }, it can be seen from the distance and time attributes of the routes: the shortest distance of the route p1, the shortest time of the route p3, the shorter time of the route p2 than p1 and the shorter distance of the route p3, namely, the routes p1, p2 and p3 cannot be mutually cancelled, and the time and the distance of the route p4 are both larger than those of the route p2, so that the route p4 is cancelled by the route p2, and if the route is calculated according to the multi-objective algorithm to obtain a Pareto route set, the Pareto route set is { p1, p2, p3 }.

The inventor of the present invention also finds in experiments that, although at least one multi-calculation-route target navigation route can be obtained by using a multi-objective algorithm and a method for obtaining a Pareto route set according to a calculation request and a preset multi-calculation-route target combination, if the number of single calculation-route targets in the preset multi-calculation-route target combination is larger, the performance of a computer system for recalling the navigation route is slower, and therefore, the method cannot be directly applied to a real-time interactive system and cannot be applied to recalling the navigation route in real time. The multi-target problem solved by the multi-target method is a Non-Deterministic Polynomial problem (NP), and the complex algorithm problem which is not solved in the Polynomial time is long in consumed time and serious in resource occupation of a computer system, and the more the number of single calculation path targets is, the more alternative routes are obtained, the larger the set of Pareto routes is, the more the resource occupation of the computer system is, the slower the system performance is, and the response requirement of the real-time interactive system cannot be met, so that the multi-target algorithm cannot be applied to the navigation route of the electronic map to recall.

In order to solve the problem that a multi-objective algorithm cannot be applied to real-time route recall of an electronic map directly, the inventor of the invention improves a route recall method, presets a multi-calculation route target combination comprising at least two single-calculation route targets, and trains by using a machine learning model to obtain K sets of single-calculation route target weight combinations corresponding to the multi-calculation route target combination, wherein K is a positive integer greater than or equal to 1. And calculating the route according to the starting point and the end point in the route calculation request and each set of single route target weight combination in the K sets of single route target weight combinations to obtain K or less multiple route target navigation routes. The navigation route recalling method can be applied to a real-time interaction system of map data, and the recalled multi-route-calculation target navigation route meets the expectation that a plurality of single route-calculation targets are considered simultaneously when a user plans a route.

Before training K sets of single calculation road target weight combinations through a machine learning model, the size of a K value can be determined according to the requirement of recalling an actual navigation route, and it needs to be noted that the larger the K value is, the more calculation road target navigation routes are likely to be obtained during the route recall, but the larger the K value is, the larger the calculation amount of a computer system is, the longer the time consumption is, the more computer resources are occupied, the higher the performance requirement on the computer system of a map data service is, and therefore, in order to meet the time requirement of a real-time interactive system, the K value cannot be set too large.

In one embodiment, referring to fig. 4, in this embodiment, the machine learning model is a ranking model, and the ranking model is obtained by training in advance in the following manner:

s401: setting a loss function by utilizing a linear rectification ReLU function according to a preset navigation target combination; obtaining;

s402: determining a Pareto route set by utilizing a selected multi-objective algorithm according to a plurality of historical route calculation requests; the Pareto route set comprises a plurality of alternative routes which meet the preset multi-target combination requirement; recalling at least one multi-calculation road target navigation route from the Pareto route set according to at least one set of single calculation road target weight combination obtained in initial setting or training;

s403: recalling at least one single calculation road target navigation route according to the plurality of historical road calculation requests and the determined single calculation road targets;

s404: inputting the obtained route set of at least one multi-route target navigation route and at least one single-route target navigation route into a sequencing model, and determining the optimal route in the route set and the optimal route in at least one single-route target navigation route;

s405: and determining a loss value of a loss function according to the attribute value of the optimal route in the route set and the attribute value of the optimal route in the single calculation route target navigation route, adjusting the weight of each single calculation route target of each set of single calculation route target weight combination and the parameters of the sequencing model according to the loss value of the loss function, and repeating the process until at least one set of single calculation route target weight combination and the parameters of the sequencing model when the loss function is minimized are obtained.

