A forward-looking method and system for scheduling heterogeneous shared autonomous vehicles

Abstract

The present invention discloses a forward-looking method and system for scheduling carpooling of heterogeneous shared autonomous vehicles, which relates to the field of intelligent transportation technology. The method includes generating candidate matching pairs of shared autonomous vehicle states and requests; constructing a regional value function based on state variables using a Markov decision method, and evaluating the regional value function; constructing a dynamic carpooling matching algorithm based on the interaction between ride requests and shared autonomous vehicles in the scheduling area; based on the evaluation results of the regional value function, using a dynamic carpooling matching algorithm to select the optimal match between the shared autonomous vehicle and the ride request based on the candidate matching pairs, and performing carpooling scheduling based on the optimal matching results. The present invention designs a carpooling scheduling framework that takes into account potential travel demand and overall carpooling efficiency to optimize the different passenger capacities during the operation of heterogeneous SAV fleets.

Description

Technical Field

[0001] The present invention relates to the field of intelligent transportation technology, and in particular to a forward-looking method and system for scheduling carpooling of heterogeneous shared autonomous vehicles. Background Art

[0002] The statements in this section merely provide background information related to the present invention and do not necessarily constitute prior art.

[0003] Ride-sharing systems have emerged as a promising solution to alleviate urban traffic congestion, enabling efficient use of fewer vehicles to meet high travel demands. The introduction of shared autonomous vehicles (SAVs) has garnered significant research attention for these systems. SAVs exhibit superior safety and operational flexibility when handling real-time requests and adjusting trip routes, unaffected by human limitations such as fatigue or distraction. Furthermore, their ability to operate continuously and adhere to precise scheduling significantly improves the efficiency of ride-sharing services. Research suggests that a fleet of SAVs can match the service capacity of a traditional taxi system while using only half the number of vehicles. Depending on the specific characteristics of the transportation network, each SAV could potentially replace seven to eleven private cars, significantly reducing private car ownership. Combining flexible demand management with strict scheduling enforcement, SAVs offer significant economic, environmental, and social advantages to urban transportation systems.

[0004] To address this trend, scholars worldwide have conducted extensive research on various aspects of ridesharing. This research covers a wide range of topics, including trip request allocation, travel time estimation, pickup point recommendation, routing, and pricing strategies. In the early stages, static ridesharing scheduling was the dominant approach. This approach requires all ridesharing requests to be known in advance, enabling optimal matching and maximizing resource utilization. Optimization objectives include minimizing total trip distance, reducing passenger wait time, and maximizing taxi utilization. Most static optimization research focuses on gaining a deeper understanding of the characteristics and impacts of ridesharing. Some researchers have developed a taxi ridesharing method designed to reduce wait and travel times during peak hours. Their results suggest that ridesharing can reduce the distance of each taxi trip by 2–3 kilometers. However, their study was limited to requests where the origin and destination (OD) distance between the two passengers is close. Other researchers have proposed a multimodal autonomous vehicle ridesharing user equilibrium model (MARUE) within a static network equilibrium framework. Based on MARUE, they analyzed the optimal supply decisions of autonomous vehicles (SAVs) in terms of fleet size, fares, routing, and allocation. Overall, static optimization provides a theoretical framework for understanding the impact of ridesharing services on transportation systems and how these services can be optimized. While static optimization offers valuable insights, its practicality is significantly limited by the accuracy of demand forecasts. Furthermore, these methods have limited adaptability to real-time fluctuations in demand and traffic conditions.

[0005] Compared to static approaches, dynamic scheduling enables the system to continuously receive and process new ride-sharing requests, matching them in real time based on current traffic conditions and vehicle locations. This adaptability improves the system's ability to handle changing ride-sharing requests and fluctuating traffic conditions. Current research on dynamic taxi ride-sharing focuses on developing efficient algorithms, as the inherent nature of dynamic ride-sharing requires rapid matching of thousands of trip requests. For example, some researchers have proposed a dynamic ride-sharing method capable of matching a large number of trip requests with taxis in real time. Similarly, some researchers have introduced a mixed integer programming model to analyze the ride-sharing problem and proposed a Lagrangian decomposition method and two heuristic methods for efficiently solving the model. Other researchers have proposed a method that only allows passengers from specific taxi hotspots to share rides with passengers traveling to similar destinations. Their research shows that 48% of ride-sharing requests from hotspots can be shared, reducing driving distance by 1.2 kilometers per shared trip. The ride-sharing matching problem is a significant technical challenge for SAV systems, as it is crucial for optimizing the response to overall travel demand in both temporal and spatial dimensions. This adaptability improves the system's ability to handle changing ridesharing requests and fluctuating traffic conditions.

