Driving state determination method and device, storage medium and electronic equipment

Abstract

The invention discloses a driving state determination method and device, a storage medium and electronic equipment. The method comprises the steps that a candidate driving state set matched with a first driving state is determined, and a flexibility cost value is determined according to the number of candidate states of second driving states included in the candidate driving state set; according to the operation state of at least one road object included in the road environment matched with the first driving state, determining a collision risk cost value corresponding to the vehicle object at the second time node; determining a track cost value according to a driving track of the vehicle object switched from the current driving state to the first driving state; and under the condition that a weighted summation result among the flexibility cost value, the collision risk cost value and the trajectory cost value meets a target condition, determining a state node in the state simulation tree according to the first driving state. The technical problem that the decision making efficiency of a related method is low is solved.

Description

Technical Field

[0001] This application relates to the field of intelligent driving, and more specifically, to a method, apparatus, storage medium, and electronic device for determining driving status. Background Technology

[0002] With the continuous improvement of vehicle intelligence, high-level autonomous driving has become a core area of ​​competition in the global automotive industry. The decision-making and planning system, as the "brain" of autonomous driving, undertakes the crucial task of integrating perception, localization, and prediction information to generate safe, comfortable, and efficient driving trajectories. Especially in structured road environments, vehicles need to frequently handle complex interaction scenarios such as lane changes and merging, placing higher demands on the real-time response capability, safety reliability, and intelligent adaptability of the decision-making system.

[0003] In the process of generating motion trajectories, decision tree search is usually used to enumerate the potential motion states of the vehicle at various future time points. However, existing methods generally adopt a strategy of "building the complete tree first and then evaluating it uniformly." In this process, a large number of action branches are simulated and calculated throughout the entire process, resulting in wasted computing resources and low decision-making efficiency.

[0004] There is currently no widely accepted and effective solution to this problem. Summary of the Invention

[0005] This application provides a method, apparatus, storage medium, and electronic device for determining driving status, in order to at least solve the technical problem of low decision-making efficiency in related methods.

[0006] According to one aspect of the embodiments of this application, a method for determining a driving state is provided, comprising: determining a set of candidate driving states matching a first driving state, and determining a flexibility value based on the number of candidate states of a second driving state included in the set of candidate driving states, wherein the first driving state is used to indicate the driving state of a vehicle object at a first time node, and the set of candidate driving states includes a second driving state that allows the vehicle object to switch directly from the first driving state; determining a collision risk value corresponding to the vehicle object at a second time node based on the operating state of at least one road object included in the road environment matching the first driving state; determining a trajectory value based on the driving trajectory of the vehicle object switching from the current driving state to the first driving state; and determining state nodes in a state simulation tree based on the first driving state, wherein the state simulation tree includes multiple state nodes, and the state nodes describe the driving state of the vehicle object at at least one time node.

[0007] According to another aspect of the embodiments of this application, a driving state determination apparatus is also provided, comprising: a first determining unit, configured to determine a set of candidate driving states matching a first driving state, and to determine a flexibility cost based on the number of candidate states of a second driving state included in the set of candidate driving states, wherein the first driving state is used to indicate the driving state of a vehicle object at a first time node, and the set of candidate driving states includes a second driving state that allows the vehicle object to switch directly from the first driving state; a second determining unit, configured to determine a collision risk cost based on the operating state of at least one road object included in a road environment matching the first driving state at a second time node; a third determining unit, configured to determine a trajectory cost based on the driving trajectory of the vehicle object switching from the current driving state to the first driving state; and a fourth determining unit, configured to determine a state node in a state simulation tree based on the first driving state, provided that the weighted sum of the flexibility cost, the collision risk cost, and the trajectory cost satisfies a target condition, wherein the state simulation tree includes multiple state nodes, and the state nodes describe the driving state of the vehicle object at at least one time node.

[0008] According to another aspect of the embodiments of this application, a computer program product or computer program is provided, which includes computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform the method for determining the driving state as described above.

[0009] According to another aspect of the embodiments of this application, an electronic device is also provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the above-described method for determining the driving state through the computer program.

[0010] In this implementation, a set of candidate driving states matching the first driving state is first determined. The flexibility cost is then calculated based on the number of directly switchable second driving states within this set. A larger number of candidate states indicates a richer behavioral choice space for the vehicle at that moment, resulting in a lower flexibility cost, and vice versa. This quantifies the abstract concept of behavioral selectivity and introduces it into the evaluation system, prioritizing driving states with ample decision space and easier optimization in subsequent node selections. Subsequently, the collision risk cost is determined based on the operational status of various road objects in the road environment corresponding to the first driving state. The collision risk cost reflects the potential threat level of external traffic bodies to the vehicle in real time, thus embedding environmental risk into node evaluation in real time and avoiding meaningless branch expansion within dangerous areas. Furthermore, the trajectory cost is calculated using the actual trajectory generated by the vehicle switching from the previous moment to the first driving state.

[0011] Finally, the value of flexibility, the value of collision risk, and the value of trajectory are weighted and summed. Only when the sum meets the target condition is the corresponding first driving state confirmed as a valid state node in the state simulation tree. Through this triple cost fusion admission mechanism, a large number of high-risk, low-flexibility, or high-energy-consumption branches can be eliminated in the early stage of tree construction, significantly reducing the computational load of subsequent fine simulation. At the same time, it ensures that the remaining nodes have high quality in terms of behavioral accessibility, environmental safety, and motion feasibility, thereby improving the overall search efficiency of the state simulation tree and the actual executability of the generated trajectory, and solving the technical problems of low decision-making efficiency and redundant computation in related technologies. Attached Figure Description

[0012] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings:

[0013] Figure 1 This is a schematic diagram of an application environment for an optional method for determining driving status according to an embodiment of this application;

[0014] Figure 2 This is a flowchart of an optional method for determining a driving state according to an embodiment of this application;

[0015] Figure 3 This is a schematic diagram of an optional simulation process according to an embodiment of this application;

[0016] Figure 4 This is a schematic diagram of an optional driving state according to an embodiment of this application;

[0017] Figure 5This is a schematic diagram of an optional state machine according to an embodiment of this application;

[0018] Figure 6 This is a flowchart of another method for determining a driving state according to an embodiment of this application;

[0019] Figure 7 This is a schematic diagram of a device for determining a driving state according to an embodiment of this application;

[0020] Figure 8 This is a schematic diagram of the structure of an optional electronic device according to an embodiment of this application. Detailed Implementation

[0021] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort should fall within the scope of protection of the present application.