In a specific embodiment, assuming that a multi-calculation route target combination is preset, wherein the multi-calculation route target combination comprises five single calculation route targets, namely, the five single calculation route targets with the fastest speed, the shortest distance, the shortest estimated time, the smallest number of traffic lights and the smallest number of turning actions, the attribute values of each alternative route in the Pareto route set obtained correspondingly comprise an avoidance congestion attribute value, a time attribute value, a distance attribute value, a traffic light number attribute value and a turning action number attribute value. And in the training process of the ranking model, adjusting the weight combination of the K sets of single calculation way targets, namely adjusting the weight of each single calculation way target in each set of single calculation way target weight combination in the K sets of single calculation way target weight combinations, wherein the weight sum of the weights of each single calculation way target in each set of single calculation way target weight combination is 1.

In one embodiment, according to K sets of single calculation road target weight combinations obtained in initial setting or training, not more than K multiple calculation road target navigation routes are screened from the Pareto route set. In specific implementation, for example, each set of single calculation route target weight combination in the K sets of single calculation route target weight combinations may be used to weight each alternative route in the Pareto route set, that is, after the weights of the single calculation route targets in each set of single calculation route target weight combination are multiplied by the target attribute values corresponding to each alternative route in the Pareto route set, the sum is obtained to obtain the comprehensive weight of each alternative route in the Pareto route set, all the alternative routes in the Pareto route set are sorted according to the size of the comprehensive weight, and the route with the minimum comprehensive weight is selected as the multiple calculation route target navigation route screened by the set of single calculation route target weight combination. Since the situation that the same multi-calculation road target navigation route is screened out by two sets of single calculation road target weight combinations may exist, when K sets of single calculation road target weight combinations are used for screening out the multi-calculation road target navigation routes from the Pareto route set, K or less multi-calculation road target navigation routes can be screened out.

In a specific embodiment, in setting the loss function, the navigation target combination used includes an optimization target and at least one target as a constraint condition, and the optimization target and the target as the constraint condition are any one or more of the following items:

estimated travel time, travel distance, toll, congestion distance, number of traffic lights in the route, and number of steering actions.

In one embodiment, the combination of navigation objectives may be, for example, that the optimization objective is predicted travel time, and the constraints are travel distance and charges, so as to achieve a control route distance and charges that are not greater than the single-calculation target navigation route obtained by the method according to the prior art, and to achieve the effect of reducing the predicted travel time. At this time, the Loss function may be set to Loss _ eta + lam _ dis _ tf.nn.relu (Loss _ dis _ online _ dis) + lam _ toll _ tf.nn.relu (Loss _ toll _ online _ toll). In the above formula, feature _ eta is a time attribute value of an optimal route in a route set composed of a multi-route target navigation route and a single-route target navigation route, feature _ dis is a distance attribute value of the optimal route in the route set composed of the multi-route target navigation route and the single-route target navigation route, feature _ toll is a charging attribute value of the optimal route in the route set composed of the multi-route target navigation route and the single-route target navigation route, online _ dis is a distance attribute value of the optimal route in the single-route target navigation route, and online _ toll is a charging attribute value of the optimal route in the single-route target navigation route; lan _ dis, tf _ sn.relu (feature _ dis-online _ dis) represents a ReLU function set according to distance for comparing the sizes of feature _ dis and online _ dis, and if feature _ dis-online _ dis is greater than 0, the function value is the real value of feature _ dis-online _ dis, and if feature _ dis-online _ dis is less than or equal to 0, the function value is 0; lam _ total _ tf.nn.relu (feature _ total-online _ total) represents a ReLU function set according to charging for comparing the sizes of the feature _ total and the online _ total, and if the feature _ total-online _ total is greater than 0, the function value is the true value of the feature _ total-online _ total, and if the feature _ total-online _ total is less than or equal to 0, the function value is 0.