[0006] However, ridesharing platforms still face key operational efficiency challenges, primarily manifesting in excessive passenger wait times and prolonged vehicle idle time. These inefficiencies not only impact service quality but also lead to higher energy consumption and environmental concerns. Research indicates that approximately 80% of Manhattan taxi trips can be shared by two passengers, highlighting the inefficiencies of the current taxi operation model. Advances in autonomous driving technology have led to the emergence of compact, two-seater SAVs (SAVs), such as the Cybercab. While extensive research has focused on designing more efficient ridesharing strategies, the vehicle type allocation problem for heterogeneous SAV fleets remains a critical yet understudied area, requiring optimization based on specific trip characteristics. This factor plays a central role in determining the performance of the overall ridesharing system. Furthermore, traditional scheduling methods have significant limitations. These methods are too short-sighted in their decision-making process and lack forward-looking planning for future ridesharing opportunities, thus limiting optimal scheduling outcomes. Notably, many studies assume the availability of complete trip information. However, this assumption is fundamentally mismatched with actual operational conditions, where real-time order changes significantly impact vehicle routing and dynamic matching decisions. Consequently, existing research still lacks an understanding of the complex interrelationships between trip characteristics, underlying travel demand, and the overall efficiency of ridesharing systems. Summary of the Invention

[0007] In response to the shortcomings of the existing technology, the purpose of the present invention is to provide a forward-looking heterogeneous shared autonomous vehicle carpooling scheduling method and system. By designing a carpooling scheduling framework, potential travel demand and overall carpooling efficiency are considered to optimize the different passenger capacities during the operation of heterogeneous SAV fleets.

[0008] In order to achieve the above object, the present invention is implemented through the following technical solutions:

[0009] A first aspect of the present invention provides a forward-looking method for scheduling heterogeneous shared autonomous driving vehicles, comprising the following steps:

[0010] Obtain ride requests, shared autonomous vehicle status, and road network information within the dispatch area;

[0011] Generate candidate matching pairs of shared autonomous vehicle status and request based on the ride request and shared autonomous vehicle status combined with road network information;

[0012] Determine the state variables of the regional value function based on the historical order information of the dispatching area, construct the regional value function based on the state variables using the Markov decision method, and evaluate the regional value function;

[0013] Build a dynamic ride-sharing matching algorithm based on the interaction between ride requests and shared autonomous vehicles within the dispatch area;

[0014] Based on the evaluation results of the regional value function, a dynamic carpooling matching algorithm is used to select the optimal match between shared autonomous vehicles and ride requests according to candidate matching pairs, and carpooling scheduling is performed based on the optimal matching results.

[0015] Furthermore, the specific steps of generating a candidate matching pair of the shared autonomous vehicle status and the request based on the ride request and the shared autonomous vehicle status combined with the road network information are as follows:

[0016] Setting request acceptance conditions based on the relationship between ride requests, shared autonomous vehicle status, and road network information;

[0017] Determine the order of boarding and disembarking based on the shared autonomous vehicle status, based on whether the ride request meets the request acceptance conditions;

[0018] Determine the shortest path for the shared autonomous vehicle based on the pick-up and drop-off sequence, and identify candidate matching pairs of the shared autonomous vehicle state and the request.

[0019] Furthermore, the Markov decision method is used to construct the regional value function according to the state variables, and the specific steps of evaluating the regional value function are as follows:

[0020] Divide the dispatching area into several small areas;

[0021] The Markov decision method is used to model and estimate the regional value function of the state variables of a small area. The temporal difference learning method is introduced to estimate the regional value function, and the parameters are updated based on the Bellman equation.

[0022] Furthermore, the regional value function represents the long-term return that a shared autonomous vehicle can obtain after entering the current area within a preset time period.

[0023] Furthermore, the state variables of the regional value function include the number of orders departing from the current region, the number of orders destined for the current region, the number of shared autonomous vehicles in the current region, and the average order waiting time in the current region.

[0024] Furthermore, the specific steps for constructing a dynamic ride-sharing matching algorithm based on the interaction between ride requests and shared autonomous vehicles within the dispatch area are as follows:

[0025] Transform the real-time ride-sharing scheduling problem using a bipartite graph matching algorithm;

[0026] Based on the matching and compatibility between shared autonomous vehicles and ride requests within the dispatch area, the advantage function is determined by considering multiple factors and constraints are set.

[0027] Furthermore, various factors are considered including:

[0028] Satisfy the most ride requests;

[0029] Reduce passenger waiting time and detour distance;

[0030] Improve the passenger load and service success rate of shared autonomous vehicles;

[0031] Incorporate long-term benefits into scheduling decisions.

[0032] A second aspect of the present invention provides a forward-looking heterogeneous shared autonomous vehicle carpooling scheduling system, comprising:

[0033] a data acquisition module configured to acquire ride requests, shared autonomous vehicle status, and road network information within a dispatch area;

[0034] a candidate matching module configured to generate a candidate matching pair of a shared autonomous vehicle state and a request based on the ride request and the shared autonomous vehicle state in combination with road network information;

[0035] a regional value evaluation module configured to determine state variables of a regional value function based on historical order information of a dispatching area, construct a regional value function based on the state variables using a Markov decision method, and evaluate the regional value function;

[0036] a matching algorithm module configured to construct a dynamic ride-sharing matching algorithm based on the interaction between ride requests and shared autonomous vehicles within a dispatch area;

[0037] The carpooling scheduling module is configured to select the optimal match between the shared autonomous driving vehicle and the ride request based on the evaluation results of the regional value function using a dynamic carpooling matching algorithm according to candidate matching pairs, and perform carpooling scheduling based on the optimal matching results.

[0038] A third aspect of the present invention provides a medium having a program stored thereon, which, when executed by a processor, implements the steps of the forward-looking heterogeneous shared autonomous vehicle carpooling scheduling method as described in the first aspect of the present invention.

