Global multi-vehicle decision system for connected and automated vehicles in dynamic environments

Abstract

Connected and automated vehicles (CAVs) show the potential to improve safety, increase road throughput, and optimize energy efficiency and emissions in several complex traffic scenarios. This invention describes a mixed integer programming (MIP) optimization method for global multi-vehicle decision and motion planning of CAVs in highly dynamic environments consisting of multiple human-driven vehicles (i.e., conventional or manual vehicles) and multiple conflict zones (e.g., merging points and intersections). The proposed method ensures safety, high throughput, and energy efficiency by solving a global multi-vehicle constrained optimization problem. The scheme provides a feasible optimal time schedule for automated vehicles to traverse road segments and conflict zones by using information from the positions, speeds, and destinations of manual vehicles that cannot be directly controlled. Despite the combinatorial complexity of MIP, the proposed scheme is still feasible for real-time implementation in infrastructure such as mobile edge computers (MECs).

Description

Technical Field

[0001] This disclosure generally relates to optimization-based control, and more specifically to methods and apparatus for optimization-based global multi-vehicle decision-making and motion planning for connected automated vehicles in dynamic environments within transportation networks. Background Technology

[0002] Automated transportation systems, even partially automated ones, can reduce road accidents and improve the efficiency of road networks. Therefore, connected autonomous vehicles (CAVs) have significant potential to improve safety and traffic flow, thereby reducing congestion, travel time, emissions, and energy consumption. While this issue has been recognized for decades, most successful developments have occurred in recent years due to technological advancements in sensing, computing, control, and connectivity. Although road scenarios are often highly dynamic (i.e., vehicle participants and their behaviors change rapidly and significantly), vehicle-to-vehicle (V2V) and vehicle-to-infrastructure (V2I) communications (also known as vehicle-to-everything (V2X) communications) enable advanced and efficient planning and decision-making by providing access to real-time information on all vehicles within a planned area.

[0003] Significant progress has been made in the planning and control of autonomous driving, which typically involves multi-layered guidance and control architectures implemented on-vehicle. At the highest level, intelligent navigation systems use traffic networks to find a route from the vehicle's current location to the desired destination. Decision-makers select appropriate driving actions at any given time based on route planning, current environmental conditions, and the behavior of other traffic participants, such as using automata combined with ensemble reachability or formal languages ​​and optimization. Given a target action (including lane following, lane changing, or stopping), motion planning algorithms compute a dynamically feasible safe trajectory, which can be tracked in real time by a low-level feedback controller. A popular approach is to combine sample-based motion planning algorithms with model predictive control (MPC) for reference tracking. For CAVs, the guidance and control architecture may resemble standard autonomous driving, but some modules may be implemented in the infrastructure (e.g., in a mobile edge computer (MEC)) and provide decision-making for multiple vehicles in the area, while other modules may still be implemented independently on each vehicle.

[0004] Coordination by cooperative agents enables traffic networks to achieve socially optimal behavior. One embodiment describes a First-Arriv-First-Served (FCFS) strategy for autonomous traffic management at intersections. Recently, coordination strategies for intersection control have been proposed using nonlinear optimization or mixed-integer linear programming (MILP). The latter has been extended to a distributed MILP algorithm for scheduling a grid of interconnected intersections. Furthermore, a MILP-based CAV ramp merging method has been proposed. Alternative techniques for CAV coordination can be found in recent work, where it can be noted that the intersection and merging control problems are fundamentally very similar. It is necessary to develop an advanced global multi-vehicle decision-making system. Summary of the Invention

[0005] This invention focuses on global multi-vehicle decision-making and motion planning for CAVs, providing target and operational information to local motion planning and tracking systems implemented independently in each autonomous vehicle in the presence of conventional (i.e., manually driven) road vehicles (referred to as uncontrolled vehicles (NCVs)). Embodiments of the invention include a MILP-based global multi-vehicle decision-making system for vehicles in interconnected networks of general conflict zones, where these zones include intersections and merging points. Unlike prior art for fully autonomous vehicle coordination, this invention focuses on more realistic scenarios of mixed traffic involving both CAVs and human-driven NCVs. For this purpose, in mixed traffic scenarios, human-driven vehicles require physical traffic lights and / or standard priority rules to cross intersections. Given a real-time sequence of vehicle routing information from an advanced algorithm, the proposed constrained optimization method directly incorporates the transportation of people and goods.

[0006] The embodiments of this invention are based on a mixed-integer optimization method for global multi-vehicle decision-making and motion planning of connected automated vehicles (CAVs) in interconnected networks with general conflict zones including both intersections and merging points, and in the presence of conventional (i.e., manually operated) road vehicles (referred to as uncontrolled vehicles (NCVs)). The proposed method, under the safety constraints of the conflict zone and the occupancy constraints of the road segments, calculates the scheduling consisting of the target speed and time for each vehicle to enter and leave the road segment towards its desired destination within a prediction window, while simultaneously optimizing the total time and energy efficiency of all controlled vehicles. In contrast to existing methods in the prior art, the proposed system and method support:

[0007] • The traffic network in the conflict zone, i.e., including one or more merging points and / or intersections;

[0008] • Mixed traffic including autonomous vehicles and human-driven vehicles, where the position, speed and heading of each vehicle can be obtained through V2X communication, but only some vehicles are controlled in local areas of multiple interconnected conflict zones;

[0009] • Global multi-vehicle decision-making and motion planning optimized based on multiple objectives (e.g., driving time, waiting time, and energy efficiency);

[0010] • CAV's rough advanced motion plan, which uses future routing information to transport people and goods.

[0011] Some embodiments of the present invention are based on a multi-layered guidance and control architecture, with some modules implemented on-board in each CAV and others operating centrally in an infrastructure (e.g., in a mobile edge computer) to leverage V2X connectivity. Specifically, in each CAV, given a target behavior representing actions such as lane following, lane changing, or parking, a motion planning algorithm calculates a dynamically feasible and safe trajectory that can be tracked in real time by the vehicle controller. Some embodiments are based on a probability-sampling-based motion planner and an MPC algorithm for reference tracking in each CAV.

[0012] Unlike fully autonomous vehicles, the implementation of this invention is based on the understanding that, in the case of CAVs, the target behavior of the motion planning algorithm should not be calculated individually for each vehicle using an onboard decision-making algorithm, but rather simultaneously for all vehicles within a global multi-vehicle decision-making module. Specifically, given real-time information about the environment from the mapping and navigation module (e.g., the real-time status and routing information of each CAV in a local area of ​​multiple interconnected conflict zones), the global multi-vehicle decision-making module simultaneously determines the target behavior of all CAVs in that local area. Note that the routing information for each CAV can be calculated independently by a centralized mapping and navigation module or by an onboard mapping and navigation module within each local CAV.

[0013] Embodiments of the present invention coordinate and schedule vehicles within a road segment network, which consists of one or more road segments serving as conflict zones and one or more road segments serving as conflict-free zones. Examples of conflict zones are intersections and / or merging points connecting multiple lanes and / or road segments. Examples of conflict-free zones are standard road segments consisting of one or more lanes allowing traffic in one or more directions.

[0014] The embodiments of the present invention are based on the understanding that information from a global multi-vehicle decision-making module can be obtained through communication between vehicles and infrastructure (i.e., V2X). This information is acquired by sensors in both CAVs and NCVs, and may also be acquired by additional sensors in the infrastructure. Specifically, the input information from V2X communication to the global multi-vehicle decision-making system may include:

[0015] • The global decision module considers map information of the traffic network, which includes lane information for each road segment and conflict zone (i.e., merging points and / or intersections);

[0016] • For each vehicle (i.e., both CAV and NCV in a local area of ​​the traffic network), the vehicle's current state, including current position, heading, and speed;

[0017] • For each NCV, a possible short-term route prediction is provided, defined as a sequence of future road segments starting from the NCV's current location and heading towards its desired destination. The length of the route (i.e., the number of road segments in the route prediction) can be fixed or time-varying, depending on the available information.

[0018] • For each CAV, a relatively long-term route plan is defined as a sequence of future road segments starting from the CAV's current location and heading towards its desired destination. The length of the planned route (i.e., the number of road segments in the future route plan) can be fixed or time-varying, depending on the information available from each CAV's mapping and navigation modules;

[0019] • For each CAV, and for each segment of its future planned route, a set of planned stop durations along each segment of the route (e.g., for transporting people and / or goods).

[0020] Similarly, the output information from a global multi-vehicle decision-making system may include:

[0021] • For each CAV, the predicted average speed sequence along the road segments of its future planned route in the traffic network;

[0022] • For each CAV, the predicted segments of the road network along its future planned route are entered into the time series;

[0023] • For each CAV, the predicted departure time series of the segment of the traffic network along its future planned route.

[0024] The length of the prediction window (i.e., the length of the route) can be chosen to be constant for all vehicles, can be chosen independently for each vehicle, or can be time-varying and different for each vehicle, depending on the information available to the global multi-vehicle decision system.

[0025] The embodiments of this invention are based on the understanding that obtaining accurate route predictions for each NCV can be challenging, depending on the infrastructure system. Therefore, in some embodiments of this invention, a global multi-vehicle decision module is implemented in a rolling time-domain manner based on the most recent information. The embodiments of this invention are based on the understanding that approximate short-term route predictions for NCVs are sufficient (e.g., up to the next conflict zone) and that any discrepancies in the predictions can be adjusted using the inherent feedback mechanism of the rolling time-domain strategy. For example, an update period of 1-2 seconds allows for real-time computation of the global multi-vehicle decision system while providing sufficiently fast updates to account for erroneous predictions of NCV behavior.

[0026] Some embodiments of the present invention are based on the understanding that the presence of other traffic participants (e.g., including bicycles and pedestrians) can be treated as obstacles by onboard modules (e.g., motion planning and / or vehicle control algorithms) in the multi-layered guidance and control architecture of each CAV. Due to their relatively low computational cost, motion planning and vehicle control algorithms can operate at a relatively fast sampling rate compared to global multi-vehicle decision systems, allowing for faster reaction times to unexpected changes in the behavior of other traffic participants (e.g., including bicycles and pedestrians). For example, vehicle control algorithms are typically executed with update intervals of 50 to 100 milliseconds.

[0027] The proposed global multi-vehicle decision module targets small to medium-sized traffic networks (e.g., local areas with multiple interconnected conflict zones and potential for congestion). Some embodiments of the invention include transport tasks for CAVs (Carrier Aerial Vehicles) carrying people and / or goods (e.g., groceries or parcels), utilizing a potentially varying number of CAVs operating in the same environment as many potentially uncontrolled traffic participants. Task assignment (i.e., the objective that each CAV must accomplish) can be performed by an independent task assignment module, for example, based on a solution to a constrained optimization problem. Based on the assigned tasks, a navigation module determines the routes of the CAVs (centrally or locally within each CAV) and can update route information in real time. Considering computational traceability (including rolling time-domain computation and effectiveness in practical applications), embodiments of the invention can handle at least 1 to 10 CAVs and 0 to 30 NCVs in local traffic networks with up to 10 conflict zones and several connecting road segments.

[0028] The embodiments of this invention are based on the understanding that, given input information from V2X communication, a global multi-vehicle decision module can be implemented by solving a constrained optimization problem to compute a coarse motion plan for each CAV. In some embodiments of this invention, the constrained optimization problem can be a mixed-integer programming (MIP) problem, such as a mixed-integer linear programming (MILP) or mixed-integer quadratic programming (MIQP) problem. In some embodiments of this invention, the MIP problem at each sampling time of the global multi-vehicle decision system can be solved using a global optimization algorithm (e.g., including branch and bound, branch cutting, and branch pricing methods). In other embodiments of this invention, heuristic techniques can be used to compute feasible but suboptimal solutions to the MIP, such as rounding schemes, feasibility pumps, approximate optimization algorithms, or the use of machine learning.

[0029] Furthermore, according to some embodiments of the present invention, a global multi-vehicle decision system can be implemented to provide real-time motion planning and coordination for one or more connected automated and / or semi-automated vehicles (CAVs) in an interconnected transportation network, the interconnected transportation network including one or more uncontrolled vehicles (NCVs), one or more conflict zones, and one or more conflict-free road segments. In this case, the global multi-vehicle decision system may include: a receiver configured to acquire infrastructure sensing signals via a roadside unit (RSU) and acquire Type 1 feedback signals regarding the status of the connected automated vehicles (CAVs) and planned future routes to one or more desired destinations of the CAVs, and Type 2 feedback signals regarding the status of the uncontrolled vehicles (NCVs) and predicted future routes; at least one memory configured to store map information and a computer-executable program including global multi-vehicle decision planning; and at least one processor connected to the at least one memory and configured to perform the following steps: based on the infrastructure... The system uses sensed signals, type 1 feedback signals, and type 2 feedback signals, along with map information, to formulate a global mixed-integer programming (MIP) problem; solves the global MIP problem to compute motion plans for each CAV and each NCV in the interconnected transportation network; computes optimal entry and exit time series and average speed series for each CAV and each NCV in each segment of a planned or predicted future route within the transportation network; computes the speed distribution and / or one or more planned stops for each CAV in the prediction time domain; and a transmitter configured to send the speed distribution to each CAV and / or send the one or more planned stops to a multi-layered guidance and control architecture for each CAV in the interconnected transportation network.