[0022] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0023] According to one aspect of the embodiments of this application, a method for determining a driving state is provided. This method for determining a driving state can be applied, but is not limited to, to applications such as... Figure 1 The hardware environment shown. Optionally, the method for determining the driving state provided in this application can be applied to a vehicle terminal. Figure 1 A side view of a vehicle terminal 101 is shown, which can travel on a driving surface 113. The vehicle terminal 101 includes a memory 102 storing an onboard navigation system 103 and a digital road map 104, a spatial monitoring system 117, a vehicle controller 109, a GPS (Global Positioning System) sensor 110, an HMI (Human / Machine Interface) device 111, and also includes an autonomous controller 112 and a telematics controller 114.

[0024] In one embodiment, the space monitoring system 117 includes one or more space sensors and systems for monitoring a visible area 105 in front of the vehicle terminal 101. The space monitoring system 117 also includes a space monitoring controller 118. The space sensors for monitoring the visible area 105 include a lidar sensor 106, a radar sensor 107, a camera 108, etc. The space monitoring controller 118 can be used to generate data related to the visible area 105 based on data input from the space sensors. The space monitoring controller 118 can determine the linear range, relative speed, and trajectory of the vehicle terminal 101 based on the input from the space sensors, for example, determining the vehicle's current speed and its relative speed to a vehicle in front. The space sensors of the vehicle terminal space monitoring system 117 may include object positioning sensing devices, which may include range sensors that can be used to locate objects in front, such as vehicles in front.

[0025] Camera 108 is advantageously mounted and positioned on vehicle terminal 101 in a location that allows for the capture of images of a visible area 105, wherein at least a portion of the visible area 105 includes the area in front of vehicle terminal 101 and a portion of the travel surface 113 of the trajectory of vehicle terminal 101. The visible area 105 may also include the surrounding environment. Other cameras may also be employed, for example, including a second camera positioned on the rear or side portion of vehicle terminal 101 to monitor the rear of vehicle terminal 101 and one of the right or left sides of vehicle terminal 101.

[0026] The autonomous controller 112 is configured to implement autonomous driving or advanced driver assistance system (ADAS) vehicle terminal functionality. Such functionality may include an onboard vehicle terminal control system capable of providing a certain level of driving automation. Driving automation may include a series of dynamic driving and vehicle terminal operations. Driving automation may include a certain level of automated control or intervention involving individual vehicle terminal functions (e.g., steering, acceleration, and / or braking). For example, the aforementioned autonomous controller may be used to determine a target pose point indicating the current position of the vehicle terminal by performing the following steps:

[0027] S102, determine a set of candidate driving states that match the first driving state, and determine the flexibility cost based on the number of candidate states of the second driving state included in the candidate driving state set.

[0028] The first driving state is used to indicate the driving state of the vehicle object at the first time node, and the candidate driving state set includes a second driving state that allows the vehicle object to switch directly from the first driving state.

[0029] S104, Based on the operating status of at least one road object included in the road environment matching the first driving state, determine the collision risk value of the vehicle object at the second time node;

[0030] The object profile is determined based on at least one driving operation of the vehicle object in the second driving cycle, and the second adjustment vector includes at least one second adjustment coefficient, which is used to perform a second adjustment operation on at least one corresponding acceleration state parameter.

[0031] S106, Determine the trajectory value based on the driving trajectory of the vehicle object when it switches from the current driving state to the first driving state;

[0032] The acceleration state parameter set includes a first distance parameter, a second distance parameter, a deceleration parameter, and a maximum acceleration parameter. The first distance parameter is the safe distance between the self-vehicle and the vehicle in front when they are stationary. The second distance parameter is the safe time distance between the self-vehicle and the vehicle in front when they are in motion.

[0033] S108, if the weighted sum of the value of flexibility, the value of collision risk and the value of trajectory satisfies the target condition, determine the state node in the state simulation tree according to the first driving state.

[0034] The state simulation tree includes multiple state nodes, which describe the driving state of the vehicle object at at least one time node.

[0035] HMI device 111 provides human-machine interaction for guiding the operation of infotainment systems, GPS (Global Positioning System) sensors 110, airborne navigation systems 103, and similar systems, and includes controllers. HMI device 111 monitors operator requests and provides the operator with status, service, and maintenance information about the vehicle terminal system. HMI device 111 communicates with and / or controls the operation of multiple operator interface devices. HMI device 111 may also communicate with one or more devices that monitor biometric data associated with the vehicle terminal operator. For simplicity, HMI device 111 is depicted as a single device, but in embodiments of the system described herein, it may be configured as multiple controllers and associated sensing devices.

[0036] Operator controls may be included in the passenger compartment of vehicle terminal 101 and, by way of non-limiting example, may include a steering wheel, accelerator pedal, brake pedal, and operator input device, which is an element of HMI device 111. The operator controls enable a vehicle terminal operator to interact with and instruct the operation of vehicle terminal 101 to provide passenger transport.

[0037] The airborne navigation system 103 uses a digital road map 104 for the purpose of providing navigation support and information to the vehicle terminal operator. The autonomous controller 112 uses the digital road map 104 for the purpose of controlling the operation of the autonomous vehicle terminal or the functions of the ADAS vehicle terminal.

[0038] The vehicle terminal 101 may include a telematics controller 114, which includes a wireless telematics communication system capable of communicating outside the vehicle terminal (including communicating with a communication network 115 with both wireless and wired communication capabilities). The wireless telematics communication system includes a non-airborne server 116 capable of short-range wireless communication with mobile terminals.

[0039] Through the above-described embodiments of this application, after fusing the first adjustment vector and the second adjustment vector, the acceleration state parameter set is synchronously corrected to the target acceleration state parameter set. The current acceleration is calculated using the current speed of the vehicle object and the target acceleration state parameter set. Since all key parameters have been dynamically updated, the determined current acceleration can maintain synchronization with traffic flow on a macroscopic level and conform to the vehicle's handling characteristics on a microscopic level, thus better matching the actual driving scenario and driving experience. This solves the technical problem of low matching degree between the determined acceleration and the actual scenario in related technologies.

[0040] According to one aspect of the embodiments of this application, a method for determining driving status is provided, such as... Figure 2 As shown, the method for determining the driving state can be as follows: Figure 1 The electronic devices included in the vehicle terminal shown execute the following specific steps:

[0041] S202, determine a set of candidate driving states that match the first driving state, and determine the flexibility cost based on the number of candidate states of the second driving state included in the candidate driving state set.