[0039] The fourth aspect of the present invention provides a device comprising a memory, a processor, and a program stored in the memory and executable on the processor. When the processor executes the program, the steps of the forward-looking heterogeneous shared autonomous driving carpooling scheduling method as described in the first aspect of the present invention are implemented.

[0040] One or more of the above technical solutions have the following beneficial effects:

[0041] This paper discloses a forward-looking method and system for scheduling heterogeneous shared autonomous vehicles (SAVs). It also proposes a forward-looking ridesharing scheduling framework for heterogeneous SAVs to improve the overall operational efficiency of heterogeneous SAV fleets in urban environments. This framework combines a rule-based approach with a Markov decision process (MDP). Through the collaborative work of three modules: ridesharing matching, path planning, and hotspot guidance, it achieves intelligent SAV scheduling and optimal resource allocation. A key feature of the framework lies in its consideration of the heterogeneous nature of SAVs, namely, differentiated scheduling decisions for vehicles of varying capacities, thereby ensuring passenger satisfaction while improving overall system efficiency. Furthermore, by constructing a regional value function and incorporating a temporal difference learning method, the system can perceive changes in urban travel demand in real time and guide idle SAVs to high-potential areas to improve order acceptance rates and resource utilization.

[0042] Advantages of additional aspects of the present invention will be given in part in the following description and in part will be obvious from the following description, or will be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] The accompanying drawings, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0044] Figure 1 This is a flow chart of a forward-looking method for scheduling heterogeneous shared autonomous driving cars for carpooling in the first embodiment of the present invention;

[0045] Figure 2 This is a schematic diagram of the change in the service success rate of the SAV at different times of the day in the first embodiment of the present invention;

[0046] Figure 3 This is a comparison chart of energy consumption changes during one day for the three methods in Example 1 of the present invention;

[0047] Figure 4 This is a graph showing a trend of the utilization rate of the SAV during a day in the first embodiment of the present invention;

[0048] Figure 5 This is a trend diagram of the seat occupancy rate of the SAV during a day in Example 1 of the present invention;

[0049] Figure 6 1 is a comparison chart of the distribution ratios of SAV passenger status under three methods in Example 1 of the present invention. DETAILED DESCRIPTION

[0050] It should be noted that the following detailed descriptions are exemplary and intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the art to which the present invention belongs.

[0051] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit the exemplary embodiments according to the present invention. As used herein, unless the context clearly indicates otherwise, the singular form is intended to include the plural form. In addition, it should be understood that when the terms "comprise" and / or "include" are used in this specification, they indicate the presence of features, steps, operations, devices, components and / or their combinations;

[0052] Example 1:

[0053] The first embodiment of the present invention provides a forward-looking method for scheduling heterogeneous shared autonomous driving vehicles. Figure 1 As shown, the following steps are included:

[0054] Step 1: Obtain ride requests, shared autonomous vehicle status, and road network information within the dispatch area.

[0055] (1) Request for ride.

[0056] In this embodiment, the set of all passenger requests is defined as The study period is divided into n time periods of equal length, and the subset of ride requests in each time period t is expressed as ,in:

[0057] .

[0058] Each ride request is recorded as , a tuple consisting of travel characteristics Indicates. Among them, The time the request was submitted. and They are the boarding and alighting locations, is the requested number of passengers. Indicates whether the passenger accepts carpooling, and its value remains unchanged throughout the planning period. Otherwise, there is . Indicates whether the passenger has been served. If the passenger has been served in time period t, then , the request will be removed from the matching pool; if it is not served, there is , the request will enter the matching pool of the next period for further matching. Indicates the number of time periods during which the request has not been met. If a request is not met within multiple consecutive time periods, the passenger cancels the request and chooses another mode of transportation, which is considered a failed rideshare. For the sake of generality, this embodiment sets the limit to 8 time periods.

[0059] (2) SAV state.

[0060] For SAV, the state of the j-th SAV in time period t is recorded as a tuple .in, is the position of SAV at time period t; is the number of passengers in the current car; The present invention considers two types of vehicles: low capacity , high capacity ; It is a 0-1 variable, indicating whether SAV is currently in service; if it is in service, ;otherwise, ; Indicates the number of consecutive idle time periods; if the SAV is idle for multiple consecutive time periods, it will be guided to the high-demand area. For the sake of generality, this embodiment is set to 8 time periods. Represents the current driving path of the SAV; the path consists of a time series of nodes, including the path from the current location to the boarding point The shortest path node set, and the shortest path node set from the boarding point to the disembarkation point The shortest path node set. The path is dynamically updated over time.

[0061] To indicate the dynamic evolution of the route, the symbol “-” is used to indicate movement between locations. For example, if the SAV is servicing a request , whose current path is If a new carpooling request is received at this time , and successfully shared the itinerary with it, then the feasible path at this time can be expressed as .

[0062] (3) Road network.

[0063] The road network is recorded as a tuple R, which is expressed as .in, is the set of road nodes, is the set of road edges, is a set of weights of road edges, which is used to calculate the shortest path and energy consumption between two locations.