[0030] In some implementations of global multi-vehicle decision systems, the optimization problem involves optimizing one or more objectives and imposing one or more equality and / or inequality constraints on the safety and efficiency of all vehicles in the traffic network. For example, constraints may include vehicle motion models, speed constraints, time constraints for route planning or route prediction, safety constraints in conflict zones, and occupancy constraints for conflict-free road segments. Objectives may include minimizing the travel time to the destination of each road segment in the route plan or route prediction, minimizing the waiting time for each vehicle, maximizing the average speed, and minimizing acceleration for energy efficiency.

[0031] The currently disclosed embodiments will be further explained with reference to the accompanying drawings. The drawings are not necessarily drawn to scale, but rather focus on illustrating the principles of the currently disclosed embodiments. Attached Figure Description

[0032] [ Figure 1 ]

[0033] Figure 1 An embodiment of a traffic scenario in a local area of ​​multiple interconnected conflict zones according to some implementation methods is shown, along with the need for global multi-vehicle decision-making for controlled and uncontrolled vehicles.

[0034] [ Figure 2A ]

[0035] Figure 2A An exemplary schematic diagram is shown illustrating traffic control in a localized area of ​​multiple interconnected conflict zones based on vehicle segment crossing time and average speed according to some implementation methods.

[0036] [ Figure 2B ]

[0037] Figure 2B This illustrates a traffic scenario where, according to some implementations, routing information for multiple connected autonomous vehicles (e.g., for transporting people and goods) is displayed in a localized area of ​​an interconnected conflict zone.

[0038] [ Figure 3 ]

[0039] Figure 3 The diagram illustrates possible interactions between a global multi-vehicle decision-making system according to some embodiments of the present invention and a mapping and navigation system, one or more controlled and uncontrolled vehicles, roadside units, passengers, and mobile edge computers.

[0040] [ Figure 4A ]

[0041] Figure 4AA schematic diagram of the feedback loop of a global multi-vehicle decision system based on infrastructure sensing, centralized mapping and navigation, and a multi-layered guidance and control architecture for each of the various controlled vehicles is shown according to some embodiments of the present invention.

[0042] [ Figure 4B ]

[0043] Figure 4B A similar schematic diagram of a feedback loop for a global multi-vehicle decision system using an independent navigation system for one or more controlled vehicles is shown according to some embodiments of the present invention.

[0044] [ Figure 5A ]

[0045] Figure 5A A block diagram of a global multi-vehicle decision system based on communication with infrastructure and / or connected vehicles is shown, according to an embodiment of the present invention, by solving a mixed-integer optimization problem.

[0046] [ Figure 5B ]

[0047] Figure 5B A schematic diagram of the inputs and outputs of a global multi-vehicle decision system according to an embodiment of the present invention is shown.

[0048] [ Figure 6A ]

[0049] Figure 6A An example of a traffic network (a local area therein) consisting of multiple conflict-free road segments interconnected via three-way intersections, each segment consisting of one or more track lanes.

[0050] [ Figure 6B ]

[0051] Figure 6B As shown Figure 6A Examples of possible routes for each of two vehicles within an interconnected transportation network are shown.

[0052] [ Figure 6C ]

[0053] Figure 6C This illustrates an embodiment of the invention that can be implemented by a global multi-vehicle decision system. Figure 6B An example of a motion plan calculated for vehicles within a traffic network is shown.

[0054] [ Figure 7 ]

[0055] Figure 7A block diagram of a global multi-vehicle decision system based on an embodiment of the present invention, which solves a mixed integer programming (MIP) optimization problem, is shown.

[0056] [ Figure 8A ]

[0057] Figure 8A An embodiment of equality and / or inequality constraints that can be enforced by a global multi-vehicle decision system for a MIP problem according to an embodiment of the present invention is shown.

[0058] [ Figure 8B ]

[0059] Figure 8B An exemplary formula for the safety constraints for each conflict zone according to an embodiment of the present invention is shown, which can be used in the MIP problem solved by a global multi-vehicle decision system.

[0060] [ Figure 8C ]

[0061] Figure 8C An exemplary formula for occupancy constraints for each road segment according to an embodiment of the present invention is shown, which can be used in the MIP problem solved by a global multi-vehicle decision system.

[0062] [ Figure 8D ]

[0063] Figure 8D An embodiment of the objective used in weighted multi-objective minimization of the MIP problem in a global multi-vehicle decision system is shown, according to an embodiment of the present invention.

[0064] [ Figure 9A ]

[0065] Figure 9A A schematic diagram of an embodiment of an integer optimization variable search tree according to some embodiments of the present invention is shown, which represents a nested search region tree for integer feasible optimal solutions for global multi-vehicle decision-making.

[0066] [ Figure 9B ]

[0067] Figure 9B A block diagram of a branch-and-bound mixed integer optimization algorithm according to some embodiments of the present invention is shown, which searches for integer feasible optimal decision solutions based on nested search region trees and corresponding lower / upper bound values.

[0068] [ Figure 10 ]

[0069] Figure 10A block diagram of a global multi-vehicle decision system for calculating the motion plans of one or more controlled vehicles in a traffic network of one or more interconnected conflict zones, according to some embodiments of the present invention, is shown.

[0070] [ Figure 11A ]

[0071] Figure 11A A schematic diagram of a vehicle is shown, incorporating principles of a multi-layered guidance and control architecture implemented in several ways.

[0072] [ Figure 11B ]

[0073] Figure 11B A schematic diagram illustrating the interaction between a multi-layered guidance and control architecture of a vehicle and other controllers according to some embodiments of the present invention is shown.

[0074] [ Figure 12A ]

[0075] Figure 12A A schematic diagram illustrating the interaction between a freight yard management system and a global multi-vehicle decision system according to some embodiments of the present invention is shown.

[0076] [ Figure 12B ]

[0077] Figure 12B An embodiment of a freight yard management system consisting of one or more components according to some embodiments of the present invention is shown. Detailed Implementation

[0078] Some embodiments of this disclosure provide a system and method for controlling one or more connected automated vehicles within a traffic network that comprises one or more interconnected conflict zones and includes a dynamic environment, such as one or more uncontrolled vehicles, traffic participants, or dynamic obstacles.

[0079] Figure 1 An embodiment of a traffic scenario (traffic state) 100 according to some implementations is illustrated, illustrating the need for global multi-vehicle decision-making for controlled and uncontrolled vehicles within a local area of ​​multiple interconnected conflict zones (e.g., at interconnected intersections of road segments). Interconnected conflict zones include physically interconnected conflict zones and communicatively interconnected conflict zones. An embodiment of a physically interconnected conflict zone is an intersection or merging point from one conflict zone to another from which vehicles can travel. An embodiment of a communicatively interconnected conflict zone is an intersection or merging point from one conflict zone to another via a communication channel for sharing traffic information.

[0080] An embodiment of traffic scenario 100 illustrates intersections 102 and 104 physically interconnected via road segments 105, and the traffic network includes additional road segments 106, 108, 110, 112, 114, and 116. Embodiments of the invention describe a system and method for making global multi-vehicle decisions for vehicles within such a road segment network, which consists of one or more road segments as conflict zones and one or more road segments as conflict-free zones. An embodiment of a conflict zone is intersections 102-104 and / or merging points, which may connect multiple lanes and / or road segments. An embodiment of a conflict-free zone is standard road segments 105-116, which consist of one or more lanes allowing one-way or multi-way traffic.

[0081] Figure 1 The diagram also illustrates multiple vehicles 126, 128, 130, 132, 134, 136, 138, 140, 142, 144, 146, and / or other traffic participants within local areas of multiple interconnected conflict zones 102 to 104 and multiple interconnected conflict-free road segments 105 to 116. Vehicles in the traffic network (e.g., vehicles 126-146) can each be autonomous, semi-autonomous, or manually operated. Autonomous and / or semi-autonomous vehicles are further referred to as connected automated vehicles (CAVs), or simply controlled vehicles, and manually operated vehicles or other traffic participants are referred to as uncontrolled vehicles (NCVs). Some embodiments of vehicles 126-146 include two-wheeled vehicles (such as e-bikes), four-wheeled vehicles (such as cars), or vehicles with more than four wheels (such as trucks, etc.).

[0082] A traffic network that includes one or more interconnected conflict zones and one or more interconnected conflict-free road segments may additionally include one or more roadside units (RSUs) for real-time sensing of the status of vehicles and other traffic participants in the local area surrounding each RSU based on infrastructure. Figure 1 An RSU 122 and another RSU 124, a core network 120 and a cloud network 118 are shown according to some embodiments of the present invention to establish a vehicle-to-everything (IoV) environment, which includes vehicle-to-vehicle (V2V) and vehicle-to-infrastructure (V2I) communications (also known as vehicle-to-everything (V2X) communications).

[0083] In some implementations, traffic scenario 100 corresponds to a public urban area where road segments 105-116 form a number (large number) of intersections, such as intersections 102 and 104. Within the urban area, traffic conditions at intersections 102 and 104 determine traffic flow, as traffic congestion typically begins in conflict zones (such as intersection 104) and propagates further to non-conflict road segments (such as road segments 105-116). Traffic conditions at interconnected conflict zones 102 and 104 are interdependent, such that changes at one conflict zone (e.g., intersection 104) propagate further to other interconnected conflict zones, such as intersection 102 (also referred to as the adjacent intersection 102 of intersection 104).

[0084] In other embodiments, traffic scenario 100 may correspond to a private transportation network, for example, including one or more parking areas and interconnected road segments for a valet parking system. Other embodiments of similar traffic scenarios (including transportation networks with multiple interconnected conflict zones and road segments) are smart distribution centers and / or freight yards. Embodiments of the CAV type include personal vehicles (e.g., in the case of a valet parking system), trucks (e.g., in the case of a parking lot management system), or shuttle buses for picking up and dropping off passengers.

[0085] Some implementations are based on the understanding that, in order to establish communication between different vehicles (e.g., a group of vehicles 126-146) in a traffic network, communication between the cloud network 118 on road segment 105 and vehicle 126 needs to propagate through RSU 122 or RSU 124 and core network 120, thereby establishing multi-hop communication. In some embodiments of the invention, the safe mobility of vehicles 126-146 is controlled by a global multi-vehicle decision system using cloud-based and / or edge-based networks (i.e., using cloud network 118 and / or core network 120). In some embodiments of the invention, the global multi-vehicle decision system is implemented using one or more mobile edge computers (MECs), which may be embedded as part of one or more RSUs, or they may be independent devices connected to RSUs 122-124, cloud network 118, and / or core network 120. Embodiments of the invention include solving a constrained optimization problem for global multi-vehicle decision-making of vehicles in a traffic network, for which computation can be performed in the cloud or in one or more MECs.

[0086] The embodiments of the present invention are based on the understanding that the on-board control device of a vehicle (such as vehicle 146) cannot obtain information about adjacent vehicles (such as vehicle 142), pedestrians, and environmental conditions outside its visible range. For example, vehicle 146 traveling on road 114 intends to cross intersection 102 after vehicle 144 (which is larger than vehicle 142) crosses intersection 102, and vehicle 142 (which is a smaller vehicle) also moves into intersection 102. In this scenario, the visibility of vehicle 142 is obstructed by vehicle 144, such as... Figure 1 As shown in the diagram. If the communication link between vehicle 142 and vehicle 146 is affected, or if vehicle 142 and vehicle 146 use different communication protocols, vehicle 142 will be noticed by vehicle 146. Therefore, vehicle 142 and vehicle 146 may collide.

[0087] In addition, some embodiments of the present invention are based on the understanding that multi-hop communication between cloud network 118 and vehicle 126 may result in long communication delays, which may be unacceptable in real-time scenarios of cloud-based vehicle control.

[0088] Some implementations are based on the understanding that different communication technologies can be used to support vehicle communication. For example, the IEEE Dedicated Short Range Communication / Wireless Access in Vehicle Environments (DSRC / WAVE) standard series for vehicle networks, 3GPP Cellular Vehicle-to-Everything (C-V2X), etc. However, due to high cost, it is impractical for vehicles (e.g., vehicles 126-146) to support more than one short range communication technology, which leads to compatibility issues in communication between vehicles. Therefore, a vehicle equipped with IEEE DSRC / WAVE cannot communicate with other vehicles equipped with 3GPP C-V2X, and vice versa. Consequently, the accuracy of real-time control decisions in the onboard multi-layered guidance and control architecture of each individual vehicle will be severely affected, as these real-time decisions will be based on incomplete information about the traffic scenario 100. Instead, embodiments of the present invention use a global multi-vehicle decision module based on real-time information from each vehicle in the traffic network to ensure safety, time, and energy efficiency.

[0089] Some implementations are based on the understanding that edge infrastructure devices (e.g., RSU 122 and RSU 124) are more advantageous in controlling multi-vehicle traffic than using only cloud networks and / or only on-board units (including multi-layered guidance and control architectures). For example, edge infrastructure devices can be installed at intersections or merging points, and they can communicate directly with vehicles approaching the intersection or merging point (e.g., vehicles 126-146). Furthermore, edge infrastructure devices can be equipped with various communication technologies to enable communication with all connected vehicles. Some implementations of the invention are based on the understanding that edge infrastructure devices are stationary, which allows them to provide reliable communication with vehicles and collect relatively high-quality environmental data.

