An automatic driving network car order delay matching method and system considering road network dynamic congestion

Abstract

The application discloses an automatic driving network car-hailing order delay matching method and system considering road network dynamic congestion, and relates to the technical field of vehicle scheduling. The Kuhn-Munkres algorithm is used for matching vehicles on the basis of the time required for the drivable vehicles around the passenger to arrive at the passenger's departure place. Compared with the strategy of matching according to Manhattan distance, the delay matching strategy can shorten the average waiting time of users, and the effect is more obvious when the road is congested, and the number of people waiting for a long time is obviously lower than that of the matching strategy according to Manhattan distance. When the number of service vehicles and passengers is relatively balanced under different delay time lengths, the appropriate selection of the delay time length can greatly reduce the platform loss rate and shorten the waiting time of passengers, and improve the user experience.

Description

Technical Field

[0001] This invention relates to the field of vehicle dispatching technology, and more specifically to a method and system for matching delayed orders for autonomous ride-hailing services that takes into account dynamic road network congestion. Background Technology

[0002] Taxis are a vital component of urban transportation. Traditional research on taxi issues largely focuses on macro-level aspects, such as market competition and regulation, fleet size, and pricing. With the development of GPS and mobile internet technologies, information collection and processing have become more convenient, leading to the emergence of ride-hailing services. Ride-hailing platforms like Didi Chuxing and Uber have become deeply integrated into daily life, and micro-level research on taxis is gradually gaining attention, including vehicle search and dispatching on these platforms. Ride-hailing platforms connect travel demand with taxi supply, using algorithms to match resources. Matching passengers and vehicles is the core of operation, and this matching problem involves complex information from various aspects, including road network conditions, passenger demand, and platform operations. Currently, research on this topic primarily relies on simulation methods.

[0003] In the past, order processing typically employed real-time matching algorithms, with vehicle assignment based on proximity. While real-time matching is easy to implement, its long-term effectiveness has been poor. Many studies, in an effort to simplify the model, have used the Manhattan distance between people and vehicles as the basis for matching, assuming that road travel time remains constant—a misconception. Therefore, addressing these technical issues requires further research by those skilled in the art. Summary of the Invention

[0004] In view of this, the present invention provides a method for introducing a mixed traffic flow of private cars and autonomous ride-hailing vehicles, taking into account road congestion, and calculating the specific time required for vehicles to pick up and drop off passengers based on a delayed batch processing model, and then performing human-vehicle matching based on this.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: A method for delay-based matching of autonomous ride-hailing orders considering dynamic road network congestion includes the following steps: Obtain road network information, initialize the delay matching interval and road surface information update time; The demand for ride-hailing services is collected according to the time delay matching interval, forming a demand set, and feasible autonomous ride-hailing services within the preset range of each demand are searched to form a feasible vehicle set. Obtain traffic flow data and calculate the actual travel time of each road segment at the current moment based on traffic flow and road segment capacity. The shortest path algorithm is used to calculate the shortest path and travel time between each demand in the demand set and the corresponding feasible vehicle in the feasible vehicle set, which are used as the passenger's expected waiting time. With the goal of minimizing the overall waiting time, the KM algorithm is used to globally match the demand set and the feasible vehicle set based on the passengers' expected waiting time, and to determine the service vehicle corresponding to each demand.

[0006] Optionally, it also includes matching vehicles to pick up and drop off passengers at their locations, returning unmatched requests to the request pool to wait for the next matching interval, and marking passengers as lost when the waiting time exceeds a preset threshold.

[0007] Optionally, the actual travel time for each road segment at the current moment is calculated using the BPR function: ; in, For road section The free circulation time, for Time and Section Traffic flow on the road, every Update once, For road section Traffic capacity.

[0008] Optionally, the shortest path algorithm is Dijkstra's algorithm, and the passenger's expected waiting time is determined based on the actual travel time of the road segment in the previous road information update interval.

[0009] Optionally, the matching objective function of the KM algorithm is: ; The constraints include: This means that a passenger can be assigned to a maximum of one vehicle. This means that a vehicle can only be matched with one passenger. , ; in, Indicates vehicle Picking up and dropping off passengers The required time, i.e. the passenger's expected waiting time, when Vehicles were assigned to passengers ,but Conversely, it is 0.

[0010] Optionally, when the number of demands m in the demand set is not equal to the number of vehicles n in the feasible vehicle set, the matching matrix is ​​constructed as a max(m,n) order square matrix by supplementing virtual demands or virtual vehicles, where the time for virtual vehicles to pick up and drop off passengers is set to 0.