[0064] In most vehicle scheduling research, graph theory is often used for modeling, while ignoring the specific structure of the road network. Typically, the distance between adjacent boarding points is calculated using the Euclidean distance formula. However, in dynamic ride-sharing scheduling, incorporating the road network as a key component of the problem solution is crucial, consistent with real-world scenarios. Therefore, in this embodiment, during the scheduling decision-making process, the distance between any two locations is determined using the road network's adjacency matrix, and the Dijkstra algorithm is used to calculate the shortest path, effectively evaluating the SAV's minimum driving distance and energy consumption.

[0065] After obtaining the above information, as attached Figure 1 As shown, the dynamic carpooling framework proposed in this embodiment specifically includes three steps: regional value learning, candidate matching pair generation, and real-time scheduling of carpooling.

[0066] The generation of candidate matching pairs aims to provide effective options for real-time carpooling scheduling, and to achieve efficient carpooling services by utilizing the regional value function and other indicator results. , perform a search to identify a set of SAV candidates that can satisfy the request This will form the initial set of SAV-order pairs ,in ,satisfy , . When a new ride request is added to the current schedule of an SAV, a detour may be required to reach the requested pick-up location. In order to optimize the travel efficiency of the SAV, its path is determined by the shortest path algorithm. In addition, the feasibility of the carpooling method is evaluated by calculating the detour ratio between existing requests and new requests. Ultimately, the evaluation process will determine a set of valid SAV-ride request pairs. By generating, screening and evaluating candidate pairs, high-quality, feasible matching pairs can be carefully selected, providing a solid foundation for real-time carpooling scheduling decisions.

[0067] Evaluating regional value functions plays a crucial role in estimating travel demand and passenger value for each region and time period. This process is achieved by establishing a value function model. First, the study area is divided into multiple subregions. Subsequently, using historical taxi travel demand data, the value function for each region is estimated using the Bellman equation and the temporal difference (TD) method. This method combines the update rule of the Bellman equation with the TD incremental learning approach to assess travel demand and value for each region. Through continuous iteration and learning, the accuracy of regional value function estimation can be improved, providing a more reliable basis for candidate matching pair generation and ridesharing scheduling decisions.

[0068] The main goal of real-time carpooling scheduling is to determine the optimal carpooling solution and scheduling plan based on candidate matching pairs. In real-world scenarios, SAVs typically receive multiple ride requests, and each request may form multiple valid matching pairs with multiple SAVs. After candidate matching is generated for each ride request, a matching graph is constructed to show the relationship between different candidate matching pairs. In order to evaluate the quality of these matchings, a weight function is introduced in the matching graph as the edge weight (i.e. ). This weighting function takes into account several key factors, including: the SAV's energy consumption, detour distance, waiting time, and the SAV's income. Together, these factors provide a comprehensive evaluation of each candidate matching pair. Ultimately, a method based on the Kuhn-Munkres algorithm (also known as the Hungarian algorithm) is used to select the SAV-ride request matching pair that maximizes the overall benefit. If an SAV is idle for eight consecutive time periods, the system will evaluate its current time period and location. Based on this evaluation, the system will select a high-demand area as the SAV's guidance destination. After receiving the guidance instruction, the SAV will go to the area with a higher ride request generation rate based on the path information provided by the system.

[0069] Step 2: Generate candidate matching pairs of shared autonomous vehicle status and request based on the ride request and shared autonomous vehicle status combined with road network information.

[0070] The generation of candidate matching pairs is a preprocessing process that aims to narrow the solution space by identifying and eliminating infeasible SAV-ride request matching pairs.

[0071] Step 2.1: Set the request acceptance conditions based on the relationship between the ride request, the shared autonomous vehicle status, and the road network information.

[0072] Consider a SAV whose current position is , it has accepted a ride request , and can reach the requested boarding point at time t1 At this time, if the SAV receives a new ride request in time period t , it will accept this request if the following conditions are met:

[0073] Condition 1: The number of remaining seats in the SAV is greater than or equal to the requested number of passengers.

[0074] (1).

[0075] Condition 2: The actual distance between the current SAV location and the new requested boarding point does not exceed the threshold .

[0076] (2).

[0077] Condition 3: The driving path deviation between two locations is defined as the ratio of the shortest driving distance to the actual driving distance. The driving path deviation of the SAV from the current location to the new requested boarding and alighting point does not exceed the threshold , as shown in formula (3). represents the shortest driving distance from node x to node y, Represents the actual driving distance from node x to node y.

[0078] (3).

[0079] Condition 4: If the SAV has not arrived , the path offset of the first request must meet the threshold condition (4).

[0080] ,if (4).

[0081] Condition 5: When the SAV has arrived When , the path offset of the first request must meet the threshold condition (5).

[0082] ,if (5).

[0083] in, Indicates the moment when the vehicle receives the first request. Indicates the current time exist When and before, Indicates the current time exist after.

[0084] Step 2.2: Determine the boarding and disembarking order based on the shared autonomous vehicle status according to whether the ride request satisfies the request acceptance conditions.

[0085] If requested When conditions 1 and 2 are met, the possible boarding and alighting sequences can be obtained. The specific order of the two requests needs to be planned based on the current state of the SAV. However, different boarding and alighting sequences will result in different path deviations, so additional checks are required based on conditions 3-5.

[0086] In a specific embodiment, when the SAV receives the request If it has not arrived yet , there are four possible get-on and get-off sequences:

[0087] ,

[0088] ,

[0089] ,

[0090] .

[0091] When the SAV has arrived There are two possible boarding and alighting sequences:

[0092] ,

[0093] .