[0090] The embodiments of the present invention are based on the understanding that edge infrastructure devices can continuously monitor multi-vehicle traffic and the environment to make accurate decisions. In some embodiments of the present invention, the edge infrastructure device uses additional sensors (e.g., rangefinders, radar, lidar, and / or cameras) and sensor fusion techniques to accurately detect vehicle status and dynamic environments, including connected and non-connected vehicles, autonomous, semi-autonomous, and manually operated vehicles, as well as other traffic participants such as bicycles and pedestrians. Therefore, edge infrastructure devices are suitable for global multi-vehicle decision-making and motion planning.

[0091] Figure 2A An embodiment of a traffic network with multiple interconnected conflict zones is illustrated, namely, including multiple interconnected intersections 200, 202a, 202b, 204a, and 204b. According to embodiments of the invention, the overall safety, time efficiency, and energy efficiency of traffic flow in this traffic network can be controlled by a global multi-vehicle decision system. In some embodiments of the invention, the global multi-vehicle decision system calculates a coarse motion plan for each CAV in the traffic network along its route from its current location to its desired destination. The coarse motion plan may include the entry and exit time series and average speed value series of each CAV in each of the conflict zones and conflict-free segments along the CAV's route from its current location to its desired destination (which may include one or more planned intermediate stops along the CAV's route).

[0092] In some embodiments of the invention, the input to the global multi-vehicle decision system may include signals from communication with one or more CAVs and / or one or more NCVs, and may include signals from communication with one or more edge infrastructure devices (e.g., RSUs with additional sensors and sensor fusion capabilities). For example, in Figure 2AIn the intersection, each of intersections 200, 202a, 202b, 204a, and 204b is equipped with RSUs 206, 208, 210, 212, and 214, respectively. Some embodiments of the invention include solutions to constrained optimization problems of global multi-vehicle decision-making for vehicles in a traffic network, the computation of which can be performed in the cloud or in one or more MECs, which can be embedded as part of one or more RSUs 206, 208, 210, 212, and 214, or can be standalone devices connected to one or more RSUs 206-214.

[0093] exist Figure 2A In the illustrative embodiment scenario, north-south traffic at intersections 202a and 202b is more congested than east-west traffic at intersections 204a and 204b. A global multi-vehicle decision system controls the planned future timing and speed trajectories of vehicles (e.g., vehicle 216) that plan to cross intersection 200 and travel north-south towards the highly congested intersection 202a. To improve the overall safety, time efficiency, and energy efficiency of traffic flow in the traffic network, the global multi-vehicle decision system can determine that vehicle 216 should slow down before and / or after crossing intersection 200, thereby predicting that vehicle 216 will arrive at intersection 202a at a later time, thus reducing traffic congestion at intersection 202a and reducing the overall waiting time for one or more vehicles in the traffic network.

[0094] Figure 2B An exemplary traffic scenario is shown within a traffic network of multiple interconnected intersections and merging points, relating to a global multi-vehicle decision system based on some embodiments of the present invention. Figure 2B A scenario is depicted where one or more controlled vehicles (referred to as CAVs, such as 231, 232, 233, and 234) and one or more uncontrolled traffic participants (such as NCVs 235, 236, 237, and 238) are included. The traffic network itself may consist of multiple interconnected intersections (such as 251(I1), 252(I2), and 253(I3)) and multiple merging points (such as 255(M1), 256(M2), and 257(M3)). In the description of embodiments of the invention, both intersections and merging points are referred to as conflict zones. Conflict zones are interconnected via multiple conflict-free road segments, each of which may consist of one or more lanes (e.g., 265(L6), 266(L47), and 267(L36)). Furthermore, Figure 2B Stop lines (e.g., stop lines 261(S1), 262(S2), and 263(S3)) are depicted, which indicate the positions where vehicles may wait for a specific period of time before entering the intersection.

[0095] Figure 2B Additionally, routing information that can be provided by a routing or navigation module for each CAV (e.g., vehicle 231) is shown. For vehicle 231, which is currently at position 240 and has a desired destination 249, the routing or navigation module provides vehicle 231 with a sequence of roads and turns indicated by arrows 241-248. Similarly, the same centralized or different independent routing or navigation modules can provide other vehicles (e.g., CAVs 232, 233, and 234) with a sequence of roads and turns from their current position to their desired destination. However, it should be noted that the sequence of roads and turns 241-248 itself does not specify a movement plan or path for vehicle 231. There are many discrete decisions to be made, such as which lane the vehicle should travel in, whether the vehicle should change lanes or remain in its current lane, whether the vehicle should begin to decelerate to stop at the stop line, and whether the vehicle is permitted to cross an intersection. Furthermore, there are many continuous decisions to be made, such as a timing sequence of the positions and orientations that the vehicle should reach as it travels from its initial point to its destination. These decisions are highly dependent on the current traffic conditions at the moment the vehicle arrives at the corresponding location. Due to the uncertainty of traffic movement and the uncertainty of when the vehicle will arrive at the location, the routing module is usually unaware of the current traffic conditions.

[0096] In some embodiments of this disclosure, a global multi-vehicle decision system may compute motion plans for one or more CAVs (e.g., 231-234), the motion plans comprising a sequence of one or more of the aforementioned discrete and / or continuous decisions along a route from the current location to the desired destination. In some embodiments of the invention, the global multi-vehicle decision system relies on real-time communication to allow vehicle-to-vehicle (V2V) coordination and / or intelligent infrastructure system-to-vehicle (V2X) communication. In some embodiments of the invention, the global multi-vehicle decision system computes a motion plan for each CAV by solving one or more constrained optimization problems (e.g., constrained mixed-integer programming problems).

[0097] According to some implementations, the global multi-vehicle decision module is designed for small to medium-sized traffic networks, such as local (public or private) areas with multiple interconnected conflict zones. Some implementations of the invention involve using a variable number of CAVs operating in the same environment as potentially many uncontrolled traffic participants to perform transportation tasks for people and / or goods, such as picking up and dropping off passengers, loading and unloading groceries or other parcels. Task assignment (i.e., the objective that each CAV must complete) can be performed by an independent task assignment module, for example, based on solutions to potentially large-constrained (mixed-integer) optimization problems. Based on the assigned tasks, the navigation module centrally or locally determines the route for each CAV and can update route information in real time. Considering the tractability of computation (including rolling time-domain computation) and its effectiveness in practical applications, implementations of the invention can handle at least 1-10 CAVs and 0-30 NCVs in local traffic networks with up to 10 conflict zones and several connected road segments.

[0098] Figure 3 A schematic diagram illustrates possible interactions between the global multi-vehicle decision system 300 and the mapping and navigation system 310, one or more controlled (CAV) vehicles 330 and uncontrolled (NCV) vehicles 340, one or more roadside units (RSU) 350, one or more passenger management systems (HMI) 360, and / or one or more mobile edge computers (MEC) 320. According to some embodiments of the invention, the mapping and navigation system calculates centralized or decentralized (e.g., on a device embedded as part of the CAV) real-time routing information for each CAV from its current location to its desired destination or sequence of desired destinations. This routing information can be transmitted from the mapping and navigation module to the global multi-vehicle decision system 311 and / or to each individual CAV 331. Similarly, the mapping and navigation system can receive up-to-date information from the global multi-vehicle decision system 311 and / or from each individual CAV 331.

[0099] According to some embodiments of the invention, the global multi-vehicle decision system calculates a coarse motion plan for each CAV in the traffic network along its desired destination or sequence of desired destinations from its current location to its desired destination. This motion plan is then transmitted directly from the global multi-vehicle decision system 332 to each controlled vehicle (CAV), or alternatively indirectly via communication with one or more MECs 301, which provides up-to-date information from the global multi-vehicle decision system to the CAV 330. Similarly, according to embodiments of the invention, the global multi-vehicle decision system relies on real-time information that can be transmitted directly from CAV 330, NCV 340, RSU 350, and HMI 360, or indirectly from MEC 320 via 301. In some embodiments of the invention, the MEC can collect real-time information about the status of traffic participants and the dynamic environment in the local area of ​​the traffic network by communicating with CAV 330, NCV 341, RSU 351, and human passengers 361 currently present in the local area of ​​the traffic network.

[0100] Some embodiments of the present invention are based on the understanding that the number of MEC 320, CAV 330, NCV 340, RSU 350, and HMI 360 can vary at each control time step of the global multi-vehicle decision system 300. Most importantly, the number of NCVs and HMIs can vary significantly as traffic participants and potential passengers enter and leave the traffic network in which the global multi-vehicle decision system operates.

[0101] Figure 4A A schematic diagram of the feedback loop of a global multi-vehicle decision system 300 is shown. Given real-time information from infrastructure sensing 401, a centralized mapping and navigation module 405, and onboard sensors 414, 424, and 434 of each independent CAV 410, 420, and 430, the global multi-vehicle decision system 300 calculates a motion plan for a multi-layered guidance and control architecture provided to each independent CAV1,2,…,n. Based on the specific task assignment of the CAV (e.g., for the transportation of people and goods in a traffic network), the centralized navigation module 405 can calculate the route for each CAV from its current location to a desired destination or a sequence of desired destinations. According to some embodiments of the invention, based on the planned future route information of each CAV and given additional real-time sensor information about other traffic participants in the traffic network and the state of the environment, a global multi-vehicle decision system 300 calculates a coarse motion plan, which may include a sequence of entry and exit times and a sequence of average speed values ​​for each CAV in each conflict zone and in each conflict-free segment along the route of the CAV from its current location to its desired destination or a sequence of desired destinations (possibly including one or more planned intermediate stops along the route of the CAV).

[0102] Some embodiments of the present invention are based on the understanding that (semi-)autonomous driving planning and control can be effectively implemented using a multi-layered guidance and control architecture, typically implemented on-vehicle for each vehicle, including one or more layers of algorithms and techniques for decision-making, motion planning, vehicle control, and / or estimation. For example, the decision layer 411 of a CAV vehicle 410, given a motion plan from a global decision maker 300, current environmental conditions, and the behavior of other traffic participants, selects an appropriate driving behavior at any given time, for example, using an automaton combined with set reachability or formal language and vehicle decision optimization. Given a target behavior (including lane following, lane changing, or stopping) from the decision layer 411, a motion planning algorithm 412 aims to calculate a dynamically feasible and safe motion trajectory that can be tracked in real time by a relatively low-level (predictive) vehicle controller 413. Real-time sensor fusion and estimation using on-vehicle and infrastructure sensing information can be performed in each vehicle (e.g., 414) to provide feedback to higher-level algorithms for decision-making, motion planning, and control. Similar but potentially different multi-layered guidance and control architectures can be used for (semi) autonomous driving in other CAVs (such as 421-424 in the second vehicle 420 or 431-434 in the nth vehicle 430).

[0103] Popular approaches for (semi-)autonomous driving utilize a combination of finite state machines (FSMs) for decision-making (411, 421, and 431), sampled motion planning algorithms (412, 422, and 432), and model predictive control (MPC) algorithms for reference trajectory tracking (413, 423, and 433). One embodiment of the sampled motion planning algorithm uses probabilistic particle filtering to sample the input space and adds additional correction terms based on one or more driving requirements. One embodiment of the vehicle control prediction algorithm uses one or more iterations of the sequential quadratic programming (SQP) method to solve a linear time-varying or nonlinear MPC problem in real time. Some embodiments of the present invention are based on the understanding that the use of predictive algorithms for motion planning and reference tracking control in (semi-)autonomous driving can more effectively benefit from predictive information in the motion plan computed by the global multi-vehicle decision system 300. Embodiments of sensor fusion and estimation algorithms are based on motion-time domain estimation (MHE), extended or linear regression Kalman filtering, or particle filtering.

[0104] In some embodiments of the invention, different components of the multi-layered guidance and control architecture 411-414 may be implemented on-vehicle in each controlled vehicle 410, while other modules (such as the global multi-vehicle decision system 300, the mapping and navigation system 405, and some or all of the sensor fusion techniques for infrastructure sensing 401) may be implemented in the infrastructure (e.g., in the cloud and / or in one or more mobile edge computers (MECs)) to provide decision and / or feedback information to multiple connected vehicles in the transportation network.

[0105] In some embodiments of the invention, infrastructure sensing 401 corresponds to one or more roadside units (RSUs) and sensor fusion techniques to accurately detect the status of vehicles and the dynamic environment in a traffic network that includes connected and non-connected vehicles, autonomous, semi-autonomous and manually operated vehicles, and other traffic participants such as bicycles and pedestrians. The RSUs include one or more sensors, such as rangefinders, radar, lidar, and / or cameras.

[0106] Some embodiments of the present invention are based on the understanding that safety constraints regarding other dynamic traffic participants (i.e., non-vehicles, including bicycles and pedestrians) can be handled by obstacle avoidance techniques in the onboard modules of the multi-layered guidance and control architecture of each CAV, for example, through motion planning and / or vehicle control algorithms. Due to relatively low computational costs, motion planning and vehicle control algorithms can operate at relatively fast sampling rates compared to the global multi-vehicle decision-making system in some embodiments of the present invention, allowing for faster reaction times to unexpected changes in the dynamic behavior of other vehicles and / or other traffic participants. For example, real-time vehicle control algorithms are typically executed with update cycles of 50 to 100 milliseconds.