[0011] Optionally, private cars can choose the shortest route based on the actual travel time of the road segment in the previous update interval at each road information update time. Their route selection and the route of ride-hailing vehicles together constitute a mixed traffic flow.

[0012] An autonomous ride-hailing order delay matching system that takes into account dynamic road network congestion includes: Road network information acquisition and initialization module: used to acquire road network information, initialize delay matching interval and road surface information update time; Demand Initiation Module: Used to collect ride-hailing travel demands according to the delay matching interval, form a demand set, and search for feasible autonomous ride-hailing vehicles within the preset range of each demand, forming a feasible vehicle set; Actual travel time calculation module: used to obtain traffic flow and calculate the actual travel time of each road segment at the current moment based on traffic flow and road segment capacity; Passenger waiting time calculation module: Used to calculate the shortest path and travel time between each demand in the demand set and the corresponding feasible vehicle in the feasible vehicle set using the shortest path algorithm, as the expected waiting time for passengers; Demand matching module: With the goal of minimizing the overall waiting time, it uses the KM algorithm to globally match the demand set and the feasible vehicle set based on the passengers' expected waiting time, and determines the service vehicle corresponding to each demand.

[0013] Optionally, an execution module is also included: for matched vehicles to go to the passenger's location to pick up and drop off the passenger, unmatched demands return to the demand pool to wait for the next matching interval, and when the waiting time exceeds a preset threshold, it is marked as passenger loss.

[0014] As can be seen from the above technical solution, compared with the prior art, the present invention provides a method and system for matching delayed orders for autonomous ride-hailing vehicles that considers dynamic road network congestion. Based on the expected travel time required for vehicles to pick up passengers, the KM algorithm is used for matching people and vehicles, optimizing the overall matching results and shortening passenger waiting time. Numerical examples show that compared with matching strategies based on Manhattan distance, the strategy proposed in this embodiment demonstrates a relative advantage in shortening passenger waiting time and reducing passenger churn rate, especially during peak congestion times, where the difference in average passenger waiting time is more pronounced. When there are sufficient service vehicles, a short delay can further shorten the average passenger waiting time, but a longer delay will cause unnecessary waiting. When the number of service vehicles and passengers is relatively balanced, appropriately selecting the delay duration will shorten the average passenger waiting time to varying degrees and reduce passenger churn. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0016] Figure 1 This is a schematic diagram of the platform's operation process; Figure 2 This is a schematic diagram of a delayed matching example; Figure 3 This is a schematic diagram of a simulated road network; Figure 4 A diagram illustrating the 24-hour travel demand for private cars and ride-hailing services; Figure 5 This is a diagram showing the speed changes of the road section over 24 hours. Figure 6 This is a diagram illustrating the average waiting time during peak hours. Figure 7 This is a diagram illustrating the average waiting time throughout the day. Figure 8 This diagram illustrates the number of people whose average waiting time exceeds 5 minutes. Figure 9 A diagram illustrating average passenger waiting times during the morning rush hour. Figure 10 A diagram illustrating off-peak passenger waiting times; Figure 11 A diagram showing the average waiting time for the 0s, 60s, and 120s scenarios; Figure 12 The diagram shows the number of passengers lost at 0s, 60s, and 120s. Detailed Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] Example 1

[0019] In a hybrid transportation system of private cars and autonomous ride-hailing vehicles, a novel ride-hailing delay matching method considering dynamic changes in road congestion is proposed based on a delay matching mechanism. The method uses the time required for feasible vehicles around the passenger to reach the passenger's departure point as the basis for vehicle matching, employing the Kuhn-Munkres algorithm. Numerical examples demonstrate that, compared to a strategy based on Manhattan distance, the proposed delay matching strategy can shorten the average waiting time for users, with a more significant effect during congested traffic. Furthermore, the number of people with longer waiting times is significantly lower than with the Manhattan distance-based matching strategy. Under different delay durations, when the number of service vehicles and passengers is relatively balanced, appropriately selecting the delay duration can significantly reduce platform churn and shorten passenger waiting times, thereby improving user experience.