[0094] Step 2.3: Determine the shortest path for the shared autonomous vehicle based on the pick-up and drop-off sequence, and determine candidate matching pairs of the shared autonomous vehicle state and the request.

[0095] Based on the limited permutation schemes mentioned above, this embodiment uses an enumeration method to determine the shortest path for SAV. Once a path is selected, SAV will strictly follow the path.

[0096] In time period t, the initial SAV-request pair set can be obtained through conditions 1 and 2 ,in To ensure feasibility, additional checks are required using conditions 3-5 to obtain a valid SAV-request pair set. , .

[0097] Step 3: Determine the state variables of the regional value function based on the historical order information of the scheduling area, use the Markov decision method to construct the regional value function based on the state variables, and evaluate the regional value function.

[0098] Step 3.1: Determine the state variables of the regional value function based on the historical order information of the scheduling area.

[0099] In this embodiment, the regional value function represents the long-term return that a shared autonomous vehicle can earn after entering the current region within a preset time period. The state variables of the regional value function include the number of orders originating from the current region, the number of orders destined for the current region, the number of shared autonomous vehicles in the current region, and the average order wait time in the current region.

[0100] Step 3.2: Use the Markov decision method to construct the regional value function based on the state variables and evaluate the regional value function.

[0101] Step 3.2.1: Divide the scheduling area into several small areas.

[0102] In a specific implementation, in order to guide SAV to high-demand areas, this embodiment introduces a regional value function. This function can quantitatively evaluate the region based on the historical order information of each region. The focus area is divided into m small areas. Represents the set of all regions. In each time period t, each region Each has a corresponding state representation, which consists of a vector of the following state variables:

[0103] (6).

[0104] in, is the number of orders departing from this area during period t, is the number of orders destined for this region in period t, is the number of SAVs in the area during time period t, is the average order waiting time in the area during period t.

[0105] Step 3.2.2: Use the Markov decision method to model and estimate the regional value function of the state variables of the small area.

[0106] Among them, the temporal difference learning method is introduced to estimate the regional value function, and the parameters are updated based on the Bellman equation.

[0107] In a specific embodiment, the state variables are input into a Markov decision process (MDP) to The regional value function is defined as , represents the long-term return that SAV can obtain after entering the region at time period t. The initial value of each region is set to 0, and it is updated as the data is traversed, and eventually converges to the region value.

[0108] To estimate the region value function, this embodiment introduces a temporal difference (TD) learning method, which is implemented by introducing the time series s and performing parameter updates based on the Bellman equation. The Bellman update equation is as follows:

[0109] (7).

[0110] in, is the learning rate; is the instantaneous return of the current period t; is the discount factor, which determines the impact of future events on the Markov process and is set to 0.9 in the present invention; is the estimated value of the next state.

[0111] Immediate returns This includes factors such as the number of orders and passengers the SAV has completed in the current area, as well as the average waiting time for passengers. Through multiple rounds of training, the system can accurately learn the value trends of each area and then provide optimized guidance recommendations for the SAV's idle state.

[0112] Step 4: Build a dynamic ride-sharing matching algorithm based on the interaction between ride requests and shared autonomous vehicles within the dispatch area.

[0113] Step 4.1: Use the bipartite graph matching algorithm to transform the real-time carpooling scheduling problem.

[0114] Step 4.2: Determine the advantage function based on the matching and compatibility between shared autonomous vehicles and ride requests within the dispatch area, taking into account multiple factors, and set constraints.

[0115] Consider various factors, including: satisfying the most ride requests; reducing passenger waiting time and detour distance; improving the passenger load factor and service success rate of shared autonomous vehicles; and incorporating long-term benefits into scheduling decisions.

[0116] The main goal of the ride-sharing scheduling problem proposed in this example is to achieve a coordinated and optimized overall benefit by finding the best match between shared autonomous vehicles (SAVs) and ride requests in each time period. This optimization aims to maximize the utilization of SAVs while minimizing their energy consumption.

[0117] In a specific embodiment, to solve this problem, the real-time ride-sharing scheduling problem is transformed into an online bipartite graph matching problem. By using a bipartite graph matching algorithm, an optimal matching solution can be determined for the problem.

[0118] In this embodiment, the bipartite graph used consists of two types of nodes: SAV nodes and ride request nodes. SAV nodes represent currently available SAVs, while ride request nodes represent ride requests waiting for service. These two types of nodes are connected. These edges are assigned weights, representing the matching or compatibility between the SAV and the ride request. These weights take into account various factors, including the SAV's energy consumption, the distance between the pick-up and drop-off, and the SAV's current and future income. The carpooling scheduling problem for each time period can be expressed as:

[0119] (8).

[0120] The constraints are as follows:

[0121] (9),

[0122] (10).

[0123] in, Indicates whether the match is successful. Advantage function represents the matching evaluation between SAV j serving ride request k and ride-sharing request i; 0-1 variable The value of 1 indicates a successful match, and the value of 0 indicates no match. Formula (9) indicates that each SAV can only serve one carpooling request at most; Formula (10) indicates that each carpooling request can only be served by one SAV at most.