[0107] Some embodiments of the present invention are based on the following understanding: Figure 4A One or more components in the architecture shown can use different vehicle motion models with different modeling accuracies and / or different computational complexities. For example, a one-dimensional kinematic model can be used to describe the motion of each vehicle in the global multi-vehicle decision system 300 as follows:

[0108] d = v(t) out -t in -t wait )

[0109] For vehicles within the road section, t in and t out t represents the time of entering and leaving the road segment, respectively. wait Let v represent the waiting time at any intermediate planned stop within the road segment, v represent the average speed of the vehicle within the road segment, and d represent the vehicle's speed from time t within the road segment.in By time t out The distance traveled.

[0110] Some embodiments of the present invention are based on the following understanding: Figure 4A Components in the lower levels of the multi-layered guidance and control architecture depicted herein may use vehicle motion models with high modeling accuracy and potentially high computational complexity. For example, a one-dimensional kinematic model may be used in the global multi-vehicle decision system 300 and / or navigation module 405, while higher-dimensional (nonlinear) kinematic models may be used to describe vehicle motion for each individual vehicle 410, 420, and 430 in decision makers 411, 421, 431, motion planners 412, 422, 432, vehicle controllers 413, 423, 433, and / or estimation algorithms 414, 424, 434. Alternatively, one or more of the lower levels of the multi-layered guidance and control architecture may use higher-dimensional (nonlinear) dynamic models to describe vehicle motion based on force-torque balance, such as in (nonlinear) MPC-based vehicle controllers 413, 423, 433.

[0111] For example, in some embodiments of the invention, a single-track nonlinear vehicle model can be used in the MPC-based vehicle controllers 413, 423, 433, whose state is described by the vehicle's two-dimensional position, longitudinal and lateral velocities, yaw angle, and yaw rate. The single-track vehicle model groups the left and right wheels on each axle together. In some embodiments of the invention, vehicle models with even higher modeling accuracy and computational complexity (e.g., based on dual-track vehicle models) can be used, enabling accurate modeling of longitudinal and lateral load transfer between the vehicle's four wheels. In some embodiments of the invention, the nonlinear relationship between longitudinal and lateral tire friction forces and slip ratio and slip angle can be modeled using the Pacejka magic formula, which exhibits typical saturation behavior of tire forces. Under combined slip conditions, the coupling between longitudinal and lateral tire forces can be modeled, for example, using a friction ellipse or a weighted function.

[0112] Figure 4B A similar schematic diagram of the feedback loop for the global multi-vehicle decision system 300 is shown, where there is no centralized navigation module 405; instead, individual navigation systems 441, 442, and 443 can calculate routes for each CAV 410, 420, and 430 from its current location to a desired destination or a sequence of desired destinations. In some embodiments of the invention, each independent navigation system (e.g., 441) is executed onboard in the corresponding controlled vehicle 410, and future planned routing information from each of the independent navigation modules 441-443 can be transmitted to the global multi-vehicle decision system 300.

[0113] Figure 5A A block diagram of a global multi-vehicle decision system 300 is shown. This system computes motion plans for global multi-vehicle decisions in a traffic network by solving a mixed-integer optimal control optimization problem 510, based on communication with infrastructure and / or connected vehicles (V2X). Inputs to the global multi-vehicle decision system 300 may include mapping information 501, i.e., location data (e.g., GPS data) of conflict-free road segments, conflict zones, and trajectory lanes within each road segment. Furthermore, inputs to the global multi-vehicle decision system 300 may include feedback signals 505 regarding the status and planned routing information of the CAV and feedback signals 506 regarding the status and predicted routing information of the NCV. In some embodiments of the invention, the latter feedback signal is obtained directly or indirectly from sensing infrastructure (e.g., RSU) and / or from connected vehicles, where the connected vehicles may be autonomous, semi-autonomous, and / or manually operated vehicles. Feedback signal 505 regarding the status and planned routing information of the CAV may be referred to as a Type 1 feedback signal. Feedback signal 506 regarding the status and predicted routing information of the NCV may be referred to as a Type 2 feedback signal.

[0114] In some embodiments of the invention, the solution to the mixed-integer optimal control optimization problem 510 can be used to calculate the optimal sequence of entry and exit times and the optimal sequence of average speeds 515 for each CAV and each NCV along each segment of the traffic network along the future planned route of each CAV and the future predicted route of each NCV. Subsequently, in some embodiments of the invention, the optimal time and speed sequences 515 can be used to calculate the speed distribution of each CAV in the prediction time domain and / or one or more planned stops 520. The latter information is sent to a multi-layered guidance and control architecture 520 for each (semi-)autonomous CAV in the interconnected traffic network.

[0115] A traffic network may include one or more fully automated CAVs and / or one or more semi-automated CAVs. In some embodiments of the invention, the speed distribution of the fully automated CAVs may be controlled by a global multi-vehicle decision system 300 to control the entry and exit times of the CAVs in each road segment, thereby improving the overall safety, time efficiency, and energy efficiency of traffic flow in the traffic network. According to some embodiments of the invention, the global multi-vehicle decision system 300 may control whether and when semi-automated CAVs are permitted to enter one or more conflict zones within the traffic network along their planned future routes; however, the speed distribution of the semi-automated CAVs cannot be directly controlled by the global multi-vehicle decision system 300. For example, at a specific time step, the global multi-vehicle decision system 300 may instruct one or more semi-automated CAVs to stop at a traffic intersection (using a speed distribution that each semi-automated CAV can determine itself, and utilizing additional information from the global multi-vehicle decision system 300 regarding when it is safe to enter the traffic intersection), but the global multi-vehicle decision system 300 cannot directly control the speed distribution of one or more semi-automated CAVs within a road segment of the traffic network.

[0116] Some embodiments of the present invention are as follows: Figure 5A The global multi-vehicle decision system described herein uses long-term future route plans for each CAV 505. Some embodiments of the invention are based on the understanding that obtaining accurate future route predictions for each NCV can be challenging depending on the infrastructure system. Therefore, in some embodiments of the invention, the global multi-vehicle decision system 300 is implemented in a rolling time-domain manner based on up-to-date information from sensing infrastructure (e.g., RSU) and / or from connected vehicles. Some embodiments of the invention are based on the understanding that approximate short-term future route predictions 506 for the NCVs are sufficient and can generally be obtained relatively easily, for example, from each NCV's current position to the next conflict zone, and any discrepancies in the predictions can be adjusted by the inherent feedback mechanism of the rolling time-domain strategy. For example, an update period of 1-2 seconds allows for real-time computation by the global multi-vehicle decision system 300 while providing sufficiently fast updates 520 for the multi-layered guidance and control architecture of each CAV to account for erroneous predictions of NCV behavior.

[0117] Some embodiments of the present invention are based on the understanding that, given input information from V2X communication, a global multi-vehicle decision system 300 can be implemented by solving a constrained optimization problem 510 to compute a motion plan for each CAV. In some embodiments of the present invention, the constrained optimization problem can be a mixed-integer programming (MIP) problem, such as a mixed-integer linear programming (MILP) or mixed-integer quadratic programming (MIQP) problem. In some embodiments of the present invention, the MIP problem of the global multi-vehicle decision system 300 at each sampling time can be solved 510 using a global optimization algorithm (e.g., including branch and bound, branch cutting, and branch pricing methods). In other embodiments of the present invention, heuristic techniques can be used to compute feasible but suboptimal solutions to the MIP, such as rounding schemes, feasibility pump methods, approximate optimization algorithms, or the use of (deep) machine learning.

[0118] In some implementations of the global multi-vehicle decision system 300, the constrained optimization problem 510 includes the optimization (minimization) of one or more objectives, and imposes one or more equality and / or inequality constraints on safety and efficiency for all vehicles in the traffic network. For example, constraints may include vehicle motion models, speed limit constraints, time constraints for future route planning or route prediction, safety constraints in conflict zones, and occupancy constraints in conflict-free road segments. Objectives may include minimizing the travel time to the destination of each road segment in the future route plan or route prediction, minimizing the waiting time for each vehicle, maximizing the average speed, and minimizing acceleration for energy efficiency.

[0119] In some embodiments of the present invention, the mixed-integer optimization (minimization) problem of global multi-vehicle decision-making can be solved using the branch and bound (B&B) optimization method. This method searches for a globally optimal solution within the search space to generate the optimal control signal. The B&B optimization iteratively partitions the search space into nested region trees to find a solution with a globally optimal (minimum) objective value. The B&B method iteratively solves convex relaxations to compute lower bounds on the objective values ​​within the regions from the nested region trees. One or more regions can be pruned when the corresponding lower bound is greater than the upper bound of the currently known globally optimal objective value. When an integer feasible solution with an objective value less than the upper bound of the currently known globally optimal objective value is found, the upper bound of the globally optimal objective value can be updated.

[0120] Figure 5BA schematic diagram of the inputs and outputs of a global multi-vehicle decision system 300 according to some embodiments of the present invention is shown. Inputs to the global multi-vehicle decision system 300 may include mapping information 501 (i.e., location data (e.g., GPS data) for conflict-free road segments, conflict zones, and trajectory lanes within each road segment) and sensing and routing information for one or more vehicles 551-555, some of which may be controlled vehicles and some of which may be uncontrolled vehicles. For example, in a traffic network at a specific time, inputs to the global multi-vehicle decision system 300 may include sensing and future routing information for n CAVs and m NCVs, resulting in a total of n+m sensing and future routing information for vehicles 551-555. Outputs from the global multi-vehicle decision system 300 according to some embodiments of the present invention may include time and speed scheduling, i.e., the entry and exit time series and average speed series of each CAV in each road segment along the future planned routes of each CAV 561-565 within the traffic network.

[0121] Figure 6A An embodiment of a traffic network (a local area within the traffic network) is shown, which consists of multiple conflict-free road segments interconnected via three-way intersections (i.e., conflict zones), and each road segment consists of one or more track lanes. For example, a road segment designated S1600 consists of multiple lanes 601-604, each lane being referred to as an independent track lane, such as t1, t2, t3, and t4 within road segment S1600. Figure 6A In this implementation, it is assumed that each lane allows traffic flow in only one direction, as in segment S1600, with trajectory lane t1601 indicated by arrow 605 (away from the intersection) and trajectory lane t4604 indicated by arrow 606 (towards the intersection). Similarly, other segments can be defined as part of the traffic network, such as segment S3610 consisting of lanes 611-614 and segment S4620 consisting of lanes 621-624. Figure 6A A three-way traffic intersection (i.e., conflict zone), which may be referred to as additional road segment S2630 according to an embodiment of the present invention, is shown, which physically connects each of the conflict-free road segments 600, 610 and 620.

[0122] In some embodiments of the invention, each trajectory lane in the conflict zone corresponds to a potential path that a vehicle can follow within the conflict zone. For example, in Figure 6AIn section S2630, track lane t1631 corresponds to the path from track lane t4624 in section S4620 to track lane t1601 in section S1600 (left turn). For example, track lane t2632 in section S2630 corresponds to the path from track lane t4624 in section S4620 to track lane 614 in section S3610 (right turn). For example, track lane t5635 in section S2630 corresponds to the path from track lane t1611 in section S3610 to track lane t1601 in section S1600 (straight ahead). For example, track lane t1631 in section S2630 corresponds to the path from track lane t1611 in section S3610 to track lane t1601 in section S1600 (straight ahead). 12 639 corresponds to the path (right turn) from track lane t4604 in road segment S1600 to track lane t1621 in road segment S4620. Similarly, each track lane in road segment S2630, such as track lanes t1, t2, t3, ..., t4604, can be defined based on the path connecting a track lane in one adjacent road segment to a track lane in another adjacent road segment. 11 ,t 12 (and possibly even more track lanes).

[0123] The embodiments of the present invention are based on the understanding that a global multi-vehicle decision-making system needs to know, for each vehicle and for each segment of the route from the vehicle's current location to its desired destination, which trajectory lane (and in which segment) the vehicle is currently traveling on and which trajectory lane the vehicle plans to travel on in the future (in each segment of its route). For example, according to some embodiments of the present invention, two vehicles traveling on the same trajectory lane in any segment cannot overtake each other and therefore cannot change their order unless one of the two vehicles is likely to switch to a different trajectory lane within that segment. Furthermore, according to some embodiments of the present invention, conflict zones (e.g., such as...) Figure 6A The risk of collision between two vehicles at the intersection shown may or may not exist, depending on the trajectory lane each vehicle follows within the conflict zone. For example, when in conflict zone S2630, one vehicle is traveling in trajectory lane t9 and the other vehicle is traveling in trajectory lane t... 11 When traveling in lane t9, there is no risk of collision between the two vehicles because the two lanes are parallel to each other and therefore do not physically intersect. However, when one vehicle is traveling in lane t9 and the other is traveling in lane t3 within the conflict zone S2630, there is a potential risk of collision between the two vehicles because the two lanes physically intersect each other.

[0124] Figure 6B As shown in Figure 6AThe same embodiment of the traffic network depicted (a local area within the traffic network) is shown, but it further illustrates possible routes that vehicle V1 640 and vehicle V2 645 might plan to follow within the traffic network. For example, vehicle V1 640 is currently traveling in track lane t3 in segment S4620 and may plan to travel along track lane t3 in conflict zone S2630 (turning left at the intersection) to continue in track lane t2 in segment S1600 after crossing the conflict zone (as shown in future vehicle position 641). Similarly, vehicle V2 645 is currently traveling in track lane t3 in segment S1600 and may plan to travel along track lane t9 in conflict zone S2630 (going straight across the intersection) to continue in track lane t3 in segment S3610 after crossing the conflict zone (as shown in future vehicle position 646).