The K sets of single-calculation-path target weight combinations are obtained by training according to loss functions set by the navigation target combinations, so that the determination of the weight of each single-calculation-path target of each set of single-calculation-path target weight combination accords with the expected effect of the navigation target combination, and when the multi-calculation-path target navigation route is recalled through the K sets of single-calculation-path target weight combinations, the recalled navigation route comprehensive multi-calculation-path target is superior to the single-calculation-path target navigation route recalled according to the single-calculation-path target recall strategy in the prior art.

In one embodiment, in the training process of the ranking model, the process of determining the optimal route in the route set composed of the at least one multi-route target navigation route and the at least one single-route target navigation route in the ranking model and the optimal route in the at least one single-route target navigation route is as follows:

when a route set consisting of the obtained multi-route target navigation routes and the single-route target navigation routes is input into the ranking model, the ranking model determined according to the parameters of the ranking model can output ranking scores of all navigation routes in the route set, all navigation routes in the route set are ranked according to the ranking scores, the navigation route with the highest ranking score is the optimal route of the route set, and meanwhile, the route with the highest ranking score in at least one single-route target navigation route is the optimal route in the single-route target navigation routes.

In a specific embodiment, the ranking model is a Deep Neural Network (DNN) model, for example, the DNN model is a multidimensional Neural Network interconnection neuron and an activation function. The specific structure of the ranking model may refer to descriptions in the prior art, and in the embodiment of the present invention, this is not particularly limited.

The training process of the ranking model is illustrated below by a specific embodiment:

referring to fig. 5, a ranking model is established, which includes an arithmetic path request obtaining module 501, a Pareto set obtaining module 502, a multi-objective route recall module 503, a single-objective route recall module 504 and a ranking model 505 to be trained, for a plurality of historical arithmetic path requests, for example, a plurality of arithmetic path requests in the early peak period of beijing, a multi-objective navigation combination including five targets of regular fastest speed, shortest distance, shortest predicted time, least number of traffic lights and least number of turning actions is preset, and the preset navigation target combination is: the method achieves the aim of shortening the predicted driving time under the condition of keeping the driving distance and the charging not higher than the single calculated road target navigation route obtained by the existing single calculated road target recall strategy. The Loss function is set to Loss _ eta + lam _ dis _ tf.nn.relu (Loss _ dis _ online _ dis) + lam _ toll _ tf.nn.relu (Loss _ toll _ online _ toll). Specifically, the training process of the ranking model is as follows:

the route calculation request obtaining module 501 obtains a plurality of historical route calculation requests in the early peak time of beijing from historical route calculation request data;

for each historical route calculation request, the Pareto set acquisition module 502 utilizes a multi-objective algorithm to calculate a Pareto route set which accords with a preset multi-calculation route target combination, wherein each alternative route in the Pareto route set is represented by attribute values corresponding to five route calculation targets;

the multi-target route recalling module 503 multiplies the weight of each single calculation route target of each set of single calculation route target weight combination in the initialized K sets of single calculation route target weight combinations by the attribute value of the corresponding target of all the alternative routes in the Pareto set, and then sums to obtain the comprehensive weight of each alternative route, and sorts the comprehensive weights of all the alternative routes from low to high, so that a route with the lowest comprehensive weight can be obtained corresponding to each set of single calculation route target weight combination, namely a multi-calculation route target navigation route; because the condition that the same multi-calculation road target navigation route is recalled by combining a plurality of single calculation road target weights may occur, K sets of single calculation road target weights are combined to obtain K or less multi-calculation road target navigation routes;

while determining K multi-route target navigation routes or less by using a multi-target algorithm and K sets of single-route target weight combinations, the single-target route recall module 504 recalls at least one single-route target navigation route for each historical route calculation request and the determined single-route target by using a single-route target recall strategy;

inputting a route set composed of K multi-route target navigation routes and at least one single-route target navigation route into a ranking model 505 determined according to initialized ranking model parameters, outputting ranking scores of all navigation routes in the route set, sequencing according to the sequence of sequencing scores from big to small, determining the optimal route of the route set and the optimal route in the single-calculation route target navigation route, obtaining the time attribute value, the distance attribute value and the charging attribute value of the optimal route of the route set, and the distance attribute value and the charging attribute value of the optimal route in the single calculation road target navigation route, calculating the loss value of the loss function, according to the loss value of the loss function, adjusting the weight of each single calculation path target weight combination in the K sets of single calculation path target weight combinations and the parameter of the sequencing model;