[0124] Ride sharing depends on the interaction between ride requests and SAVs within the dispatch area. The matching results are reflected in the advantage function of the ride sharing model. The ride sharing model designed in this embodiment comprehensively considers the following aspects: 1) satisfying the most ride requests; 2) reducing passenger wait times and detour distances, improving travel efficiency and user experience; 3) increasing SAV passenger load factors and service success rates, achieving efficient resource utilization and reducing energy consumption; and 4) incorporating long-term benefits into dispatch decisions.

[0125] Considering the above aspects, the following advantage function is proposed:

[0126] (11).

[0127] By solving the advantage function, the most suitable scheduling solution is found. It has a positive impact on the advantage function, meaning that serving ride requests with high-value areas can generate greater revenue. represents the actual arrival time and location of SAV j after serving ride requests i and k. The function has a negative impact on the advantage function. SAVs in areas with lower value are more likely to be selected to serve ride requests. represents the time and location at which SAV j arrives after serving ride request k. It represents the immediate benefit of SAV j after serving ride request i and has a positive impact on the advantage function. It represents the energy consumption of SAV j in providing services to ride request i and has a negative impact on the advantage function. Indicates the resource utilization of two ride requests sharing. A smaller value indicates a higher matching degree. Indicates that the pick-up and drop-off points of two ride requests are at the same location. and Represent the completion time periods of carpooling requests i and k respectively. Usually But if , indicating that the driving path of the SAV does not change after the inclusion of the carpooling request i.

[0128] Step 5: Based on the evaluation results of the regional value function, a dynamic carpooling matching algorithm is used to select the optimal match between the shared autonomous driving car and the ride request based on the candidate matching pairs, and carpooling scheduling is performed based on the optimal matching results.

[0129] In a specific embodiment, the energy consumption of the SAV is mainly affected by two factors: the distance traveled and the type of the SAV vehicle, which can be specifically expressed as follows:

[0130] (12).

[0131] Where, represents the energy consumed by the n-type SAV traveling from node x to node y; represents the energy consumption per unit distance of the n-type SAV, in liters per kilometer (L / km); Represents the shortest distance between nodes x and y.

[0132] Consider a scenario where the position of SAV j when receiving ride request i is At this time, the energy unit price is expressed as If the SAV service distance is within two kilometers, the passenger will be charged a fixed total fee. On the other hand, if the distance traveled exceeds 2 kilometers, the passenger must pay an additional fee, which is Obviously, the energy cost of the SAV accepting a ride request and traveling from its current location to the requested ride location can be calculated as:

[0133] (13).

[0134] Similarly, the energy cost of the SAV to complete the ride request can be defined as:

[0135] (14).

[0136] The total energy cost of the SAV to complete the ride request is:

[0137] (15).

[0138] The direct revenue of SAV after completing a ride request is the amount the passenger needs to pay, as shown in the following formula:

[0139] (16).

[0140] Although carpooling can increase SAV utilization, reduce energy consumption, and improve passenger service efficiency, it can also negatively impact service quality, primarily in two ways: first, detours may occur during service, and second, carpooling's resource utilization efficiency may be low. Therefore, when carpooling occurs, the system feeds the relevant penalty value back into the advantage function to minimize detours and improve carpooling efficiency. The penalty term is defined as follows:

[0141] Low carpool utilization penalty:

[0142] (17).

[0143] Detour distance penalty:

[0144] (18).

[0145] Total penalties:

[0146] (19).

[0147] in, Smaller values indicate a more favorable carpooling match between two ride requests. Conversely, larger values indicate less favorable carpooling results, indicating a decreased likelihood of a shared ride between the two requests. By analyzing and planning carpooling results, we can obtain an optimal scheduling solution, significantly improving the efficiency of the carpooling process and the passenger experience.

[0148] To evaluate the performance of the proposed model, this embodiment conducts an empirical study based on real taxi operation data in Manhattan, New York City. The data comes from the New York City Taxi and Limousine Commission (NYC TLC) and contains passenger pick-up and drop-off information, travel time, fares, and number of passengers for the entire day on June 1, 2022. The Manhattan area is further divided into 265 small areas to construct a spatial network. The present invention uses travel records during the peak period from 7:00 to 9:00 on that date, which includes a total of 40,823 ride requests. Based on the actual road structure of New York City, a directed graph containing 4,092 nodes and 9,453 road edges is constructed as the road network in the study.

[0149] In the experiment, two vehicle types were considered: SAV-A with a capacity of 2 passengers and SAV-B with a capacity of 4 passengers. The vehicles were initially evenly distributed across different road nodes. The simulation scheduling cycle was 1 minute, and the entire scheduling process lasted 120 time slots (i.e., 2 hours). The parameters for generating candidate matching pairs were set as follows: the maximum detour ratio threshold accepted by passengers was 1.3, and the maximum additional detour ratio accepted by the SAV was 1.5.

[0150] To evaluate the effectiveness of the carpooling method proposed in this example, an experiment was conducted using three SAV service methods: the SAV carpooling method proposed in this invention, a variant of the SAV carpooling method, and a non-carpooling method, defined as MSRM, VSRM, and NRM, respectively. VSRM is defined as matching two ride requests whose origin and destination distances fall within a certain threshold. For the initial ride request matching process, MSRM and VSRM employ the same methodology as NRM. To evaluate the performance of the proposed method in optimizing the dynamic carpooling scheduling problem, the following evaluation metrics were used: service success rate (SSR), energy consumption (EC), average energy consumption per passenger (AEC), SAV utilization rate (TUR), pickup time (PT), seat occupancy rate (OT), SAV company net profit (TCNP), ride-hailing platform profit (RPP), and average passenger cost (APC). AEC is defined as total energy consumption divided by the number of passengers served, TUR is defined as the distance an SAV transports passengers from their origin to their destination divided by the total distance traveled by the vehicle, and APC is defined as total passenger fares divided by the number of passengers carried.