[0125] According to some embodiments of the invention, if both vehicle V1 640 and vehicle V2 645 are connected automated vehicles (CAVs), the global multi-vehicle decision system can calculate the motion plan for each CAV and determine whether vehicle V1 640 should cross the conflict zone S2630 along trajectory lane t3 before vehicle V2 645 may cross the conflict zone S2630 along trajectory lane t9, or whether vehicle V1 640 should cross the conflict zone S2630 along trajectory lane t3 after vehicle V2 645 has completed crossing the conflict zone S2630 along trajectory lane t9, in order to avoid any possible collision between the two vehicles, since trajectory lanes t3 and t9 physically intersect each other in the conflict zone S2630. Alternatively, if one or more vehicles are uncontrolled vehicles (NCVs), according to some embodiments of the invention, the global multi-vehicle decision system can calculate the motion plan for each CAV while predicting the motion plan for each NCV, for example assuming that each NCV is intended to follow every traffic rule applied within the traffic network. In some cases, the movement plan may be a rough movement plan for each CAV along a planned future route. The implementation of this invention is based on the understanding that approximate short-term forecasts of the NCV are sufficient (e.g., until the next conflict zone) and that any discrepancies in the forecasts can be adjusted through the inherent feedback mechanism implemented in the rolling time domain of a global multi-vehicle decision system.

[0126] Figure 6C Some embodiments of the invention are shown that can be derived from... Figure 6B An example of a global multi-vehicle decision system calculated motion plans for vehicle V1 660 and another vehicle V2 670 within a traffic network is shown. Figure 6CThe motion plans of vehicles V1660 and V2670 are shown by plotting the travel distance 650 for each trajectory relative to time 655. Figure 6C As shown in embodiment 680, vehicle V1 640 only begins to cross the conflict zone S2664 along the track lane t3 after vehicle V2 645 has completed crossing the conflict zone S2672 along the track lane t9, in order to avoid any possible collision between the two vehicles, since the track lanes t3 and t9 physically intersect each other in the conflict zone S2630. Figure 6C The motion plan of vehicle V1 is shown to consist of one or more portions of a trajectory corresponding to each of the road segments and trajectory lanes along the planned future route of the vehicle, such as trajectory 661 on trajectory lane t3 in road segment S4, followed by trajectory 664 on trajectory lane t3 in conflict zone S2, and then trajectory 665 on trajectory lane t2 in road segment S1. Similarly, the motion plan of vehicle V2 can consist of, for example, trajectory 671 on trajectory lane t3 in road segment S1, trajectory 672 on trajectory lane t9 in conflict zone S2, and trajectory 673 on trajectory lane t3 in road segment S3.

[0127] Some embodiments of the present invention are based on the understanding that for any vehicle within a road segment, there are typically multiple trajectory lanes, each with a different average speed value, in order to reach the end of the road segment before a specific time. For example, as Figure 6C As shown in trajectory 663, multiple trajectories have different average speed values ​​for vehicle V1 640 on trajectory lane t3 in segment S4, so that vehicle V1 can reach the end of segment S4 before the instant 680 when it can safely enter conflict zone S2, because vehicle V2 645 completes its crossing of conflict zone S2 along trajectory lane t9672 at that instant. According to some embodiments of the invention, each of the various trajectories 663 can be calculated by a global multi-vehicle decision system as part of the motion plan for vehicle V1 (on trajectory lane t3 in segment S4) to optimize one or a combination of multiple objectives (e.g., travel time, waiting time, and / or energy efficiency). More specifically, from Figure 6C It can be observed that before vehicle V1 enters the conflict zone S2, trajectory 661 can be calculated with a relatively large average speed and therefore a small travel time but a relatively large waiting time. Alternatively, before vehicle V1 enters the conflict zone S2 in its planned motion, trajectory 662 can be calculated with a relatively small average speed and therefore a relatively large travel time but a small waiting time or even no waiting time at all.

[0128] Some embodiments of the present invention are based on the understanding that for CAVs in a traffic network, there is a trade-off between minimizing total travel time (e.g., for each vehicle in each segment along its route) and minimizing total waiting time (e.g., for each vehicle before it reaches the end of a segment along its route and enters a subsequent segment). Specifically, according to some embodiments of the present invention, a global multi-vehicle decision system can instruct CAVs to travel as fast as possible in each segment along their route while adhering to all safety constraints and speed limits, which will result in relatively short travel times, but may additionally result in relatively long waiting times. Alternatively, according to some embodiments of the present invention, the global multi-vehicle decision system can instruct one or more CAVs to travel slower (immediately or in the future) when it is predicted that a CAV needs to wait at the end of a segment before safely entering a subsequent segment along its route, in order to reduce total waiting time. Some embodiments of the present invention are based on the understanding that the global multi-vehicle decision system can use alternative or additional objectives to calculate the motion plan for each CAV in the traffic network, such as minimizing the total amount of acceleration and / or deceleration of each vehicle along its route in order to optimize total energy efficiency.

[0129] Figure 7 A block diagram of a global multi-vehicle decision system 300 is shown. This system computes a motion plan by solving a mixed-integer programming (MIP) problem 720, based on communication with infrastructure and / or connected vehicle (V2X) systems. The motion plan comprises, for example, the time and speed scheduling of each of the controlled and / or semi-controlled vehicles 730 in a traffic network. Inputs to the global multi-vehicle decision system 300 may include mapping information 501, i.e., location data (e.g., GPS data) of conflict-free road segments, conflict zones, and trajectory lanes within each road segment. Furthermore, inputs to the global multi-vehicle decision system 300 may include feedback signals 705 regarding sensing and (planned or predicted) routing information for each of the controlled, uncontrolled, and / or semi-controlled vehicles in the traffic network. This is according to some embodiments of the invention, and as... Figure 7 As shown, the input to the global multi-vehicle decision system 300 can be used to construct matrices and vectors in the MIP data to define the objective, equality and inequality constraints 710, so as to construct and solve the resulting mixed integer programming (MIP) problem 720.

[0130] In some embodiments of the present invention, the MIP problem can be a mixed-integer linear programming (MILP) problem or a mixed-integer quadratic programming (MIQP) problem. For example, Figure 7The MIP shown is a MIQP consisting of a linear quadratic objective function 721, defined by a positive semi-definite Hessian matrix H ≥ 0 and a gradient vector h. The MIP may additionally include one or more linear inequality constraints 722 defined by a constraint matrix C and a constraint vector c, and one or more linear equality constraints 723 defined by a constraint matrix F and a constraint vector f. Finally, the MIQP 720 for the global multi-vehicle decision system 300 may include one or more integrity constraints 724, which constrain one or more optimization variables z. j The restriction is to the set of integer values, that is, In some embodiments of the present invention, the objective function 721 is a linear function (i.e., H = 0), such that the resulting MIP 720 is a MILP optimization problem.

[0131] Figure 8A An embodiment of equality and / or inequality constraints 800, which can be implemented by a global multi-vehicle decision system 300 to solve a MIP problem 720, is shown. In some embodiments of the invention, the MIP problem 720 may include vehicle dynamic equality constraints 801, vehicle routing inequality constraints 802, speed limit constraints 803 for controlled and / or semi-controlled vehicles (CAVs), speed limit constraints 804 for uncontrolled vehicles (NCVs), safety constraints 805 for each conflict zone, and occupancy constraints 806 for each (conflicting and / or non-conflicting) road segment in the traffic network.

[0132] In some embodiments of the invention, the global multi-vehicle decision system is based on a solution of MIP 720 that includes inverse velocities in the optimization variables, for example, such that a one-dimensional kinematic model of each of the CAV and NCV can be defined by the following linear vehicle dynamics equation constraint 801:

[0133]

[0134] Among them, variables and These represent the entry time and the exit time, respectively, and the variables... Represents road segments j∈J in the transportation network s For each vehicle i∈I v The inverse velocity. Furthermore, the value... and These are fixed quantities calculated and / or provided to the global multi-vehicle decision system, and they respectively represent each road segment j∈J in the traffic network. s For each vehicle i∈I v The (planned or predicted) distance to be traveled and the (planned or predicted) waiting time.

[0135] In some embodiments of the present invention, at each time step of the rolling time-domain implementation of the global multi-vehicle decision system, if the vehicle is currently located within a road segment, the distance to be traveled is... It can be calculated as from vehicle i∈I v The current position to the segment j∈J along the track lane s The distance to the destination. If vehicle i∈I v Currently located on road segment j∈J s Beyond that, the distance to travel It can be calculated as vehicle i∈I v In road segment j∈J s The total distance to be traveled along the (planned or predicted) trajectory lane. More specifically, in some embodiments of the invention, the quantity... It can be defined as follows:

[0136]

[0137] Where, if vehicle i∈I v There are no planned or predicted road segments j∈J that are not in the transportation network. s If proceeding in the middle, then Otherwise, the distance to travel It can be calculated as vehicle i∈I v In road segment j∈J s The total distance traveled along the (planned or predicted) trajectory lane minus the distance traveled by vehicle i∈I v In road segment j∈J s The current position p in i Distance traveled D j (p i That is, the vehicle dynamics equality constraint 801 in MIP problem 720.

[0138] In some embodiments of the present invention, the waiting time is defined at each time step of the rolling time-domain implementation of the global multi-vehicle decision-making system. The road segments j∈J can be calculated as part of the transportation network. s For each vehicle i∈I v The planned or predicted total waiting time. More specifically, if vehicle i∈I v There is currently no plan or prediction that the route will not be in the transportation network along the road segment j∈J. s If the internal waiting time is less than the total waiting time, then the total waiting time is less than the total waiting time. Alternatively, if a vehicle is planned or predicted to travel along its route in the traffic network on road segment j∈J s If an intermediate stop is made (e.g., for picking up or dropping off passengers), the waiting time... The total time required to make this intermediate stop along its route is predicted or estimated in order to define the vehicle dynamics equation constraint 801 for MIP problem 720.

[0139] Some embodiments of the present invention are based on the understanding that one or more of these quantities (such as values) and The time steps can vary from one time step to the next; these time steps are those implemented by the global multi-vehicle decision-making system in the rolling time domain. Any discrepancies in the prediction of vehicle behavior can be adjusted through the inherent feedback mechanism of the rolling time domain strategy.

[0140] In some embodiments of the present invention, the global multi-vehicle decision system is based on the solution of MIP 720, which includes vehicle routing inequality constraints 802 to enforce the routing of each vehicle i∈I. v Planning or predicting future routes Leaving and entering subsequent road sections φ i (k-1) and φ i The order between (k). For example, some embodiments of the present invention use the following formula of vehicle routing inequality constraint 802:

[0141]

[0142] This forces the vehicle i∈I to be executed. v Within the transportation network, leave the road segment φ along its route. i After (k-1), you can only enter section φ. i (k).

[0143] In some embodiments of the invention, the global multi-vehicle decision system is based on a solution of MIP 720 including one or more speed limit constraints 803 for controlled or semi-controlled vehicles (CAVs). Some embodiments of the invention limit the speed of each CAV to remain positive (excluding, for example, parking maneuvers where speeds may alternate between positive and negative values), and the speed can be limited to remain below a maximum permissible value that may differ for each segment within the traffic network. For example, with respect to the inverse speed variable, the speed limit constraint 803 for the CAV can be formulated in MIP 720 as follows:

[0144] In some embodiments of the invention, the global multi-vehicle decision system is based on a solution to a MIP 720 that includes one or more speed limit constraints 804 for uncontrolled vehicles (NCVs). Some embodiments of the invention are based on the understanding that the speed of an NCV cannot be directly controlled, preventing the global multi-vehicle decision system from freely choosing any sequence of speed values ​​for each NCV along its predicted route. Instead, some embodiments of the invention aim to limit the predicted speed of each NCV to keep it relatively close to its current speed value provided by the state estimation and sensing module (e.g., by the RSU in the sensing infrastructure).

[0145] In some embodiments of the invention, the inverse speed variable of the NCV corresponds to an inverse prediction of the average speed of each NCV along each segment of its predicted route, for example based on the current speed and / or acceleration value of the NCV.

[0146] For example, some embodiments of the present invention include the following speed limiting constraint 804 for NCVs.

[0147]

[0148] in, It can be for NCVi∈I nc The measured or estimated current speed value. In some embodiments of the invention, the speed value in the NCV speed limit constraint 804. It can be computed by a predictor module, for example, based on (deep) machine learning techniques, which is designed to predict the most likely driving behavior of each NCV based on recently collected past (real-time) data on the driving behavior of each NCV in the traffic network.

[0149] Figure 8B An embodiment of the safety constraint formula 805 for each conflict zone in the MIP problem 720 solved by the global multi-vehicle decision system 300 is shown. In some embodiments of the invention, for each road segment j∈J that is a conflict zone... s (That is, j∈C in the traffic network), if it is planned or predicted that vehicles i and k will travel on conflict trajectory lanes within the conflict zone j∈C in the traffic network, then safety constraint 810 restricts the two vehicles i,k∈I. v They will never be simultaneously located within the conflict zone j∈C. More specifically, if it is planned or predicted that vehicle i and vehicle k will travel on the conflict trajectory lane within the conflict zone j∈C, then vehicle i will leave the conflict zone before vehicle k enters, or vehicle k will leave the conflict zone before vehicle i enters.