[0020] Specifically, the following steps are included: Obtain road network information, initialize the delay matching interval and road surface information update time; The demand for ride-hailing services is collected according to the time delay matching interval, forming a demand set, and feasible autonomous ride-hailing services within the preset range of each demand are searched to form a feasible vehicle set. Obtain traffic flow data and calculate the actual travel time of each road segment at the current moment based on traffic flow and road segment capacity. The shortest path algorithm is used to calculate the shortest path and travel time between each demand in the demand set and the corresponding feasible vehicle in the feasible vehicle set, which are used as the passenger's expected waiting time. With the goal of minimizing the overall waiting time, the KM algorithm is used to globally match the demand set and the feasible vehicle set based on the passengers' expected waiting time, and to determine the service vehicle corresponding to each demand.

[0021] Furthermore, in this embodiment, the matched vehicle goes to the passenger's location to pick up and drop off the passenger, and the unmatched demand returns to the demand pool to wait for the next matching interval. When the waiting time exceeds a preset threshold, it is marked as passenger loss.

[0022] This embodiment also provides an autonomous driving ride-hailing order delay matching system that takes into account dynamic road network congestion, including: Road network information acquisition and initialization module: used to acquire road network information, initialize delay matching interval and road surface information update time; Demand Initiation Module: Used to collect ride-hailing travel demands according to the delay matching interval, form a demand set, and search for feasible autonomous ride-hailing vehicles within the preset range of each demand, forming a feasible vehicle set; Actual travel time calculation module: used to obtain traffic flow and calculate the actual travel time of each road segment at the current moment based on traffic flow and road segment capacity; Passenger waiting time calculation module: Used to calculate the shortest path and travel time between each demand in the demand set and the corresponding feasible vehicle in the feasible vehicle set using the shortest path algorithm, as the expected waiting time for passengers; Demand matching module: With the goal of minimizing the overall waiting time, it uses the KM algorithm to globally match the demand set and the feasible vehicle set based on the passengers' expected waiting time, and determines the service vehicle corresponding to each demand.

[0023] Furthermore, it also includes an execution module: used for matched vehicles to pick up and drop off passengers at their locations, unmatched requests to return to the request pool to wait for the next matching interval, and when the waiting time exceeds a preset threshold, it is marked as passenger churn.

[0024] Example 2

[0025] This embodiment further illustrates the concept using a specific scenario: Assume the system consists of private cars and ride-hailing vehicles, all of which are autonomous vehicles and centrally dispatched. Consider a road network. , As a road network node, For road segment collection, For the aggregation of vehicles in the road network, private cars and ride-hailing vehicles are respectively used and express, The demand for travel can be categorized into private car demand. and ride-hailing demand Both are generated randomly and dynamically. Individual vehicle service Individual travel needs.

[0026] First, the route selection rules for private cars are given. (Each interval...) This generates a batch of private car demand, which follows a Poisson distribution, originating from... and destination Randomly generated. At that time, private car travelers based on the previous stage Travel time was estimated, and the shortest route was selected for all trips. The travel time for each route segment was calculated using the BPR function. Time and Section The travel time is (1) in, For road section The free circulation time, for Time and Section Traffic flow on the road, every Update once, For road section Traffic capacity.

[0027] The operation process of ride-hailing platforms is as follows: Figure 1 As shown, it is mainly divided into three parts: Requirements gathering: exist At any moment, the traveler Call a ride randomly via a mobile app and provide your starting point for the trip. and the end point The time the request is submitted is the earliest departure time the traveler hopes to take. (Collected by the center) Demand within a time period constitutes a demand set. In terms of demand collection, this embodiment adopts a delayed batch processing method, dividing the time into several intervals. Demand within time intervals such as 30s or 60s is accumulated and matched simultaneously. While batch processing may extend the waiting time for some passengers, it optimizes the matching of people and vehicles from a system perspective, reducing overall passenger waiting time. Figure 2 At 7:00 AM, traveler 1 submits a ride request at point 6. At this time, vehicles A, B, and C are all available, with an arrival time of 2 for each. If an instant matching method is used, vehicle B is randomly assigned to traveler 1. When traveler 2 submits a ride request at 7:01 AM at point 14, there are no available vehicles nearby, and the nearest available vehicle is vehicle C, with a waiting time of 4. The total waiting time is 6. If a batch processing method is used, processing traveler 1 and traveler 2 at 7:01 AM, assigning vehicle A or C to traveler 1 and vehicle B to traveler 2, the total waiting time is reduced to 5, thus shortening the overall waiting time.