[0151] Figure 2The figure shows the variation in the service success rate of SAVs at different times of the day. The service success rate is defined as the percentage of completed ride requests out of the total number of requests. Requests that remain unanswered for more than four minutes are considered service failures. The figure shows that the MSRM method generally outperforms other methods in terms of service success rate. In contrast, the service success rate of the NRM method fluctuates significantly, while the MSRM method maintains a relatively stable service level. This demonstrates that the MSRM method is more robust and can effectively cope with dynamic fluctuations in ride demand. The MSRM method's advantages are even more pronounced during periods of high ride request volume, particularly the evening peak (6:00 PM–10:00 PM), demonstrating its ability to ensure service quality under high demand conditions. The MSRM method better meets passenger needs and improves overall service quality during demand surges. This advantage is primarily due to the fact that the MSRM method optimizes SAV utilization efficiency and passenger travel experience from a global and long-term perspective within its scheduling strategy. It achieves a better trade-off between accepting new requests and controlling the degree of detours, thereby improving overall SAV utilization and service efficiency. Furthermore, the VSRM method's service success rate surpasses the NRT method, further validating the effectiveness of the carpooling strategy in improving service coverage.

[0152] Figure 3 The energy consumption of the three methods over the course of a day is shown. Overall, for each method, SAV's energy consumption is positively correlated with the number of ride requests it completes. During peak ride demand periods, SAV's energy consumption increases significantly due to the large number of requests; conversely, energy consumption decreases during periods of lower demand. Compared to VSRM, MSRM completes more service requests with lower energy consumption: the number of service requests increases by 4.0%, but energy consumption decreases by 1.43%. Compared to NRT, MSRM performs even better, with an 11.6% increase in service requests and a 3.67% decrease in energy consumption.

[0153] Figure 4 and Figure 5 The daily trends of SAV utilization and seat occupancy are shown. It is clear that MSRM consistently outperforms VSRM and NRT in both SAV utilization and seat occupancy. This finding suggests that MSRM's advantage lies in its efficient scheduling and utilization of SAV resources, thereby improving vehicle utilization efficiency. Ride-sharing services demonstrate significant energy-saving advantages, particularly in medium- and long-distance travel scenarios.

[0154] Figure 6 The distribution of SAV passenger capacity under the three methods is shown. NRT mode achieved a 90.1% passenger capacity for SAVs, while VSRM achieved a 70% capacity. MSRM significantly improved SAV passenger capacity. However, 86.5% of the time, each vehicle still had fewer than two passengers.

[0155] This paper, belonging to the field of intelligent transportation technology, provides a forward-looking scheduling framework that combines a rule-based approach with a Markov decision process (MDP) to enable real-time ride-sharing operations. Built around three interrelated components: ride-sharing matching, route planning, and hotspot guidance, the framework aims to improve operational efficiency while maintaining an optimal balance between immediate passenger satisfaction and the long-term benefits of SAVs. A comprehensive case study demonstrates the effectiveness of the framework using real-world data from New York City taxi operations and the city's road network. Experimental results demonstrate significant improvements over traditional approaches in multiple areas, including a 4% increase in service success rate, a 1.43% reduction in energy consumption, a 4.1% increase in average vehicle load factor, and a 16.8% increase in the number of passengers served per SAV. These results strongly demonstrate the transformative potential of heterogeneous SAV services and provide valuable insights for transportation planners to implement more efficient and sustainable urban mobility solutions.

[0156] Example 2:

[0157] A second embodiment of the present invention provides a forward-looking heterogeneous shared autonomous driving carpooling scheduling system, including:

[0158] a data acquisition module configured to acquire ride requests, shared autonomous vehicle status, and road network information within a dispatch area;

[0159] a candidate matching module configured to generate a candidate matching pair of a shared autonomous vehicle state and a request based on the ride request and the shared autonomous vehicle state in combination with road network information;

[0160] a regional value evaluation module configured to determine state variables of a regional value function based on historical order information of a dispatching area, construct a regional value function based on the state variables using a Markov decision method, and evaluate the regional value function;

[0161] a matching algorithm module configured to construct a dynamic ride-sharing matching algorithm based on the interaction between ride requests and shared autonomous vehicles within a dispatch area;

[0162] The carpooling scheduling module is configured to select the optimal match between the shared autonomous driving vehicle and the ride request based on the evaluation results of the regional value function using a dynamic carpooling matching algorithm according to candidate matching pairs, and perform carpooling scheduling based on the optimal matching results.

[0163] Example 3:

[0164] Embodiment 3 of the present invention provides a medium having a program stored thereon, which, when executed by a processor, implements the steps of the forward-looking heterogeneous shared autonomous driving carpooling scheduling method as described in Embodiment 1 of the present invention.