[0042] The first driving state is used to indicate the driving state of the vehicle object at the first time node, and the candidate driving state set includes a second driving state that allows the vehicle object to switch directly from the first driving state.

[0043] S204, Based on the operating state of at least one road object included in the road environment matching the first driving state, determine the collision risk value of the vehicle object at the second time node;

[0044] The object profile is determined based on at least one driving operation of the vehicle object in the second driving cycle, and the second adjustment vector includes at least one second adjustment coefficient, which is used to perform a second adjustment operation on at least one corresponding acceleration state parameter.

[0045] S206, Determine the trajectory value based on the driving trajectory of the vehicle object when it switches from the current driving state to the first driving state;

[0046] The acceleration state parameter set includes a first distance parameter, a second distance parameter, a deceleration parameter, and a maximum acceleration parameter. The first distance parameter is the safe distance between the self-vehicle and the vehicle in front when they are stationary. The second distance parameter is the safe time distance between the self-vehicle and the vehicle in front when they are in motion.

[0047] S208, under the condition that the weighted sum of the value of flexibility, the value of collision risk and the value of trajectory satisfies the target condition, the state node in the state simulation tree is determined according to the first driving state.

[0048] The state simulation tree includes multiple state nodes, which describe the driving state of the vehicle object at at least one time node.

[0049] In the above embodiments, the self-vehicle object can be the vehicle of interest in this embodiment, that is, the vehicle that needs to be located and driven. The other vehicle object can be any vehicle object in the road environment where the self-vehicle object is currently located, other than the self-vehicle object. In addition, the preceding vehicle object can be any vehicle object located in front of the self-vehicle object among the other vehicle objects.

[0050] It is understood that the first driving state and the second driving state in the above embodiments can be vehicle driving states corresponding to different time points. The vehicle driving state can include, but is not limited to, the lateral and longitudinal movement states of the vehicle. For example, the lateral movement states can include, but are not limited to, KeepLane state, lane change initiation state, lane change post-initiation state, and lane change abort state. The lane change state can be further divided into left change and right change. Similarly, the longitudinal movement states can include one or more of acceleration, deceleration, and constant speed states. In other words, in this embodiment, the vehicle movement state of the vehicle at different time points can be accurately described by combining the lateral and longitudinal movement states of the vehicle, such as by (starting left lane change state, acceleration).

[0051] It should be noted that the above-described embodiments of this application can be applied to the process of generating a state simulation tree for trajectory simulation. The current state node can be the state node currently traversed during the generation of the state simulation tree. Correspondingly, the state node determined in step S208 can be the first state node in the next level of the state simulation tree connected to the current state node, used to indicate that the vehicle object can switch from the current driving state corresponding to the current state node to the first driving state corresponding to the first state node.

[0052] It should be further noted that the second driving state included in the candidate driving state set that matches the first driving state in step S202 is a vehicle driving state that the vehicle object is allowed to switch from from the first driving state to.

[0053] For example, if the vehicle is in the late stage of a lane change, it is only allowed to switch from the late stage of a lane change to the KeepLane state. That is, the second driving state included in the candidate driving state set is only the KeepLane state, and the vehicle is no longer allowed to switch to the aborted lane change state, thereby preventing the vehicle from performing dangerous driving operations in the road environment.

[0054] When the vehicle is in the initial lane-changing state, it is only allowed to switch from the late lane-changing state to the late lane-changing state and the aborted lane-changing state. That is, the second driving state included in the candidate driving state set can simultaneously include the late lane-changing state and the aborted lane-changing state. In other words, when the vehicle is in the initial stage of lane changing, it is allowed to switch to the aborted lane-changing state, and it is also allowed to continue to perform subsequent lane-changing operations.

[0055] The following combination Figure 3 The relationship between the first and second driving states described above will be explained. For example... Figure 3 In the state simulation tree shown, each level of state node indicates the simulated driving state at a given time point. For example, the root node N0 indicates the current state at the current time point, the first level includes three state nodes indicating the simulated driving state 1 second later, the second level includes multiple state nodes indicating the simulated driving state 2 seconds later, and the third level includes multiple state nodes indicating the simulated driving state 3 seconds later.

[0056] Assuming the current node determined by the simulation is the LeftChange node in the first layer, the current driving state can be determined as the vehicle driving state described by the LeftChange node. Then, the state nodes connected to the LeftChange node in the second layer can be determined using the above implementation method. Specifically, if it is examined whether a state node corresponding to the (LeftChange, A1) state is generated, then (LeftChange, A1) can be determined as the first driving state corresponding to the first time node, and a candidate driving state set corresponding to (LeftChange, A1) can be determined, along with the corresponding flexibility value.

[0057] In addition, as in step S204 above, besides determining the flexibility value, it is also necessary to further determine the collision risk value corresponding to the second time node, where the second time node can be a time node after the first time node.

[0058] Furthermore, as in step S206, in this embodiment of the application, the trajectory value corresponding to the trajectory generated during the transition from the current driving state to the first driving state (LeftChange, A1) will be further examined.

[0059] Furthermore, in this application embodiment, when the three types of values ​​corresponding to the first driving state (LeftChange, A1) are determined through the above-described implementation method, it is possible to determine whether to generate the corresponding state node based on the weighted sum of the above-described flexibility value, collision risk value, and trajectory value.

[0060] In one optional implementation, after determining the trajectory value based on the driving trajectory of the vehicle object switching from the current driving state to the first driving state, the method further includes:

[0061] Method 1: If the weighted sum of the value of flexibility, the value of collision risk, and the value of trajectory is less than or equal to the value threshold, the first driving state is determined to meet the target condition.

[0062] Method 2: When the weighted sum of the value of flexibility, the value of collision risk, and the value of trajectory is greater than the value threshold, a reference driving state is obtained from the reference driving state set as the next vehicle driving state to be simulated. The reference driving state set includes a fourth driving state that allows the vehicle to switch directly from the third driving state. The third driving state is the vehicle driving state corresponding to the previous time node.

[0063] It is understandable that, in the above implementation, a cost-value determination mechanism can be further introduced to dynamically determine whether the first driving state is qualified to continue to expand, thereby eliminating inferior branches in the early stage of tree building and improving search efficiency.

[0064] Specifically, the cost of flexibility, the cost of collision risk, and the cost of trajectory can be weighted and summed to obtain a comprehensive cost. If this comprehensive cost is less than or equal to a preset cost threshold, it indicates that the first driving state meets the requirements in terms of behavioral accessibility, environmental safety, and motion smoothness, and can be retained as a valid node in the state simulation tree, and can continue to expand downwards.