[0150]

[0151] Where ∈≥0 is a small positive fixed value, which defines the tolerance of the time constraint. In some embodiments of the present invention, if it is planned or predicted that vehicles i and k will travel on non-conflict trajectory lanes within the conflict zone j∈C in the traffic network, then both vehicles i,k∈I may be allowed to travel on lanes without conflict. v Both vehicles are allowed to be within the conflict zone j∈C. For example, if two vehicles are traveling on the same track lane within the conflict zone, or if they are traveling on two track lanes that do not physically intersect each other, then both vehicles are allowed to be within the conflict zone simultaneously.

[0152] Some embodiments of the present invention are based on the following understanding: for each pair of two vehicles i,k∈I in each conflict zone j∈C in the traffic network v MIP problem 720 may include the use of the Big M rule 820 regarding conflict zones and the use of a binary decision variable b. i,j,k ∈{0,1}, i.e., decision variable b i,j,k The value is restricted to a safety constraint of 0 or 1: 810

[0153]

[0154] Where M>0 is a positive fixed value, which is chosen to be sufficiently large, and ∈≥0 is a positive fixed value, which defines the tolerance of the time constraint and is generally chosen to be relatively small. Some embodiments of the invention are based on the understanding that the security constraints with respect to the conflict zone 805 are generally symmetric, and therefore these constraints only need to be implemented for each unique pair of vehicles in the MIP problem 720. In some embodiments of the invention, alternative formulas for the security constraints with respect to each conflict zone 805 can be used in the MIP problem 720, for example, using the convex hull formula to produce a more stringent convex relaxation, although the cost is that it may lead to a high-dimensional optimization problem.

[0155] Figure 8C An embodiment of the occupancy constraint formula for each road segment 806 in the MIP problem 720 solved by the global multi-vehicle decision system 300 is shown. In some embodiments of the invention, for each road segment j∈J in the traffic network... s Occupation constraint 830 restricts two vehicles i,k∈I v In the road segments j∈J that they plan or predict in the transportation network s When vehicles are traveling on the same track lane within a given area, their order cannot be exchanged at any time. More specifically, if it is planned or predicted that vehicle i and vehicle k are in the same lane on road segment j∈J... sIf two vehicles i, k∈I are traveling on the same track lane, then vehicle i enters and leaves segment j before vehicle k enters and leaves segment j, or vehicle i enters and leaves segment j after vehicle k enters and leaves segment j. 830. In some embodiments of the invention, if it is planned or predicted that two vehicles i, k∈I v In each of its planned or predicted routes, in the same subsequent segment l∈J s The subsequent road segment j∈J s If vehicles are traveling on the same track lane, then vehicle i enters and leaves track j and l before vehicle k enters and leaves track j and l, or vehicle i enters and leaves track j and l after vehicle k enters and leaves track j and l.

[0156]

[0157] Wherein, ∈≥0 is a small positive fixed value, which defines the tolerance of the time constraint. In some embodiments of the invention, the occupancy constraint 830 is enforced only for specific trajectory lanes in road segments where no overtaking between two vehicles is permitted (e.g., by performing a double lane change).

[0158] Some embodiments of the present invention are based on the following understanding: for each road segment l∈J in the transportation network s Neutralize / or neutralize every pair of vehicles i,k∈I in each conflict-free road segment j∈F. v MIP problem 720 may include information about using the Big M rule 840 for road segments and using a binary decision variable b. i,j,k ∈{0,1}, i.e., decision variable b i,j,k The value is restricted to 0 or 1 for road segment occupancy constraint 830:

[0159]

[0160]

[0161] Where M>0 is chosen as a sufficiently large positive fixed value, and ∈≥0 is a positive fixed value that defines the tolerance of the time constraint and is typically chosen to be relatively small. Some embodiments of the invention are based on the understanding that the occupancy constraints with respect to road segment 806 are generally symmetric, and therefore, these constraints need to be implemented only for each unique pair of vehicles in the MIP problem 720. In some embodiments of the invention, alternative formulas for the occupancy constraints of each road segment 806 can be used in the MIP problem 720, for example, using the convex hull formula to produce a more stringent convex relaxation, although the cost is that it may lead to a high-dimensional optimization problem.

[0162] Figure 8DAn embodiment of the present invention is shown, which may include an objective in a weighted multi-objective minimization 850 of a MIP problem 720 solved by a global multi-vehicle decision system 300. For example, the weighted multi-objective minimization function 850 in the MIP problem 720 may be as follows:

[0163]

[0164] For each vehicle in the transportation network, J t This corresponds to minimizing the travel time to 851, J. w This corresponds to minimizing the waiting time to 852, J. p This corresponds to minimizing the velocity tracking penalty by 853, and J a This can correspond to minimizing acceleration and / or deceleration values, i.e., minimizing energy consumption 854.

[0165] In some embodiments of the present invention, the travel time minimization 851 in the MIP problem 720 is performed as minimizing the sum of the entry and / or exit times of each vehicle in the transportation network. In some embodiments of the present invention, the travel time minimization 851 in the MIP problem 720 is performed as minimizing the weighted sum of the times each vehicle takes to reach its desired destination in the transportation network. For example, in some embodiments of the present invention, the travel time minimization 851 in the MIP problem 720 is as follows:

[0166]

[0167] in, Defined as vehicle i∈I v The total distance traveled within the planned or predicted route of the transportation network up to the end of the k-th segment. The latter objective function J t 851 aims to minimize the travel time of each vehicle along its planned or predicted route to the end of each segment, with corresponding weights ω1≥0, and considering the total distance traveled by each vehicle i and segment k. Normalization.

[0168] Some embodiments of the present invention are based on the following understanding: for each vehicle i∈I v And for planned or predicted routes of vehicles within the transportation network. Each pair of subsequent road segments φ i (k-1) and φ i (k) defines the waiting time as the entry time. and departure time The difference between them. The latter difference corresponds to the expected vehicle i∈I. v From segment φ along the vehicle's planned or predicted routei (k-1) Cross to the subsequent road segment φ i (k) The time period of waiting (i.e., at speed 0). In some embodiments of the invention, the weighted multi-objective minimization 850 of the MIP problem 720 includes minimizing the sum of the waiting times for each vehicle and for each pair of subsequent road segments in its planned or predicted future route in the traffic network 852:

[0169]

[0170] Minimizing the waiting time for CAV and NCV 852 has corresponding weight values ​​ω2≥0 and ω4≥0, respectively. In some embodiments of the invention, the weight value ω4 is chosen to be significantly greater than ω2 (i.e., ω4>>ω2) because the global multi-vehicle decision system cannot directly control whether and when the NCV will wait between two subsequent segments along its predicted future route, nor can it directly control the time the NCV will wait between two subsequent segments along its predicted future route.

[0171] In some embodiments of the invention, the weighted multi-objective minimization 850 in the MIP problem 720 includes, for example, minimizing the speed tracking penalty 853 for uncontrolled vehicles (NCVs) to limit the speed of each NCV to remain relatively close to the predicted average speed of the NCV along each segment of its future route within the traffic network. For example, in some embodiments of the invention, the MIP objective 850 includes minimizing the sum of the inverse speed variables of each NCV in each segment of the traffic network:

[0172]

[0173] The corresponding weight value is ω3≥0, and this minimization of the sum of the inverse speed variables corresponds to the maximization of the speed of each NCV, so as to limit the speed of each NCV to remain relatively close to the maximum value enforced by the (inverse) speed limit constraint 804 (using the predicted average speed of the NCV along each segment of its future route within the traffic network).

[0174] In some embodiments of the invention, the weighted multi-objective minimization 850 in the MIP problem 720 solved by a global multi-vehicle decision system includes minimizing the energy consumption 854 of each vehicle in the traffic network to improve the overall energy efficiency of interconnected traffic flows. For example, in some embodiments of the invention, the MIP objective 850 includes minimizing the acceleration and / or deceleration of each vehicle in the traffic network, which is closely related to energy consumption. If the motion model of the system dynamics for each vehicle includes acceleration and / or deceleration variables, the latter objective function can be implemented directly. Alternatively, in some embodiments of the invention, the minimization of energy consumption 854 is performed for each vehicle and each pair of subsequent road segments (e.g., vehicle i ∈ i v φ i (k-1) and φ i The approximation is made by minimizing the sum of the absolute changes in the average (inverse) velocity of (k), as follows:

[0175]

[0176] The corresponding weight value is ω5≥0, and it can be defined as follows: That is, it means that for each vehicle i∈I v The reciprocal of the current or predicted average speed value. Some embodiments of the invention are based on the understanding that the 1-norm in energy consumption minimization 854 can be reconstructed using additional optimization variables and additional inequality constraints from MIP problem 720. Alternatively, in some embodiments of the invention, energy consumption minimization 854 is approximated by minimizing the sum of the squared changes in the average (inverse) speed of each vehicle with respect to each pair of subsequent road segments along its future planned or predicted route within the traffic network.

[0177] In some embodiments of the present invention, a method for using n s Interconnected road segments are composed and support up to n c Controlled vehicles (CAVs) and up to n nc The global multi-vehicle decision system for uncontrolled vehicle (NCV) traffic networks, which needs to be solved at each time step of the global multi-vehicle decision system, can include the following optimization variables, equality and inequality constraints:

[0178] ● Multiple continuous optimization variables: 3n s ( nc +n nc For example, this includes information about each vehicle i∈I in the transportation network. v and each road segment j∈J s The entry and exit time variables and (inverse) velocity variables;

[0179] • Multiple binary optimization variables: This includes a unique binary optimization variable for each pair of vehicles on each road segment, for example, to enforce safety constraints on conflict zone 805 and / or enforce occupancy constraints on road segment 806;

[0180] Multiple equality constraints n s (n c +n nc For example, it includes vehicle dynamics equation constraints 801 for each vehicle in each road segment;

[0181] • Multiple inequality constraints (excluding simple boundary conditions for variables):

[0182] · Examples include vehicle routing inequality constraint 802, safety constraints of conflict zone 805, and / or occupancy constraints of road segment 806;

[0183] Where |C| and |F| represent the total number of conflict zones and non-conflict road segments, respectively. Some embodiments of the present invention are based on the understanding that, for practical applications of global multi-vehicle decision-making, it may be desirable for the MIP problem 720 to have a fixed dimension, such as supporting a (relatively large) upper limit on road segments, conflict zones, controlled and uncontrolled vehicles in the traffic network.

[0184] Some embodiments of the present invention are based on the understanding that redundant optimization variables can be automatically removed by pre-solution routines in numerical optimization algorithms used to solve the MIP problem 720 in the global multi-vehicle decision system 300. In some embodiments of the present invention, one or more redundant optimization variables can be explicitly fixed to specific values ​​by adjusting the corresponding simple boundary of each redundant optimization variable in the MIP problem 720.

[0185] For example, in some embodiments of the present invention, redundant optimization variables can be fixed and removed according to simple rules, for example,

[0186]

[0187]

[0188]

[0189]

[0190]

[0191] The first equation corresponds to the road segment j∈J that already exists. s Vehicle i∈I v This allows the corresponding entry time to be fixed at zero, i.e. Similarly, for road segments j∈J that are not planned or predicted to enter...s Any vehicle i∈I v (Second equation, based on the distance to be traveled) Both the entry and exit time variables can be fixed at zero, i.e. The third equation corresponds to the situation where vehicle i or vehicle k does not plan or predict entering road segment j∈J. s That is, if or If true, the fixed binary variable b i,j,k =0, which can be used to remove binary variables and corresponding inequality constraints from MIP problem 720. As a final example, both vehicle i and vehicle k currently exist in road segment j∈J. s Within this context, and assuming the fixed value of the binary optimization variable depends on the order of vehicles within the road segment, based on equations four and five above, the binary variable can be fixed to either 0 or 1, i.e., b i,j,k =0 or b i,j,k =1.

[0192] Some embodiments of the present invention are based on the understanding that after fixing one or more binary optimization variables, one or more corresponding inequality constraints may become redundant and can be removed by setting the corresponding lower bound to -∞ and / or by setting the corresponding upper bound to ∞. Some embodiments of the present invention are based on the understanding that one or more vehicle dynamic equality constraints 801 can be used to remove one or more continuous optimization variables; for example, vehicle dynamic equality constraints 801 can be used to remove (inverse) velocity variables from MIP problem 720.

[0193] Figure 9A A schematic diagram of an embodiment of a binary decision variable search tree according to some implementations is shown, which represents a nested search region tree for integer feasible solutions to MIP problem 720 in a global multi-vehicle decision system 300. Figure 9A A schematic representation of the branch and bound method is illustrated by showing a binary search tree 900 under a specific iteration of a mixed-integer optimization algorithm. This branch and bound method can be used to implement a global multi-vehicle decision system in some embodiments. The main idea of ​​the branch and bound (B&B) method is to sequentially create partitions of the original MIP problem 720 and then attempt to solve those partitions, where each partition corresponds to a specific region of the search space of discrete optimization variables. In some embodiments of the invention, the branch and bound method selects partitions or nodes and selects discrete optimization variables to branch that partition into smaller partitions or search regions, thereby producing nested partition trees or search region trees.