[0028] Vehicle search: Assume the central system possesses all basic vehicle information, including whether it's a match, whether it's carrying passengers, its parking location, its current latitude and longitude, and speed. By default, parking is allowed at all nodes without restrictions. Vehicles stop at their current node upon reaching their destination. When a passenger request is received, the central system begins searching for distance requirements. The feasible vehicles within the scope constitute the feasible vehicle set. In this embodiment, the feasible vehicle is an unmanned vehicle that is not paired with another vehicle. If If no feasible vehicles are available within the specified range, the request is returned to the demand pool to await the next round of matching. If the waiting time exceeds [a certain period], the request will be terminated. If a passenger has not yet been matched with a partner, the passenger can cancel the order, which will be recorded as a passenger loss.

[0029] Driver-vehicle matching: Once all demand and available vehicles are determined at a certain point in time, the demand... and feasible vehicles Two sets are formed, and the central task is to match the two sets. Existing literature often directly uses Manhattan distance as the basis for matching. The difference in this embodiment is that it introduces mixed traffic flow, considers road congestion, calculates the waiting time for feasible vehicles to reach passengers, and matches passengers and vehicles based on the principle of minimizing the total waiting time of passengers in the current interval. Each passenger is matched with only one vehicle at a time (this embodiment does not consider carpooling). Unmatched passengers return to the demand pool and wait for the next round.

[0030] In the process of matching people and vehicles, this embodiment uses the Kuhn-Munkres (KM) algorithm with the goal of minimizing the overall waiting time. The time-matching problem is expressed as follows: (2) st (3) (4) (5) in, Indicates vehicle Picking up and dropping off passengers The required time is the passenger's expected waiting time. The shortest route and travel time for pick-up and drop-off are determined using Dijkstra's algorithm. Expected waiting time for a moment Based on the previous stage The road conditions determine that, ,in Indicates vehicle Picking up and dropping off passengers The set of road segments for the shortest path. Constraint 1 means that a passenger can be assigned to at most one vehicle, and constraint 2 means that a vehicle can only be matched with one passenger. The car was assigned to passengers ,but Conversely, it is 0.

[0031] During the algorithm matching process, if the vehicle Then the virtual demand matrix is ​​completed as follows: Rank; if Then complete the virtual vehicle matrix as follows: The time for virtual vehicles to pick up passengers The value is 0. If a passenger is assigned to a virtual vehicle, they enter the waiting area; if a vehicle is assigned to a virtual demand, it means the vehicle is not assigned. After a vehicle is matched, it immediately proceeds to the passenger's location, simultaneously updating the demand pool and vehicle information. Once the passenger boards, the shortest path is calculated using Dijkstra's algorithm, and the passenger travels to their destination.

[0032] Example 3

[0033] The autonomous ride-hailing order matching process proposed in this embodiment is as follows: Step 0: Initialization. Set the delay interval. and road information update time The 24-hour period is divided into equal intervals based on the size of the interval. It is assumed that the vehicle stops at the center position at 00:00:00. The initial allocation is based on the free-flow time of each road segment, with intervals of... Update the speed of the road segment once.

[0034] Step 1: Private car route selection. Private cars select routes based on... The system analyzes road conditions within a given area, estimates travel time, and selects the shortest path between origin and destination (OD) locations. After traffic allocation, it updates the data using the BPR function. Travel time for each segment of the route at any given time, every Updated once.

[0035] Step 2: Requirements Gathering. At any time, the center collects All demands within the specified interval, plus previously unmatched demands, constitute the demand set. Search for current distance to various needs feasible vehicle set within the range .

[0036] Step 3: Calculate the passenger's expected waiting time. Based on the travel time of the road segment calculated in Step 1, Dijkstra's algorithm is used to calculate the shortest path between the passenger and the corresponding feasible vehicle, as well as the path travel time. .

[0037] Step 4: Matching drivers and vehicles. Based on the expected waiting time. Based on this, the KM algorithm is used for passenger-vehicle matching. After completion, unmatched passengers return to the demand pool to wait for the next round of matching. If the time exceeds [a certain threshold], [further action is taken]. If no match is found, the passenger is considered lost. Update vehicle information, collect demand for the next segment, and proceed to Step 2.

[0038] Step 5: The matched vehicles use the Dijkstra algorithm to select the shortest path to reach the destination and stop at the destination.

[0039] The matching strategy proposed in this embodiment is compared with the distance-based matching strategy. Furthermore, the matching strategies for different numbers of service vehicles and different delay intervals will also be compared. The matching results demonstrate the necessity of delayed matching under supply-demand equilibrium conditions. The example was run on a computer configured with an Intel Core i5 3.10GHz processor and 16GB of RAM, and was implemented using Python programming.