[0065] Conversely, if the overall cost exceeds the cost threshold, it indicates that the state either has a narrow range of selectable behaviors, significantly conflicts with surrounding traffic bodies, or lacks sufficient smoothness in its trajectory. In this case, the system will no longer continue to expand in that direction, but will instead select a new reference driving state from the set of reference driving states as the next node to be simulated.

[0066] In one alternative implementation, the aforementioned cost threshold can be preset and dynamically adjusted according to the complexity of the scenario. For example, a higher cost threshold can be set for high-speed scenarios and a lower cost threshold can be set for urban scenarios.

[0067] In another optional implementation, the aforementioned cost threshold can also be determined based on the sorting results of the comprehensive cost values ​​corresponding to multiple fourth driving states in the reference driving state set. For example, the comprehensive cost value of the fourth driving state ranked kth can be determined as the cost threshold after sorting the comprehensive cost values ​​from smallest to largest, thereby controlling the inclusion of a limited number of fourth driving states in the subsequent expansion process of the state simulation tree and avoiding excessive growth of the state simulation tree.

[0068] Continue with Figure 3 For example, when the current node is the LeftChange node in the first layer, the comprehensive value corresponding to each of the three fourth driving states (LeftChange, A1), (LeftChange, A2), and (LeftChange, A3) included in the reference driving state set can be determined. Among them, the comprehensive value corresponding to (LeftChange, A3) is higher than the above-mentioned value threshold, so a pruning operation is performed, and the (LeftChange, A3) node is no longer generated, nor is it expanded further based on (LeftChange, A3).

[0069] Through the above-described implementation method of this application, the value of flexibility, the value of collision risk, and the value of trajectory are weighted and summed. Only when the sum meets the target condition is the corresponding first driving state confirmed as a valid state node in the state simulation tree. This triple-cost fusion admission mechanism eliminates a large number of high-risk, low-flexibility, or high-energy-consumption branches early in the tree construction process, significantly reducing the computational load of subsequent detailed simulations. Simultaneously, it ensures that the remaining nodes possess high quality in terms of behavioral reachability, environmental safety, and motion feasibility, thereby improving the overall search efficiency of the state simulation tree and the actual executability of the generated trajectory. This solves the technical problems of low decision-making efficiency and excessive redundant computation in related technologies.

[0070] The following provides a further explanation of the methods for determining the three cost values ​​in the above embodiments.

[0071] In the process of determining the flexibility cost value, the aforementioned determination of the candidate driving state set matching the first driving state can be achieved by determining multiple time nodes based on a fixed time interval, and then determining the candidate driving state set corresponding to each of the multiple time nodes.

[0072] In another optional implementation, the aforementioned time interval may also be related to the current road scenario, including:

[0073] S1, obtain the target time interval that matches the road environment, where multiple environment identifiers each correspond to a time interval;

[0074] S2, based on the first time node and the target time interval, determine the second time node after the first time node;

[0075] S3, determine a candidate driving state set based on at least one second driving state that the vehicle object is allowed to switch to at the second time node.

[0076] It is understandable that this implementation further refines the process of determining the flexibility cost value by introducing a target time interval that dynamically matches the road environment, thereby giving the flexibility cost value a more physically meaningful measurement in different traffic scenarios.

[0077] Specifically, the first step is to obtain the target time interval corresponding to the current road environment. This interval can be obtained through a preset mapping table or an online recognition module: for example, in high-speed scenarios, due to high vehicle speeds and short reaction windows, the target time interval is determined to be 0.2 seconds; in general urban roads, due to moderate vehicle speeds and complex traffic flow, the interval is expanded to 0.7 seconds; and in congested urban roads, due to frequent vehicle starts and stops and high behavioral uncertainty, the interval is further relaxed to 1.0 second. Each time interval corresponds to an environmental identifier, forming multiple configurations such as high-speed -0.2 seconds, urban -0.7 seconds, and congested -1.0 seconds, ensuring that subsequent evaluations scale synchronously with the level of external risk.

[0078] Subsequently, the target time interval is superimposed on the first time node, i.e., the time corresponding to the current node to be evaluated, to obtain the second time node. Optionally, the second time node can be the next discrete layer immediately following the decision tree, or it can be a future slice on the physical time axis, with its position dynamically floating according to the road type, thereby breaking through the fixed step size limit between layers.

[0079] At the second time point, combining state machine constraints, road geometry, and traffic regulations, all second driving states that the vehicle is allowed to switch to directly are enumerated, forming a candidate driving state set. The number of actions in this set serves as the core input for the flexibility cost: the larger the number, the richer the behavioral choices at the target time, the higher the system flexibility, and the lower the corresponding cost; conversely, the lower the cost, the higher the cost, indicating insufficient scalability in that direction, and it should be retained with caution.

[0080] Through the above-described embodiments of this application, by coupling the time interval used to determine the candidate driving state set with the environment, the flexibility evaluation focuses on short-term rapid response in high-speed scenarios and long-term prediction in congested scenarios. This avoids the evaluation distortion caused by fixed step size and keeps the scale of candidate actions consistent with the real behavior space, thereby significantly improving the search efficiency and trajectory rationality of the state simulation tree in complex scenarios.

[0081] Furthermore, the flexibility value is determined based on the number of candidate states of the second driving state included in the candidate driving state set, including one of the following:

[0082] Method 1: The product of the number of candidate states and the flexibility weight is determined as the flexibility cost.

[0083] Method 2: When at least one reference driving state set is determined based on the candidate driving state set, the flexibility cost is determined based on the weighted sum between the number of candidate states and the number of at least one reference state. Here, the first reference driving state in the at least one reference driving state set is obtained by simulation deduction based on the second driving state, and the i-th reference driving state in the i-th reference driving state set in the at least one reference driving state set is obtained by simulation deduction based on the (i-1)-th reference driving state.

[0084] In the first method described above, the number of states in the candidate driving state set can be directly multiplied by a preset flexibility weight to obtain the flexibility value. For example, the flexibility value... It can be determined in the following way:

[0085]

[0086] in, To enumerate the number of all legal lateral movements in the next time step, starting from the current node state, according to the lane change state machine rules. For example, a node in KeepLane mode... (KeepLane, LeftChange, RightChange can continue). A node in the LeftChange state that has entered the later stages of a lane change... (LeftChange can only continue). Therefore, KeepLane nodes will receive a greater flexibility reward than lane-change nodes in the later stages of a lane change.