[0194] For example, partition P1901 represents a discrete search region that can be divided or branched into two smaller partitions or regions P2902 and P3903, i.e., a first region and a second region nested within a common region. The first and second regions are disjoint, i.e., their intersection is empty P2∩P3=φ907, but together they form the original partition or region P1, i.e., after branching, the union P2∪P3=P1906 holds. The branch-and-bound method then solves an integer relaxation optimization problem for both the first and second partitions or regions in the search space, producing two solutions (local optima) that can be compared with each other and with a currently known upper bound on the optimal objective value. If the performance metrics of the first and / or second partitions or regions are lower than the currently known upper bound on the optimal objective value of the MIP problem, the first and / or second partitions or regions can be pruned. If the first region, the second region, or both regions produce a discrete feasible solution to the MIP problem, the upper bound can be updated. The branch-and-bound method then continues by further partitioning the remaining regions in the currently nested region tree.

[0195] While solving each partition may still be challenging, it is quite efficient to obtain a local lower bound for the optimal objective value by solving for local relaxations of mixed integer programming (MIP) or by using duality. If the MIP solver happens to obtain an integer feasible solution while solving for local relaxations, it can use that integer feasible solution to obtain a global upper bound for the mixed integer solutions of the original MIP problem in a global multi-vehicle decision system. This can help avoid solving or branching over some partitions that have already been created; that is, these partitions or nodes can be pruned. The general algorithmic idea for such partitioning can be represented as a binary search tree 900, including a root node (e.g., P1901 at the top of the tree) and leaf nodes (e.g., P4904 and P5905 at the bottom of the tree). Furthermore, nodes P2902 and P3903 are usually called the direct children of node P1901, while node P1901 is called the parent node of nodes P2902 and P3903. Similarly, P4904 and P5905 are child nodes of their parent node 902. In some embodiments of the present invention, the MIP problem can be a mixed-integer linear programming (MILP) problem or a mixed-integer quadratic programming (MIQP) problem.

[0196] Figure 9B A block diagram is shown of a branch-and-bound mixed integer optimization algorithm, according to some embodiments, for searching for integer feasible optimal solutions based on nested search region trees and corresponding lower / upper bound values. In some embodiments, Figure 9BThe block diagram of the branch-and-bound mixed-integer optimization algorithm shown can be used to implement a global multi-vehicle decision system. The branch-and-bound method, based on MIP data 710 consisting of matrices and vectors, initializes the branch search tree information of the mixed-integer programming (MIP) at the current time step of the global multi-vehicle decision system 910. Initialization can additionally utilize branch search tree information and MIP solution information 912 from previous time steps to generate a warm-start initialization 910 for the current time step. The main objective of the optimization algorithm is to construct a lower bound and an upper bound for the target values ​​of the mixed-integer solution. At step 911, if the gap between the lower bound and the upper bound is less than a specific tolerance value, a mixed-integer optimal solution 955 is found.

[0197] As long as the gap between the lower and upper bounds at step 911 is greater than a specific tolerance value, and the optimization algorithm has not yet reached its maximum execution time, the branch and bound method continues to iteratively search for the optimal solution for mixed integers 955. Each iteration of the branch and bound method begins by selecting the next node in the tree corresponding to the next region or partition of the search space for integer variables, where the possible variables are fixed based on a pre-solved branching technique 915. After selecting the node, the corresponding integer relaxation problem is solved, where the possible variables are fixed based on a pre-solved branching technique 920.

[0198] If the integer relaxation problem has a feasible solution, the resulting relaxed solution provides a lower bound on the target value for that particular region or partition of the integer variable search space. At step 921, if the target is determined to be greater than the currently known upper bound of the target value of the optimal mixed integer solution, the selected node is pruned or removed from the branch tree 940. However, at step 921, if the target is determined to be lower than the currently known upper bound, and the relaxed solution is integer feasible 925, then at step 930, the currently known upper bound and the corresponding mixed integer solution estimate are updated.

[0199] If the integer relaxation problem has a feasible solution and the objective is below the currently known upper bound 921, but the relaxation solution is not yet integer feasible, then the global lower bound of the objective can be updated to the minimum objective value 935 of the remaining leaf nodes in the branch tree, and the selected node 940 can be pruned from the tree. Additionally, starting from the current node, discrete variables with fractional values ​​are selected for branch 945 according to a specific branching strategy, so as to create and supplement the resulting subproblems corresponding to the region or partition of the discrete search space as child nodes 950 of that node in the branch tree.

[0200] A crucial step in branch and bound methods is creating partitions, specifically choosing which node (915) and which discrete variable to use for branching (945). Some implementations are based on branching one of the binary optimization variables with fractional values ​​in the solutions with integer relaxation. For example, if a particular binary optimization variable d∈{0,1} has a fractional value as part of the optimal solution with integer relaxation, some implementations create two partitions for mixed integer programming by adding the equality constraint d=0 to one subproblem and the equality constraint d=1 to another subproblem, respectively. Some implementations are based on a reliability branching strategy for variable selection (945), which aims to predict future branching behavior based on information from previous branching decisions.

[0201] Some implementations are based on branch-and-bound methods using a depth-first node selection strategy, which can be implemented using a Last-In-First-Out (LIFO) buffer. The next node to be solved is selected as one of the children of the current node, and this process is repeated until a node is pruned (i.e., the node is infeasible, optimal, or dominated by the currently known upper bound), then a backtracking process is performed. Conversely, some implementations are based on branch-and-bound methods using an optimal-first strategy, selecting the node with the current lowest local lower bound. Some implementations employ a combination of depth-first and optimal-first node selection methods, where a depth-first node selection strategy is used until an integer feasible solution is found, followed by an optimal-first node selection strategy in subsequent iterations of the branch-and-bound-based optimization algorithm. The motivation for the latter implementation is to find an integer feasible solution early in the branch-and-bound process (depth-first) to allow for early pruning, followed by a more greedy search for better feasible solutions (optimal-first).

[0202] The branch and bound method continues to iterate until one or more of the following conditions are met:

[0203] • Reach the processor's maximum execution time.

[0204] • All nodes in the branch search tree are pruned, making it impossible to select new nodes to solve for convex relaxation or to branch.

[0205] • For mixed integer solutions, the optimality gap between the global lower bound and the global upper bound of the objective is smaller than the tolerance.

[0206] Figure 10A block diagram of a global multi-vehicle decision system 1000 for calculating the motion plans of one or more controlled vehicles in a traffic network with one or more interconnected conflict zones, according to some embodiments of the present invention, is shown. The global multi-vehicle decision system 1000 is executed in the cloud or in one or more mobile edge computers (MECs). The system may additionally require one or more edge devices (e.g., RSUs for infrastructure-based sensing) or operatively connected to such a set of sensors to collect traffic information near one or more interconnected conflict zones in the traffic network. In some embodiments of the invention, the global multi-vehicle decision system is designed to send an optimal sequence 1022 of entry and exit times and average speeds and / or a speed distribution and a planned stop sequence 1024 to each connected autonomous vehicle (CAV) along a future planned route in the traffic network.

[0207] The global multi-vehicle decision system 1000 includes multiple interfaces for connecting the decision system 1000 to other systems and devices. For example, the decision system 1000 includes a network interface controller (NIC) 1002 adapted to connect the decision system 1000 to a network 1006 via a bus 1004 that connects the decision system 1000 to one or more devices 1008. Examples of such devices include, but are not limited to, vehicles, traffic lights, traffic sensors, roadside units (RSUs), mobile edge computers (MECs), and passenger mobility devices. Furthermore, the decision system 1000 includes a transmitter interface 1010 that uses a transmitter 1012 and / or one or more devices 1008, configured to transmit a sequence and / or speed distribution and a planned stop sequence 1024, determined by one or more processors 1014, to each connected autonomous vehicle (CAV) along a future planned route in the traffic network. In each CAV, commands received from the global multi-vehicle decision system can be used by a multi-layered guidance and control architecture to control the movement of the vehicle in order to improve the overall safety, time and energy efficiency of traffic flow in the transportation network.

[0208] Using network 1006, the global multi-vehicle decision system 1000 receives real-time traffic data 1032 via receiver interface 1028 connected to receiver 1030. The decision system 1000 can receive traffic information from one or more interconnected conflict zones and road segments in the traffic network. Traffic data 1032 may include vehicle status (e.g., acceleration, position, heading, speed) and information on planned and / or predicted future routes (e.g., future road segments, trajectory lanes, desired destinations, sequences of waiting times) for each vehicle in the traffic network. Additionally or alternatively, the decision system 1000 may include a control interface 1034 configured to send commands to one or more devices 1008 to change their respective states (e.g., acceleration, speed, etc.). Control interface 1034 may use transmitter 1012 for sending commands and / or any other communication device.

[0209] In some embodiments of the present invention, a human-machine interface (HMI) 1040 connects the decision system 1000 to a keyboard 1036 and a pointing device 1038, wherein the pointing device 1038 may include a mouse, trackball, touchpad, joystick, pointing stick, stylus, or touchscreen, etc. The decision system 1000 may also be linked via a bus 1004 to a display interface adapted to connect the decision system 1000 to a display device (such as a computer monitor, camera, television, projector, or mobile device, etc.). The decision system 1000 may also be connected to an application interface adapted to connect the decision system 1000 to one or more devices for performing various power distribution tasks.

[0210] The decision system 1000 may include: one or more processors 1014 configured to execute stored instructions; and a memory (at least one memory) 1016 storing instructions executable by the processors (at least one processor) 1014. The processor 1014 may be a single-core processor, a multi-core processor, a computing cluster, a network of multiple connected processors, or any other configuration. The memory 1016 may include random access memory (RAM), read-only memory (ROM), flash memory, or any other suitable memory system. The processor 1014 may be connected to one or more input and output devices via a bus 1004. These instructions implement a method for making global multi-vehicle decisions in local areas of interconnected conflict zones. In some embodiments of the invention, the decision system 1000 includes a map configuration 1018. For example, the map configuration 1018 may include location data (e.g., GPS data) for conflict-free road segments, conflict zones, and trajectory lanes within each segment of the traffic network.

[0211] Decision system 1000 includes constraints and objectives 1020 of a mixed integer programming (MIP) problem 720 solved at each time step of a global multi-vehicle decision system, such as... Figures 8A to 8D As described in [the document]. For example, MIP constraints and objective 1020 can be configured to enforce physical constraints, vehicle speed limits, and / or safety constraints, and minimize a weighted combination of travel time, waiting time, and / or energy consumption for each vehicle in the traffic network. In some embodiments of the invention, MIP constraints and objective 1020 enable the generation of solutions to mixed-integer linear programming (MILP) or mixed-integer quadratic programming (MIQP) problems. Based on the solutions to the MIP, one or more processors 1014 can determine optimal sequences 1022 of entry and exit times and average speeds and / or speed distributions and planned stop sequences 1024 for each connected automated vehicle (CAV) along a future planned or predicted route in the traffic network.

[0212] Figure 11A A schematic diagram of a vehicle 1101 including a multi-layered guidance and control architecture 1102 according to some embodiments of the present invention is shown. This architecture 1102 controls the movement of the vehicle based on a future route plan and a corresponding motion plan that can be calculated by a global multi-vehicle decision system for a traffic network. As used herein, the vehicle 1101 can be any type of wheeled vehicle, such as a bus, truck, shuttle bus, public bus, or off-road vehicle. Furthermore, the vehicle 1101 can be an autonomous or semi-autonomous vehicle. For example, some embodiments control the movement of the vehicle 1101 based on a motion plan that can be calculated by a global multi-vehicle decision system. Examples of movement include lateral movement of the vehicle controlled by a steering system 1103 of the vehicle 1101. In some embodiments of the invention, the steering system 1103 is controlled by the multi-layered guidance and control architecture 1102. Additionally or alternatively, the steering system 1103 can be controlled by a (human) driver of the vehicle 1101.

[0213] The vehicle may also include an engine 1106, which may be directly controlled by the multi-layer guidance and control architecture 1102 or by other components of the vehicle 1101. The vehicle may also include one or more onboard sensors 1104 to sense the surrounding environment. Embodiments of sensor 1104 include rangefinders, radar, lidar, and cameras. The vehicle 1101 may also include one or more onboard sensors 1105 to sense its current motion and internal state. Embodiments of sensor 1105 include global positioning systems (GPS), accelerometers, inertial measurement units, gyroscopes, shaft rotation sensors, torque sensors, deflection sensors, pressure sensors, and flow sensors. The onboard sensors provide information to the multi-layer guidance and control architecture 1102. According to some embodiments of the invention, the vehicle may be equipped with a transceiver 1106 that enables the communication capabilities of the multi-layer guidance and control architecture 1102 via wired or wireless communication channels, for example, for communication between the vehicle 1101 and a global multi-vehicle decision system.

[0214] Figure 11B A schematic diagram illustrating the interaction between a multi-layered guidance and control architecture 1102 (i.e., an algorithmic and technical layer for decision-making, motion planning, vehicle control, and / or estimation) according to some embodiments of the present invention and other controllers 1120 of vehicle 1101. For example, in some embodiments, the controllers 1120 of vehicle 1101 are a steering controller 1125 and a brake / throttle controller 1130 that control the rotation and acceleration of vehicle 1120, respectively. In this case, the multi-layered guidance and control architecture 1102 outputs control inputs to controllers 1125 and 1130 to control the state of vehicle 1101. Controller 1120 may also include a higher-level controller (e.g., a lane-keeping assist controller 1135) that further processes the control inputs of the multi-layered guidance and control architecture 1102. In both cases, controller 1120 uses the outputs of the multi-layered guidance and control architecture 1102 to control at least one actuator of vehicle 1101 (such as the steering wheel and / or brakes of vehicle 1101) to control the movement of vehicle 1101. In some embodiments of the invention, vehicle 1101 is one of multiple CAVs in a traffic network, and multi-layer guidance and control architecture 1102 determines inputs to vehicle 1101 based on a motion plan calculated by a global multi-vehicle decision system, wherein inputs to vehicle 1101 may include one or a combination of vehicle 1101's acceleration, vehicle 1101's engine torque, braking torque, and steering angle.