in the same manner as in the above process, the multi-objective route recall module 503 obtains K multi-calculated-route target navigation routes or less again, inputs a route set composed of the obtained K multi-calculated-route target navigation routes or less and at least one single-calculated-route target navigation route into the ranking model 505 determined according to the trained parameters of the ranking model, further obtains the optimal routes of the corresponding route set and the optimal routes of the single-calculated-route target navigation routes, calculates the loss value of the loss function, and readjusts each single-calculated-route target weight of each single-calculated-route target weight combination in the K single-calculated-route target weight combinations and the parameters of the ranking model according to the loss value of the loss function. And continuously continuing the process until K sets of single-calculation road target weight combinations and parameters of the ranking model when the loss function is minimized are obtained, and ending the training process of the ranking model 505.

In the embodiment of the invention, the training index of the ranking model is the optimal K sets of single-calculation road target weight combination and the parameters of the ranking model when the loss function is minimized by adjusting the K sets of single-calculation road target weight combination and the parameters of the ranking model to reduce the loss value of the loss function. That is to say, in the training process of the ranking model, the loss value of the loss function is optimized by adjusting the weight combination of the K sets of single-computation-path targets and the parameters of the ranking model, and the weight combination of the K sets of single-computation-path targets and the parameters of the ranking model when the loss function is minimized are obtained. When a route calculation request is received, calculating routes by using K sets of single route calculation target weight combinations according to start and end point information carried in the route calculation request, and determining at least one multi-route calculation target navigation route corresponding to each set of single route calculation target weight combination according to a route calculation result; and then sequencing at least one multi-route target navigation route and at least one single-route target navigation route which are input by utilizing a sequencing model determined according to the parameters of the sequencing model obtained by training, and finally determining at least one recommended navigation route pushed to a user.

Based on the same inventive concept, embodiments of the present invention further provide a navigation route obtaining apparatus, a navigation route obtaining system, a related storage medium, and a device, and because the principles of the problems solved by these apparatuses, systems, related storage media, and devices are similar to those of the aforementioned route recalling method, the implementation of the method, apparatuses, systems, related storage media, and devices may refer to the implementation of the aforementioned route recalling method, and repeated details are not repeated.

Based on the same inventive concept, the present embodiment provides a navigation route obtaining device, which may be installed in a local server or a cloud server, and obtains at least one navigation route by executing the navigation route obtaining method, so as to push the navigation route to a navigation client for route planning and navigation. Referring to fig. 6, the navigation route acquisition device includes:

the multi-objective recall module 601 is used for retrieving a multi-computed route target navigation route according to the computed route request and at least one set of single-computed route target weight combination, wherein the set of single-computed route target weight combination comprises weights respectively corresponding to at least two single-computed route targets;

a single-target recall module 602, configured to recall a single-route calculation target navigation route according to the route calculation request and the determined single-route calculation target;

the sorting module 603 is configured to sort the multiple route target navigation routes and the single route target navigation route, and determine a recommended navigation route pushed to the user.

In an embodiment, the multi-objective recall module 601 is further configured to determine, according to a route preference index carried in the route calculation request, a single route calculation objective corresponding to the route preference index;

Wherein the route preference metrics include: the traffic time is shortest, the driving distance is shortest, the number of traffic lights is least, the steering action is least, and congestion is avoided.