[0087] In another alternative implementation, the determination of the flexibility cost is no longer limited to single-point counting, but incorporates both short-term and medium-to-long-term behavioral spaces into the assessment through a hierarchical deduction mechanism, thereby more accurately depicting the vehicle's true decision-making flexibility in the current state.

[0088] To further improve evaluation accuracy, a multi-layered reference driving state set is introduced. Specifically, the candidate driving state set can be used as the first layer, representing the possible states for the next step. Then, forward simulation is performed on each state in the first layer to deduce the second layer of reference driving state sets; and so on, generating the third, fourth, and even more layers of reference state sets. The number of states in each layer participates in the flexibility calculation, but its corresponding weight monotonically decreases with increasing layer level. For example, an exponential decay or linear discount strategy is used to ensure that more distant branches contribute less to the current flexibility, preventing excessive amplification of long-term uncertainty.

[0089] In this application's implementation, the value of flexibility It can be determined in the following way:

[0090]

[0091] in, It can be determined in the following way:

[0092]

[0093] in, , These represent the number of states in the i-th layer reference driving state set. This is the preset maximum number of simulation layers.

[0094] In the implementation method described above, the flexibility cost is determined by the weighted sum of the number of candidate states and the number of reference states at each layer, with the weight decreasing along the hierarchy. This measures not only how many actions are possible at the next moment, but also how many compliant paths these actions can evolve into in subsequent steps. A richer path pool indicates greater behavioral redundancy at the current decision point, higher system flexibility, and a lower corresponding cost. Conversely, if a dead end or a single path appears in the early layers, even with a considerable number of states at the first layer, the long-term weighted result will still converge rapidly, increasing the cost and indicating limited scalability in that direction.

[0095] By using the above-mentioned hierarchical weighting method, it is possible to identify two typical cases, namely wide entrance-narrow channel or narrow entrance-wide channel, in the early stage of tree construction. This avoids wasting computing resources on subtrees with poor long-term reachability, thereby significantly improving the search efficiency of the state simulation tree and the rationality of the generated trajectory.

[0096] The following section provides further explanation of the method for determining the collision risk cost value.

[0097] In one optional implementation, determining the collision risk value of the vehicle object at a first time point based on the operating state of at least one road object included in the road environment matching the first driving state includes:

[0098] S1, determine the current road object from at least one road object;

[0099] S2, obtain the rate of change of the estimated collision time between the vehicle object and the current road object;

[0100] S2-1, when the time change rate is greater than 0, the negative value of the time change rate is used as the risk cost of the object matched with the current road object;

[0101] S2-2, when the rate of change over time is less than or equal to 0, 0 is taken as the risk value of the object matched with the current road object;

[0102] S3, given that the risk cost of at least one road object is determined, the collision risk cost is determined based on the weighted sum of the risk cost values ​​of at least one object.

[0103] It is understood that, in the embodiments of this application, the determination of the collision risk cost value no longer depends on a single instantaneous collision time value, but focuses on the risk change trend at a certain future moment, so as to capture the interaction situation that is about to deteriorate in advance.

[0104] Specifically, in the current set of road objects, one or more current road objects are selected as the focus of risk analysis based on the filtering strategy of being closest to the vehicle longitudinally, having the highest relative speed, or having the lowest TTC.

[0105] Next, for each current road object, the first derivative of the estimated collision time TTC with respect to time, i.e., the rate of change of time, is calculated. This rate of change reflects the speed at which risk intensifies or mitigates—if If positive, TTC is increasing, and the interaction is becoming more secure; if... The TTC is negative, indicating that the risk is accumulating rapidly.

[0106] Furthermore, when the rate of change is greater than 0, it indicates that the danger is mitigating and the object poses no further threat to the vehicle; therefore, its risk cost is set to 0 to avoid unnecessary conservative costs. Conversely, when the rate of change is less than or equal to 0, it indicates that the danger remains unchanged or is intensifying; in this case, the rate of change is set to negative. As a risk-value object;

[0107] Understandably, if multiple road objects exist simultaneously, the risk cost of each object is weighted and summed. The weights can be dynamically allocated based on longitudinal distance, relative speed, or collision probability. The final output is the collision risk cost, which is used for subsequent node cost sorting and pruning.

[0108] In this application's implementation, the collision risk cost It can be determined in the following way:

[0109]

[0110] In the above formula, Let TTC be the rate of change of the vehicle and the j-th obstacle over time. As TTC decreases, the risk increases, so this term is positive and incurs a cost.

[0111] Through the above-described implementation methods of this application, the approaching danger trend can be captured before the TTC falls below the safety threshold, and a high cost can be assigned in advance. This guides the decision tree to prioritize the behavioral branches that cause the TTC curve to rise or at least stop falling, thereby achieving trend-level risk prediction and improving trajectory safety and foresight.

[0112] In the above implementation, two future cost functions are introduced. and It focuses on decision-making flexibility and situational quality. It assesses the magnitude of "future possibilities" opened up by the current state. The flexibility cost can be quantified based on the vehicle's current state and behavioral state machine, calculating the number of all executable subsequent lateral actions starting from that node. This cost rewards nodes that retain more future options for the system. The collision risk cost is used to comprehensively assess the future interaction environment with surrounding traffic participants caused by the current node state, including overall safety pressures predicted based on relative kinematics.

[0113] The following provides a further explanation of the method for determining the trajectory value in step S206.

[0114] In one optional implementation, the trajectory value is determined based on the driving trajectory of the vehicle as it switches from the current driving state to the first driving state, including at least one of the following:

[0115] Method 1: Obtain the lane offset cost value corresponding to each of the multiple trajectory points in the driving trajectory; determine the navigation offset cost value matching the driving trajectory based on the weighted sum of the multiple lane offset cost values. The trajectory cost value is determined based on the navigation offset cost value, and the lane offset cost value of the trajectory point is determined based on the ratio between the trajectory point offset and the reference offset. The trajectory point offset is the distance between the trajectory point position and the centerline of the target lane.

[0116] In the first determination method described above, the navigation offset cost is used to measure the lateral deviation of the vehicle from the centerline of the target lane, ensuring that the vehicle deviates from the lane as little as possible. For a single trajectory point, the corresponding lane offset cost can be... The navigation offset cost, determined by the weighted sum of multiple lane offset cost values, can be determined as follows:

[0117]

[0118] in, Let be the lateral distance between the i-th trajectory point and the center of the lane. The maximum allowable offset, The navigation cost weights used in determining the trajectory cost based on the navigation offset cost.