[0165] Example 4:

[0166] Embodiment 4 of the present invention provides a device, including a memory, a processor, and a program stored in the memory and executable on the processor. When the processor executes the program, the steps of the forward-looking heterogeneous shared autonomous driving carpooling scheduling method as described in Embodiment 1 of the present invention are implemented.

[0167] The steps involved in the above embodiments 2, 3 and 4 correspond to those in the method embodiment 1. For the specific implementation methods, please refer to the relevant description part of the embodiment 1.

[0168] Those skilled in the art will appreciate that the modules or steps of the present invention described above can be implemented using a general-purpose computer device. Alternatively, they can be implemented using program code executable by a computing device, which can then be stored in a storage device and executed by the computing device. Alternatively, they can be fabricated into separate integrated circuit modules, or multiple modules or steps can be fabricated into a single integrated circuit module for implementation. The present invention is not limited to any specific combination of hardware and software.

[0169] Although the above describes the specific embodiments of the present invention in conjunction with the accompanying drawings, it is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art on the basis of the technical solution of the present invention without any creative work are still within the scope of protection of the present invention.

Claims

1. A forward-looking method for scheduling heterogeneous shared autonomous vehicles, characterized in that: The following steps are involved: Obtain ride requests, shared autonomous vehicle status, and road network information within the dispatch area; Based on the ride request and shared autonomous vehicle status combined with road network information, candidate matching pairs of shared autonomous vehicle status and request are generated. The specific steps are as follows: Setting request acceptance conditions based on the relationship between ride requests, shared autonomous vehicle status, and road network information; Determine the order of boarding and disembarking based on the shared autonomous vehicle status, based on whether the ride request meets the request acceptance conditions; Determine the shortest path for the shared autonomous vehicle based on the pickup and drop-off sequence, and identify candidate matching pairs of the shared autonomous vehicle state and the request; The state variables of the regional value function are determined based on the historical order information of the dispatching area. The regional value function is constructed based on the state variables using the Markov decision method and evaluated. The specific steps are as follows: Divide the dispatching area into several small areas; The Markov decision method is used to model and estimate the regional value function of the state variables of a small area. The temporal difference learning method is introduced to estimate the regional value function, and the parameters are updated based on the Bellman equation. The regional value function represents the long-term return that a shared autonomous vehicle can obtain after entering the current area within a preset time period. Build a dynamic ride-sharing matching algorithm based on the interaction between ride requests and shared autonomous vehicles within the dispatch area; Based on the evaluation results of the regional value function, a dynamic carpooling matching algorithm is used to select the optimal match between shared autonomous vehicles and ride requests according to candidate matching pairs, and carpooling scheduling is performed based on the optimal matching results.

2. The forward-looking heterogeneous shared autonomous driving carpooling scheduling method according to claim 1, characterized in that: The state variables of the regional value function include the number of orders departing from the current region, the number of orders destined for the current region, the number of shared autonomous vehicles in the current region, and the average order waiting time in the current region.

3. The forward-looking heterogeneous shared autonomous driving carpooling scheduling method according to claim 1, characterized in that: The specific steps for building a dynamic ride-sharing matching algorithm based on the interaction between ride requests and shared autonomous vehicles within the dispatch area are: Transform the real-time ride-sharing scheduling problem using a bipartite graph matching algorithm; Based on the matching and compatibility between shared autonomous vehicles and ride requests within the dispatch area, the advantage function is determined by considering multiple factors and constraints are set.

4. The forward-looking heterogeneous shared autonomous driving carpooling scheduling method according to claim 3, characterized in that: Consider a variety of factors including: Satisfy the most ride requests; Reduce passenger waiting time and detour distance; Improve the passenger load and service success rate of shared autonomous vehicles; Incorporate long-term benefit considerations into scheduling decisions.

5. A forward-looking heterogeneous shared autonomous vehicle carpooling scheduling system, using the forward-looking heterogeneous shared autonomous vehicle carpooling scheduling method according to any one of claims 1 to 4, characterized in that: include: a data acquisition module configured to acquire ride requests, shared autonomous vehicle status, and road network information within a dispatch area; a candidate matching module configured to generate a candidate matching pair of a shared autonomous vehicle state and a request based on the ride request and the shared autonomous vehicle state in combination with road network information; a regional value evaluation module configured to determine state variables of a regional value function based on historical order information of a dispatching area, construct a regional value function based on the state variables using a Markov decision method, and evaluate the regional value function; a matching algorithm module configured to construct a dynamic ride-sharing matching algorithm based on the interaction between ride requests and shared autonomous vehicles within a dispatch area; The carpooling scheduling module is configured to select the optimal match between the shared autonomous driving vehicle and the ride request based on the evaluation results of the regional value function using a dynamic carpooling matching algorithm according to candidate matching pairs, and perform carpooling scheduling based on the optimal matching results.

6. A computer-readable storage medium, characterized in that Multiple instructions are stored therein, which are suitable for being loaded by a processor of a terminal device and executed by the forward-looking heterogeneous shared autonomous driving carpooling scheduling method described in any one of claims 1-4.

7. A terminal device, characterized in that: The invention comprises a processor and a computer-readable storage medium, wherein the processor is used to implement each instruction; the computer-readable storage medium is used to store multiple instructions, wherein the instructions are suitable for being loaded by the processor and executed by the forward-looking heterogeneous shared autonomous driving carpooling scheduling method described in any one of claims 1 to 4.

Images