[0040] Table 1. Traffic Capacity of Road Sections

[0041] This embodiment adopts Figure 3 The road network shown contains 16 nodes, 48 ​​road segments, and 240 origin-destination (OD) pairs. The length of each road segment is... Free flow time is The road segment capacity is shown in Table 1. It is assumed that vehicles are evenly distributed across nodes 6, 7, 10, and 11 at 00:00:00. The longest distance for vehicle search is... Longest waiting time before passenger matching ,Pick The number of service vehicles was set at 20, 24, 28, 32, and 36 respectively. Travel demand was based on the 2017 US NHTS resident 24-hour travel data distribution, with private car demand taking... The demand for ride-hailing services The demand follows a Poisson distribution, with OD pairs generated randomly every minute. The 24-hour distribution of demand is as follows: Figure 4 As shown.

[0042] Private cars use Dijkstra's algorithm to calculate the shortest path. When , Figure 5 The average speed changes of road segments [1,2], [8,9], [9,10], and [14,10] over a 24-hour period are shown. These segments are more congested between 7:00-10:00 and 16:00-19:00, forming morning and evening rush hours. Figure 4 The distribution of demand basically corresponds.

[0043] The following section will compare the time-based matching strategy proposed in this embodiment with the distance-based matching strategy. , Figure 6 and Figure 7 The average waiting time during peak hours (16:00-19:00) and the average waiting time throughout the day are shown for different numbers of service vehicles under two strategies. As the number of service vehicles increases, the average waiting time for passengers decreases. During peak hours, due to road congestion, the overall average waiting time is longer than the total average waiting time, and the difference between the two strategies is also significant. This is because the Manhattan distance strategy ignores current road congestion and matches users with the vehicle closest to their needs. Although the path distance is short, road congestion leads to a longer waiting time. The strategy proposed in this embodiment takes road conditions into account.

[0044] Figure 8 The number of people whose actual waiting time exceeded 5 minutes under both strategies was statistically analyzed over a 24-hour period. As the number of service vehicles increased, the number of people with waiting times exceeding 5 minutes effectively decreased, and service quality gradually improved. It is evident that the delayed matching strategy proposed in this embodiment resulted in significantly fewer people with actual waiting times exceeding 5 minutes compared to the distance-based matching strategy, indicating a higher overall service quality.

[0045] The difference between delayed matching and instant matching lies in how demand is handled. Instant matching occurs as soon as the system receives an order. Matching is performed, while delayed matching involves waiting for a period of accumulated demand before processing all requests simultaneously. During the morning peak hours (7:00-10:00), the total passenger demand was 392, accounting for 19.5% of the total demand. During the off-peak hours (20:00-22:00), the total passenger demand was 110, accounting for 5.5%. Figure 9 The average passenger wait time during the morning rush hour was compared for different delay durations and different numbers of service vehicles. As the number of service vehicles decreased, the average passenger wait time gradually increased. When there were many service vehicles (36 vehicles), a delay of 15 seconds resulted in the shortest wait time. When the delay exceeded 1 minute, the average passenger wait time surpassed that of the immediate matching solution. This is because when there are sufficient vehicles, delays actually cause unnecessary waiting. When the number of service vehicles is relatively balanced (28 and 32 vehicles), the optimal delay time is 30 seconds. When service vehicles are scarce (20 and 24 vehicles), a delay of 45 seconds results in the shortest average passenger wait time. When there are fewer service vehicles, passengers' inherent wait times are longer; therefore, considering longer-waiting needs together actually leads to higher efficiency.

[0046] Figure 10 The average passenger wait time was compared during off-peak hours. During off-peak hours, total demand is about one-quarter of that during peak hours, and there are relatively sufficient service vehicles, so there is no need for "forced" waiting. The average wait time is shortest at 15 seconds.

[0047] Figure 11 Choose between 60s, 120s, and instant matching strategies A comparison reveals that the average waiting time with a 60-second delay is significantly lower than that with both a 120-second delay and instant matching. When there are fewer service vehicles, the 120-second delay strategy is slightly better than instant matching, indicating that overall optimization is more beneficial to the system when resources are scarce. However, when the number of service vehicles increases, the 120-second delay actually increases user waiting time, with the average waiting time exceeding that of instant matching.

[0048] Figure 12The changes in the number of lost passengers were compared when the delay time was 0s, 60s, and 120s. The overall passenger churn rate was lowest with a 60s delay and highest with instant matching. As the number of taxis increased, the difference in lost passengers between a 120s delay and instant matching gradually narrowed. This is because with more service vehicles, the overall waiting time for passengers is shortened, and a 120s delay actually forces some passengers to wait, resulting in unnecessary churn.