[0119] Method 2: Determine the speed cost based on the ratio between the average speed of the trajectory corresponding to the driving trajectory and the reference speed; determine the distance cost based on the ratio between the first distance indicated by the trajectory endpoint and the reference distance; determine the traffic efficiency cost of the driving trajectory matching based on the weighted sum of the speed cost and the distance cost, wherein the trajectory cost is determined based on the traffic efficiency cost.

[0120] In the second determination method described above, the traffic efficiency cost is used to evaluate the difference between the vehicle's average speed and the desired speed during the simulation, preventing vehicles from "dawdling" and encouraging efficient traffic flow. The first cost, speed, penalizes low speeds, while the second cost, distance, directly rewards traveling further along the road. The speed cost can be expressed as... The distance cost can be expressed as Therefore, the cost of traffic efficiency can be determined in the following way:

[0121]

[0122] in, For the target speed, The average velocity of the trajectory points . Let S be the endpoint in the Frent (SL) coordinate system of the current node simulation. This represents the theoretical longitudinal displacement that can be achieved within the current simulation duration.

[0123] Method 3: Obtain the acceleration and rate of change of acceleration corresponding to multiple trajectory points in the driving trajectory, and obtain the maximum trajectory curvature; determine the acceleration cost based on the acceleration corresponding to each of the multiple trajectory points; determine the acceleration change cost based on the rate of change of acceleration corresponding to each of the multiple trajectory points; determine the comfort cost based on the weighted sum of the acceleration cost, the acceleration change cost, and the maximum trajectory curvature, wherein the trajectory cost is determined based on the comfort cost.

[0124] In the second determination method described above, the comfort cost is used to ensure the smoothness of acceleration and jerk (rate of change of acceleration, jerk), making the generated trajectory smooth and stable, avoiding sudden acceleration, sudden braking, and sharp turns. The acceleration cost can be expressed as... The cost of acceleration change can be expressed as The maximum trajectory curvature can be expressed as Therefore, the cost of comfort can be determined in the following way:

[0125]

[0126] in, Let be the curvature corresponding to the i-th trajectory point. The maximum absolute curvature in the entire trajectory;

[0127] Method 4: Obtain the trajectory curvature and the rate of change of trajectory curvature for each of the multiple trajectory points in the driving trajectory; determine the curvature cost based on the trajectory curvature for each of the multiple trajectory points; determine the rate of change of curvature based on the rate of change of trajectory curvature for each of the multiple trajectory points; determine the curvature continuity cost based on the weighted sum of the curvature cost and the rate of change of trajectory curvature, wherein the trajectory cost is determined based on the curvature continuity cost.

[0128] In the fourth determination method described above, the curvature continuity cost is used to penalize excessively large or drastic changes in trajectory curvature, thereby generating a dynamically feasible trajectory as much as possible. The curvature cost can be expressed as... The rate of change of curvature can be expressed as Therefore, the cost of curvature continuity can be determined in the following way:

[0129]

[0130] Method 5: Obtain the collision time cost and relative distance cost corresponding to each of the multiple trajectory points in the driving trajectory; based on the exponential decay summation results of the multiple collision time cost values ​​and the multiple relative distance cost values, determine the safety cost matching the driving trajectory. Among them, the trajectory cost is determined based on the safety cost, the collision time cost is determined based on the exponential decay value between the estimated collision time of the trajectory point and the collision time safety threshold, and the relative distance cost is determined based on the exponential decay value between the relative distance of the trajectory point and the distance safety threshold.

[0131] In the fifth method described above, the safety cost is determined based on the minimum distance to surrounding vehicles and the time to collision (TTC). An exponential function is used to convert collision risk and distance into cost. When the TTC is small or the distance is very close, the cost increases sharply, creating a "hazard field" that strongly penalizes any unsafe behavior. This can be determined as follows:

[0132]

[0133] Where N is the number of trajectory points and M is the number of targets. For the TTC of the i-th trajectory point and the j-th obstacle, For TTC security threshold, Let be the relative distance between the i-th trajectory point and the j-th obstacle. The safe distance threshold between targets.

[0134] In one optional implementation, the trajectory cost can be determined by a combination of one or more of the above five determination methods; in a preferred implementation, when the trajectory cost is determined by a combination of the above five determination methods, the trajectory cost... It can be determined in the following way:

[0135]

[0136] Through the above-described implementation methods of this application, the rationality of the trajectory can be combined to accurately identify different vehicle states in advance. Based on the current state, a lightweight evaluation mechanism can quickly and accurately predict the long-term benefits of the subtree, thereby achieving efficient and intelligent pruning in the early stages of decision-making.

[0137] In an optional implementation, before obtaining a reference driving state from the set of reference driving states as the next vehicle driving state to be simulated, the method further includes:

[0138] S1, Obtain the target state machine diagram, which includes multiple state nodes. The state nodes are used to indicate a vehicle driving state, and the directed connections between the state nodes are used to indicate that state switching is allowed between two vehicle driving states.

[0139] S2, based on the current state node in the target state machine diagram that matches the third driving state, determine at least one target state node that the current state node points to;

[0140] S3, determine the reference driving state set based on at least one target state node.

[0141] In the embodiments of this application, such as Figure 4 As shown, the driving status may include, for example: Figure 4 The default state (lane keeping, stage 0) shown in Figure (a) is as follows: Figure 4 The first lane change state (stage 1) shown in Figure (b) is as follows: Figure 4 The second lane change state (stage 2) shown in diagram (c) is as follows: Figure 4 The abortion state shown in diagram (d) is as follows: Figure 4 The third lane change state (stage3) is shown in Figure (e), and the trajectory decision is constrained by the state machine.

[0142] In the above embodiments of this application, the rules of the five-state lane-changing state machine can be deeply embedded into each step of the decision tree expansion process. Specifically, when the trajectory decision tree expands to any node, the current state of the state machine is first queried, and a set of legal lateral and longitudinal actions is generated according to the mapping relationship shown in Table 1, instead of enumerating all possible actions. Subsequently, forward state deduction is performed only for these legal actions, and the state of the state machine is updated synchronously. This mechanism ensures from the source of decision sequence generation that all candidate behaviors comply with driving logic and safety regulations. For example, when the state machine is in the third lane-changing state (stage 3), the system will prohibit the generation of the "terminate lane change" action, thereby avoiding dangerous disjointed behavior of the vehicle after it has crossed the lane line.