In an embodiment, the multi-target recall module 601 is specifically configured to determine, according to start and end point information carried in the route calculation request, route segments included in the multi-target navigation route to be recalled by using each set of single route calculation target weight combination;

In an embodiment, the multi-target recall module 601 is specifically configured to divide a road network into a plurality of grids in advance;

weighting the attribute values of each route calculation target in each route segment corresponding to each grid passing from the starting point to the end point by using the weight of each route calculation target in each set of route calculation target weight combination to obtain the weight of each route segment corresponding to each grid;

In an embodiment, the ranking module 603 is specifically configured to input the at least one multi-route target navigation route and the at least one single-route target navigation route into a trained ranking model, and output ranking scores of the at least one multi-route target navigation route and the at least one single-route target navigation route;

In one embodiment, the navigation route obtaining apparatus may further include a training module 600, configured to obtain the ranking model by:

Wherein the navigation target combination comprises an 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:

Based on the same inventive concept, the present embodiment further provides a navigation route obtaining system, as shown in fig. 6, including: a server 1 and at least one client 2;

the client 2 is used for sending a route calculation request and receiving a recommended navigation route pushed by the server 1;

the server 1 is used for recalling the multi-route target navigation route according to the route calculation request and at least one set of single calculation route target weight combination, wherein the set of single calculation route target weight combination comprises weights respectively corresponding to at least two single calculation route targets; recalling the single route calculation target navigation route according to the route calculation request and the determined single route calculation target; and sequencing the multi-route target navigation route and the single-route target navigation route, determining a recommended navigation route pushed to the user and pushing the recommended navigation route to the client 2.

In the embodiment of the present invention, the server 1 is provided with the navigation route acquisition device described in the above embodiment, and is configured to execute the above navigation route acquisition method, obtain at least one recommended navigation route pushed to the user, and push the recommended navigation route to the client for route planning and navigation.

Based on the same inventive concept, the present embodiment also provides a computer-readable storage medium on which computer instructions are stored, which when executed by a processor implement the navigation route acquisition method as described in the above embodiments.

Based on the same inventive concept, the present embodiment further provides a computer device, including: a processor, a memory for storing processor executable commands; wherein the processor is configured to execute the navigation route obtaining method according to the embodiment.

As will be appreciated by one skilled in the art, embodiments of the present invention may be provided as a method, system, or computer program product. Accordingly, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, optical storage, and the like) having computer-usable program code embodied therein.

The present invention is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each flow and/or block of the flow diagrams and/or block diagrams, and combinations of flows and/or blocks in the flow diagrams and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may 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 flow or flows and/or block diagram block or blocks.

These computer program instructions may 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 instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.

These computer program instructions may 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 functions specified in the flowchart flow or flows and/or block diagram block or blocks.

It will be apparent to those skilled in the art that various changes and modifications may be made in the present invention without departing from the spirit and scope of the invention. Thus, if such modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include such modifications and variations.

Claims

1. A navigation route acquisition method includes:

2. The navigation route acquisition method according to claim 1, wherein the method further comprises:

3. The method of claim 2, wherein the recalling the multi-route target navigation route according to the route calculation request and at least one set of single-route target weight combination comprises:

4. The navigation route obtaining method according to claim 3, wherein the determining, according to the start and end point information carried in the route calculation request, the route segments included in the multi-objective navigation route to be recalled by using each set of single-calculation route target weight combination includes:

dividing a road network into a plurality of grids in advance;

5. The navigation route acquisition method according to claim 2, wherein the route preference index includes:

6. The navigation route acquiring method according to claim 1, wherein the step of ranking the multi-route target navigation route and the single-route target navigation route to determine a recommended navigation route to be pushed to the user comprises:

7. The navigation route acquisition method according to claim 6, wherein the ranking model is obtained by:

8. The navigation route acquisition method according to claim 7, wherein the 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 being selected from the following targets:

9. A navigation route acquisition apparatus, comprising:

the multi-objective recall module is used for retrieving the multi-calculation road target navigation route according to the calculation road request and at least one set of single calculation road target weight combination, wherein the set of single calculation road target weight combination comprises weights respectively corresponding to at least two single calculation road targets;

10. A computer-readable storage medium having stored thereon computer instructions, wherein the instructions, when executed by a processor, implement the navigation route acquisition method according to any one of claims 1 to 8.

11. A computer device, comprising: a processor, a memory for storing processor executable commands; wherein the processor is configured to execute the navigation route acquisition method according to any one of claims 1 to 8.