[0215] Figure 12AA schematic diagram illustrating the interaction between a global multi-vehicle decision-making system 300 and a freight yard management system 1200 is shown. The global multi-vehicle decision-making system 300 and the freight yard management system 1200 can be used in combination for efficient job scheduling, motion planning, and decision-making of controlled and uncontrolled vehicles in a freight yard. In some embodiments of the invention, the freight yard management system calculates task assignments and plans routes for one or more trucks and / or tractors to perform one or more of the following tasks:

[0216] • Route planning toward one or more docks, for loading or unloading one or more trailers by one or more dockworkers;

[0217] • Route planning used to move one or more trailers from one location in a freight yard to another;

[0218] • Route planning toward specific parking spaces in the parking area for picking up or placing one or more trailers in the parking area;

[0219] • Route planning for the interconnected transportation network used to enter and leave the freight yard area.

[0220] In some embodiments of the invention, among other components, the freight yard management system may include one or more of the following components: reservation scheduling 1205, security gate management 1210, dock management 1215, and tractor dispatching 1220. The reservation scheduling 1205 system ensures that each truck scheduled to arrive at the freight yard for loading or unloading its cargo is reserved. The security gate management 1210 system may, with or without a driver, incorporate a security or verification process for each truck registration. The security gate management system ensures that both the freight yard management system 1200 and the global multi-vehicle decision system 300 are aware of all trucks and drivers currently present in the freight yard area. The dock management 1215 system is based on a task assignment or job scheduling system that determines which particular dock can be used for loading or unloading cargo in a particular trailer by one or more dockworkers. The tractor dispatching 1220 system determines whether a particular tractor should be instructed to acquire a trailer to move it to a specific parking space or a specific dock in the freight yard.

[0221] In some embodiments of the invention, task allocation, job scheduling, mapping, and / or navigation in a freight yard management system can be implemented using solutions to mixed integer programming (MIP) problems. The MIP for task allocation or job scheduling can be solved independently or in conjunction with an MIP for motion planning in a global multi-vehicle decision system, for example, using one or more mobile edge computers. In other embodiments of the invention, task allocation or job scheduling in a freight yard management system can be implemented based on heuristic search techniques to significantly reduce the computational cost of freight yard management for calculating feasible but potentially suboptimal schedules.

[0222] Figure 12B A schematic diagram of a freight yard management system 1250 is shown, which includes an interconnected traffic network 1255 for one or more vehicles to travel within a freight yard area between security gates 1265, one or more parking spaces 1260, and one or more docks 1270. The freight yard management system is designed to organize the movement of trucks within a manufacturing facility, warehouse, or distribution center. In some embodiments of the invention, each of the vehicles may be a autonomous or semi-autonomous vehicle (CAV) or an uncontrolled vehicle (NCV), and each of the vehicles may be a truck 1261 (with or without a trailer 1262), a tractor 1263, or a yard operator. The freight yard management system 1250 assigns each task to one or more of the truck 1261, trailer 1262, tractor 1263, or dock worker 1264, and one or more of the parking spaces 1260 or docks 1270. The task assignment results can be transmitted to one or more dockworkers 1264 and one or more of trucks 1261 and tractors 1263, and / or to one or more drivers of trucks 1261 and tractors 1263. For example, vehicle-to-everything (V2X) communication equipment and technology can be used in conjunction with electronic communication between the freight yard management system and a mobile device or computer, which may be held by one or more dockworkers and / or drivers. The global multi-vehicle decision system then calculates a rough movement plan for each of the CAVs and NCVs along its planned or predicted route within the interconnected transportation network to perform the task within the freight yard.

[0223] Some embodiments of the present invention are based on the understanding that a priority hierarchy may exist among vehicles due to the specific tasks assigned to each vehicle by the freight yard management system. In some embodiments of the present invention, the global multi-vehicle decision system computes a motion plan based on the solution of the MIP problem, which prioritizes one or more vehicles with higher priority (e.g., determined by the freight yard management system), wherein the MIP problem is based on weighted multi-objective minimization. For example, in the weighted multi-objective minimization of the MIP problem in the global multi-vehicle decision system, the weights of the travel time and / or waiting time of one or more vehicles with higher priority can be chosen to be relatively large.

[0224] The embodiments described above can be implemented in any of a variety of ways. For example, these embodiments can be implemented using hardware, software, or a combination thereof. When implemented in software, the software code can be executed on any suitable processor or set of processors, whether provided in a single computer or distributed across multiple computers. Such a processor can be implemented as an integrated circuit, wherein one or more processors are implemented within an integrated circuit component. However, the processor can be implemented using circuitry of any suitable format.

[0225] Furthermore, the various methods or processes outlined herein can be encoded as software that can execute on one or more processors employing any of a variety of operating systems or platforms. Additionally, such software can be written using a variety of suitable programming languages ​​and / or programming or scripting tools, and such software can also be compiled into executable machine language code or intermediate code that executes on a framework or virtual machine. Typically, the functionality of program modules can be combined or distributed in various implementations as needed.

[0226] Furthermore, embodiments of the present invention can be embodied as the methods of the provided embodiments. Actions performed as part of the method can be ordered in any suitable manner. Therefore, embodiments can be constructed that perform actions in a different order than those shown, which may include performing certain actions simultaneously, even if these actions are shown as sequential actions in the illustrative embodiments.

[0227] Although the invention has been described by way of preferred embodiments, it should be understood that various other adjustments and modifications can be made within the spirit and scope of the invention. Therefore, the appended claims are intended to cover all such changes and modifications within the true spirit and scope of the invention.

Claims

1. A global multi-vehicle decision-making system, the global multi-vehicle decision-making system being used to provide real-time motion planning and coordination for one or more connected autonomous and / or semi-autonomous vehicles (CAVs) in an interconnected transportation network, the interconnected transportation network including one or more uncontrolled vehicles (NCVs), one or more conflict zones, and one or more conflict-free road segments, the global multi-vehicle decision-making system comprising: The receiver is configured to acquire infrastructure sensing signals via a roadside unit (RSU) and acquire Type 1 feedback signals regarding the status of connected autonomous vehicles (CAVs) and planned future routes to one or more desired destinations, as well as Type 2 feedback signals regarding the status of uncontrolled vehicles (NCVs) and predicted future routes. The receiver is also configured to perform data communication with the freight yard management system, which includes a centralized task allocation, job scheduling, mapping and navigation system connected to the global multi-vehicle decision system to provide map information, long-term planned future route information for one or more CAVs and / or short-term predicted future route information for one or more NCVs. At least one memory, the at least one memory configured to store the map information and a computer-executable program, the computer-executable program including a global multi-vehicle decision program; as well as At least one processor, connected to the at least one memory, is configured to perform the following steps: A global mixed-integer programming (MIP) problem is formulated based on the infrastructure sensing signals, the type 1 feedback signals, the type 2 feedback signals, and the map information. The MIP problem performs weighted multi-objective minimization, which ensures a hierarchy of priorities among vehicles performing tasks within the interconnected transportation network of the freight yard by selecting relatively large weight values ​​for the travel time and / or waiting time of one or more vehicles with higher priorities. The motion plan for each CAV and each NCV in the interconnected transportation network is calculated by solving the global MIP problem. The motion plan is calculated for one or more trucks, guide vehicles or yard towing vehicles, each of which can be a controlled or uncontrolled vehicle along a route within the freight yard for efficient loading and unloading of goods. Calculate the optimal entry and exit time series and average speed series for each CAV and each NCV along each segment of the planned or predicted future route within the traffic network; Calculate the speed distribution and / or one or more planned stops for each CAV in the prediction time domain; and A transmitter configured to send the speed distribution to each CAV and / or send one or more planned stops to the multi-layer guidance and control architecture of each CAV in the interconnected transportation network.

2. The global multi-vehicle decision-making system according to claim 1, wherein, The uncontrolled vehicle (NCV) can include one or more human-driven vehicles and / or one or more semi-autonomous vehicles that cannot be controlled by the global multi-vehicle decision system.

3. The global multi-vehicle decision-making system according to claim 1, wherein, The conflict zone may include one or more merging points and / or traffic intersections, which may physically connect multiple lanes in multiple segments of the interconnected traffic network.

4. The global multi-vehicle decision-making system according to claim 1, wherein, The receiver uses vehicle-to-everything (V2X) communication to obtain the Type 1 feedback signal from one or more CAVs and / or one or more RSUs.

5. The global multi-vehicle decision-making system according to claim 1, wherein, The receiver uses vehicle-to-everything (V2X) communication to obtain the Type 2 feedback signal from one or more NCVs and / or one or more RSUs.

6. The global multi-vehicle decision-making system according to claim 1, wherein, The global multi-vehicle decision-making process is executed by one or more processors in one or more mobile edge computers (MECs) in a rolling time domain.

7. The global multi-vehicle decision-making system according to claim 1, wherein, The global multi-vehicle decision-making process is executed by one or more processors in a cloud-based computing system in a rolling time domain implementation.

8. The global multi-vehicle decision-making system according to claim 1, wherein, The centralized task allocation, mapping, and navigation system is connected to the global multi-vehicle decision system to provide the map information, long-term planned future route information for one or more CAVs, and / or short-term predicted future route information for one or more NCVs.

9. The global multi-vehicle decision-making system according to claim 1, wherein, Safety constraints regarding dynamic traffic participants, including bicycles and pedestrians, are enforced by the onboard modules for motion planning and / or vehicle control within the multi-layered guidance and control architecture of each CAV.

10. The global multi-vehicle decision-making system according to claim 1, wherein, The infrastructure sensing signals can include signals from one or more sensors, including rangefinders, radar, lidar, and / or cameras.

11. The global multi-vehicle decision-making system according to claim 1, wherein, The status information in the Type 1 feedback signal of the CAV and the status information in the Type 2 feedback signal of the NCV can include the current position, current angle of travel, and / or current speed of each vehicle.

12. The global multi-vehicle decision-making system according to claim 1, wherein, The map information includes location data of conflict-free road segments, location data of conflict zones including one or more merging points and / or traffic intersections, and location data of one or more track lanes within each road segment and each conflict zone of the traffic network.

13. The global multi-vehicle decision-making system according to claim 12, wherein, The planned future route information or the predicted future route information in the Type 1 feedback signal and the Type 2 feedback signal includes a sequence of future road segments and future trajectory lanes within those road segments, starting from the current road segment and trajectory lane of each vehicle, and the vehicle's direction toward the desired destination.

14. The global multi-vehicle decision-making system according to claim 1, wherein, The MIP problem is a mixed-integer linear programming (MILP) or mixed-integer quadratic programming (MIQP) problem, and the optimization variables include the entry and exit time variables, inverse average speed variables, and one or more binary optimization variables for each CAV and each NCV in each of the road segments and conflict zones of the traffic network.

15. The global multi-vehicle decision-making system according to claim 14, wherein, The MIP problem includes vehicle dynamics equality constraints for each of the CAV and / or NCV, as well as one or more speed limit constraints.

16. The global multi-vehicle decision-making system according to claim 14, wherein, The MIP problem includes vehicle routing inequality constraints to enforce the order between departure and entry segments in the planned or predicted future routes for each of the CAVs and NCVs.

17. The global multi-vehicle decision-making system according to claim 14, wherein, The MIP problem includes one or more safety constraints for each conflict zone in the traffic network to avoid potential collisions in the motion plan by prohibiting a pair of vehicles from traveling simultaneously on the conflict trajectory lane in the same conflict zone.

18. The global multi-vehicle decision-making system according to claim 17, wherein, The safety constraints are based on the Big M rule in a set of mixed integer linear inequalities of the MIP problem and are implemented using a binary optimization variable for each pair of vehicles and each conflict zone in the traffic network.

19. The global multi-vehicle decision-making system according to claim 14, wherein, The MIP problem includes one or more occupancy constraints for each segment in a traffic network to enforce a fixed order between any pair of vehicles that are planned or predicted to travel on the same trajectory lane within the same segment.

20. The global multi-vehicle decision-making system according to claim 19, wherein, The occupancy constraint is based on the Big M rule in a set of mixed integer linear inequality constraints of the MIP problem and is implemented using a binary optimization variable for each pair of vehicles and each road segment in the traffic network.

21. The global multi-vehicle decision-making system according to claim 14, wherein, The MIP problem is subjected to weighted multi-objective minimization, which includes a weighted sum of one or more objective terms to minimize a weighted combination of travel time, waiting time and / or energy consumption for each vehicle in the transportation network.

22. The global multi-vehicle decision-making system according to claim 1, wherein, The MIP problem can be solved using branch and bound algorithms, branch cutting algorithms, or branch pricing optimization algorithms.

23. The global multi-vehicle decision-making system according to claim 1, wherein, The exercise plan is calculated by using heuristic optimization algorithms to compute feasible but suboptimal solutions to the MIP problem and based on integer rounding schemes, feasibility pumps, approximate optimization algorithms, or by using machine learning techniques.

Images