[0143] Table 1

[0144]

[0145]

[0146] The transitions between different states in the above five-state machine can be achieved through methods such as... Figure 5 The description is as shown.

[0147] Optionally, this application proposes an interaction mechanism for state machine and decision tree node expansion, aiming to achieve efficient joint search of horizontal and vertical actions with limited computing resources. This mechanism operates in each planning cycle and mainly includes four stages: initialization, node selection and expansion loop, forward simulation and intelligent pruning, and loop termination and optimal output. The specific process is as follows:

[0148] During the initialization phase, upon receiving the planning instruction for a new cycle, the state machine either resets or inherits the final state of the previous cycle to form the current state. At the same time, the decision tree establishes its root node. It stores the vehicle's current pose, speed, acceleration, and environmental perception information, serving as the starting point for the entire search tree.

[0149] The system then enters a node selection and expansion loop. This loop continues until the decision tree reaches a preset maximum depth or the computation time exceeds the budget. In each loop, the system first selects a node N_current to be expanded from the leaf node list; then it queries the state machine to obtain the current state S_m corresponding to N_current; based on this state, the action generation module filters out a subset A_legal of legal actions that meet the state constraints from the complete horizontal and vertical action space A_all. Here, A_all covers Cartesian combinations of horizontal {keeping lane, changing lanes left, changing lanes right, aborting lane change} and vertical {accelerating, decelerating, maintaining speed}. Filtering rules include avoiding two consecutive lane changes, prohibiting exceeding road boundaries, and speed limit constraints, ensuring that the generated A_legal is both rich and safe.

[0150] After obtaining A_legal, the system performs an adaptive intelligent IDM forward simulation based on multi-dimensional perception for each legal action pair: using the end state of N_current as the initial value, it runs along the candidate action trajectory in simulation steps of 0.5-1.5s, updating the relative position, speed, and acceleration of the vehicle and surrounding vehicles in real time, and simultaneously calculating the costs of navigation deviation, traffic efficiency, comfort, curvature continuity, and safety. After the simulation is completed, a lightweight heuristic function is used to quickly score all newly generated leaf nodes, sort them from low to high cost, and only the Top-K nodes are retained for the next round of expansion, while the remaining nodes are immediately pruned, thus keeping the search space within an acceptable range.

[0151] When the loop condition is no longer met, the system terminates the expansion and initiates a refined forward simulation on the few remaining candidate trajectories: recalculating the dynamics and costs using a smaller time step to obtain a more accurate total cost; then, the trajectory with the minimum total cost is selected as the optimal trajectory, and its acceleration sequence and lane-changing command are output to the longitudinal and lateral controllers for execution. Simultaneously, the state machine state S_m_final corresponding to the end of this optimal trajectory is recorded and used as the initial state for the next planning cycle, achieving a smooth transition between cycles. Through this interactive mechanism, the decision tree, guided by state machine constraints, completes a rapid search of the joint longitudinal and lateral space, ensuring both real-time performance and improved trajectory quality.

[0152] It is understandable that, in the process of determining the state simulation tree through the above implementation method, the vehicle driving state corresponding to different trajectory points in the trajectory can be determined in the following way.

[0153] In an optional implementation, before determining the trajectory value based on the driving trajectory of the vehicle object switching from the current driving state to the first driving state, the method further includes:

[0154] Given the starting point of the driving trajectory and the first driving state, the vehicle acceleration corresponding to multiple timestamps is determined sequentially according to the target step size. The vehicle acceleration is determined based on the target acceleration state set and the previous vehicle speed. The target acceleration state set is obtained by adjusting at least one acceleration state parameter in the acceleration state parameter set according to a first adjustment vector matching the current road environment and a second adjustment vector matching the object profile of the vehicle object.

[0155] Furthermore, the above-mentioned determination of the vehicle acceleration corresponding to multiple timestamps according to the target step size includes:

[0156] S1, obtain a first adjustment vector, wherein the current road environment is determined based on at least one driving state of at least one other vehicle object in the first driving cycle, and the first adjustment vector includes at least one first adjustment coefficient, and the at least one first adjustment coefficient is used to perform a first adjustment operation on at least one corresponding acceleration state parameter;

[0157] S2, obtain the second adjustment vector, wherein the object profile is determined based on at least one driving operation of the vehicle object in the second driving cycle, and the second adjustment vector includes at least one second adjustment coefficient, which is used to perform a second adjustment operation on at least one corresponding acceleration state parameter;

[0158] S3, adjust at least one acceleration state parameter in the acceleration state parameter set according to the first adjustment vector and the second adjustment vector to obtain the target acceleration state parameter set, wherein the acceleration state parameter set includes a first distance parameter, a second distance parameter, a deceleration parameter, and a maximum acceleration parameter. The first distance parameter is the safe distance between the self-vehicle object and the preceding vehicle object in a stationary state, and the second distance parameter is the safe time distance between the self-vehicle object and the preceding vehicle object in a moving state.

[0159] S4. Based on the vehicle speed and target acceleration state parameter set corresponding to the previous timestamp, determine the vehicle acceleration corresponding to the current timestamp.

[0160] The following further explains the method for determining the acceleration corresponding to each trajectory point during trajectory simulation. In the above implementation, the first distance parameter included in the acceleration state parameters (which can be specifically expressed as...) ), the second distance parameter (which can be specifically expressed as ), deceleration parameters (which can be specifically expressed as ), maximum acceleration parameter (which can be specifically expressed as It can be used to determine real-time acceleration.

[0161] Among them, the maximum acceleration parameter It can be used to characterize the upper limit of a vehicle's driving capability. On the one hand, when the maximum acceleration parameter... Increasing the maximum acceleration parameter allows for a corresponding increase in the acceleration margin, enabling the vehicle to approach the target speed more quickly; conversely, decreasing the maximum acceleration parameter... This can reduce acceleration power, resulting in a more conservative driving style.

[0162] deceleration parameters It can be used to control the comfort of the braking process. When the deceleration parameter... When the setting is larger, the absolute value of deceleration is smaller when deceleration is triggered, resulting in a smoother ride. When the deceleration parameter... When the setting is smaller, the vehicle can decelerate later and more abruptly, and the peak negative acceleration will increase accordingly, which can cope with emergency situations, but the comfort will decrease.

[0163] First distance parameter It can be used to determine the minimum parking distance. A larger value allows for a greater following distance in low-speed or congested crawling scenarios. Decrease the first distance parameter. This allows for closer parking distances and increases road capacity, but places higher demands on sensor error and execution latency.