[0049] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to the method section.

[0050] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined in these embodiments may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for matching delayed orders for autonomous ride-hailing services considering dynamic road network congestion, characterized in that, Includes the following steps: Obtain road network information, initialize the delay matching interval and road surface information update time; The demand for ride-hailing services is collected according to the time delay matching interval, forming a demand set, and feasible autonomous ride-hailing services within the preset range of each demand are searched to form a feasible vehicle set. Obtain traffic flow data and calculate the actual travel time of each road segment at the current moment based on traffic flow and road segment capacity. The shortest path algorithm is used to calculate the shortest path and travel time between each demand in the demand set and the corresponding feasible vehicle in the feasible vehicle set, which are used as the passenger's expected waiting time. With the goal of minimizing the overall waiting time, the KM algorithm is used to globally match the demand set and the feasible vehicle set based on the passengers' expected waiting time, and to determine the service vehicle corresponding to each demand.

2. The method for matching delayed orders for autonomous ride-hailing services considering dynamic road network congestion as described in claim 1, characterized in that, It also includes matching vehicles going to the passenger's location to pick up and drop off the passenger, unmatched requests returning to the request pool to wait for the next matching interval, and when the waiting time exceeds a preset threshold, it is marked as passenger churn.

3. The method for matching delayed orders for autonomous ride-hailing services considering dynamic road network congestion as described in claim 1, characterized in that, The actual travel time for each road segment at the current moment is calculated using the BPR function: ； in, For road section The free circulation time, for Time and Section Traffic flow on the road, every Update once, For road section Traffic capacity.

4. The method for matching delayed orders for autonomous ride-hailing services considering dynamic road network congestion as described in claim 1, characterized in that, The shortest path algorithm is Dijkstra's algorithm, and the expected waiting time for passengers is determined based on the actual travel time of the road segment in the previous road information update interval.

5. The method for matching delayed orders for autonomous ride-hailing services considering dynamic road network congestion as described in claim 1, characterized in that, The matching objective function of the KM algorithm is: ； The constraints include: This means that a passenger can be assigned to a maximum of one vehicle. This means that a vehicle can only be matched with one passenger. ； in, Indicates vehicle Picking up and dropping off passengers The required time, i.e. the passenger's expected waiting time, when Vehicles were assigned to passengers ,but Conversely, it is 0.

6. The method for matching delayed orders for autonomous ride-hailing services considering dynamic road network congestion as described in claim 1, characterized in that, When the number of demands m in the demand set is not equal to the number of vehicles n in the feasible vehicle set, the matching matrix is ​​constructed as a max(m,n) square matrix by completing virtual demands or virtual vehicles, where the time for virtual vehicles to pick up and drop off passengers is set to 0.

7. The method for matching delayed orders for autonomous ride-hailing services considering dynamic road network congestion as described in claim 1, characterized in that, Private cars choose the shortest route based on the actual travel time of the road segment in the previous update interval at each road information update time. Their route selection, together with the route of ride-hailing vehicles, constitutes a mixed traffic flow.

8. An autonomous driving ride-hailing order delay matching system considering dynamic road network congestion, characterized in that, include: Road network information acquisition and initialization module: used to acquire road network information, initialize delay matching interval and road surface information update time; Demand Initiation Module: Used to collect ride-hailing travel demands according to the delay matching interval, form a demand set, and search for feasible autonomous ride-hailing vehicles within the preset range of each demand, forming a feasible vehicle set; Actual travel time calculation module: used to obtain traffic flow and calculate the actual travel time of each road segment at the current moment based on traffic flow and road segment capacity; Passenger waiting time calculation module: Used to calculate the shortest path and travel time between each demand in the demand set and the corresponding feasible vehicle in the feasible vehicle set using the shortest path algorithm, as the expected waiting time for passengers; Demand matching module: With the goal of minimizing the overall waiting time, it uses the KM algorithm to globally match the demand set and the feasible vehicle set based on the passengers' expected waiting time, and determines the service vehicle corresponding to each demand.

9. A method for matching delayed orders for autonomous ride-hailing services considering dynamic road network congestion as described in claim 8, characterized in that, It also includes an execution module: used for matched vehicles to pick up and drop off passengers at their locations, unmatched requests to return to the request pool to wait for the next matching interval, and when the waiting time exceeds a preset threshold, it is marked as passenger loss.

Images