[0164] Second distance parameter It can be used to determine the following distance during driving. Increasing the second distance parameter... This means proportionally scaling up the desired distance at any speed, causing the vehicle to exhibit "long-distance" driving behavior, which is beneficial for safety and energy conservation. The second distance parameter is reduced. This reduces following distance, but may increase the risk of collision and lead to more frequent acceleration and deceleration.

[0165] Furthermore, the adjustment method described in the above embodiments may involve first performing a first adjustment operation on at least one acceleration state parameter based on a first adjustment vector, and then performing a second adjustment operation on at least one acceleration state parameter obtained after the first adjustment operation based on a second adjustment vector; alternatively, it may involve first performing a second adjustment operation on at least one acceleration state parameter based on a second adjustment vector, and then performing a first adjustment operation on at least one acceleration state parameter obtained after the second adjustment operation based on a first adjustment vector; optionally, it may involve performing a first adjustment operation and a second adjustment operation on at least one acceleration state parameter based on a first adjustment vector and a second adjustment vector respectively, and then performing aggregation processing on the operation results of the first adjustment operation and the second adjustment operation to obtain a target acceleration state parameter set, and determining the current acceleration based on the target acceleration state parameter set and the current speed of the vehicle object.

[0166] In a preferred embodiment, a first adjustment operation can be performed on at least one acceleration state parameter based on a first adjustment vector, and then a second adjustment operation can be performed on at least one acceleration state parameter obtained after the first adjustment operation based on a second adjustment vector.

[0167] In the embodiments of this application, during the determination of the acceleration of each trajectory point in the trajectory simulation process, the acceleration corresponding to each trajectory point can be determined based on the above-mentioned acceleration state parameters in the following manner.

[0168]

[0169] in, The acceleration of the current trajectory point. The current speed of the vehicle object. For the desired speed of the vehicle, The acceleration index is used to adjust the smoothness of acceleration behavior. This represents the actual headway between the vehicle and the vehicle in front. Furthermore, in the above formula... This can be further determined in the following ways:

[0170]

[0171] in, The preceding vehicle's acceleration term allows the vehicle to respond to the preceding vehicle's acceleration or deceleration earlier.

[0172] It should be noted that the above method is used to determine Previously, the first and second adjustment operations could be performed on each acceleration state parameter using the aforementioned implementation method.

[0173] The following explains the first adjustment operation.

[0174] In this embodiment of the application, the congestion description parameter can be expressed as: This can be used to quantify the congestion status of the current road environment. In one alternative approach, the aforementioned congestion description parameters... It can be directly derived from V2X messages received in real time at the vehicle (such as the congestion level field in SPAT / MAP), or it can be determined by calculating the average time distance between vehicles by the vehicle's sensors within a specific length spatial window.

[0175] The average speed of a lane can be described as It is used to reflect the longitudinal movement intensity of the current lane's vehicle group. It can be obtained by performing a "time-space" weighted average of the speed sequences of other vehicles output by forward radar or cameras, or by directly referencing the real-time traffic speeds provided by navigation maps.

[0176] Road type parameters can be described as This can be output in real time from a high-precision map or vehicle positioning unit, and its enumerated values ​​can include at least two of the following road types: "highway arterial road", "urban expressway", "urban ordinary road", "ramp", "auxiliary road", and "elevated road". By introducing... This ensures that subsequent adjustment coefficients maintain a baseline that matches the road design speed and legal speed limit under the same traffic conditions, avoiding abnormal combinations such as "high-speed close distance" or "low-speed long distance".

[0177] Next, using the aforementioned three-dimensional parameters as a joint index, nearest neighbor matching can be performed in a pre-defined pattern table to obtain the current pattern identifier. This pattern table can be generated offline from historical natural driving data clustering or obtained through traffic simulation calibration. Each pattern identifier corresponds to a pre-defined adjustment vector, which contains distance adjustment coefficients, time-distance adjustment coefficients, deceleration adjustment coefficients, and acceleration adjustment coefficients, all of which act on the first distance parameter. Second distance parameter deceleration parameters Maximum acceleration parameters .

[0178] In one alternative implementation, the current mode identifier can be determined in the following way:

[0179]

[0180] The current mode identifier is determined through the above implementation method. Given the specific values, the corresponding first adjustment vector can be determined by looking up the values ​​in Table 2 below. :

[0181] Table 2

[0182]

[0183] For example, in determining through the above implementation method In this case, the first adjustment vector can be determined accordingly. .

[0184] Furthermore, after determining the first adjustment vector In this case, the first adjustment method mentioned above can be implemented in the following way:

[0185]

[0186]

[0187]

[0188]

[0189] in, These are the first distance parameter, the second distance parameter, the deceleration parameter, and the maximum acceleration parameter before the first adjustment operation, respectively.

[0190] These are the first distance parameter, the second distance parameter, the deceleration parameter, and the maximum acceleration parameter after the first adjustment operation.

[0191] The second adjustment operation is further explained below. First, the second adjustment vector can be determined as follows:

[0192] S1, determine a driving description vector based on at least one operation data sequence matched with the second driving cycle, wherein the operation data sequence includes operation quantities corresponding to multiple timestamps respectively, and at least one description parameter included in the driving description vector is determined based on at least one operation data sequence respectively;

[0193] S2, Based on the similarity between the driving description vector and multiple image description vectors, determine the target image description vector from the multiple image description vectors;

[0194] S3, the adjustment vector that matches the target profile description vector is determined as the second adjustment vector, wherein the second adjustment vector includes an aggression adjustment coefficient, a patience adjustment coefficient, an economic adjustment coefficient, and a risk propensity adjustment coefficient.

[0195] In the above embodiments, at least one operational data sequence can be used to indicate multiple driving decision behaviors of the current driving object within the first driving cycle; the at least one operational data sequence includes parameters and data reflecting the driving behavior of the current driving object, such as lane change frequency, braking / acceleration frequency, average absolute acceleration / deceleration, average / standard deviation of acceleration, standard deviation of lane centering error, following distance, and TTC longitudinal speed with the preceding vehicle, etc., and the driving data of the current driving object can be collected in real time by vehicle sensors (such as cameras, radar, ABS systems, etc.).

[0196] Furthermore, a driving description vector for describing the object profile can be determined based on at least one of the above-mentioned operational data sequences, and a target profile description vector can be determined based on the driving description vector.

[0197] In one alternative implementation, a large amount of actual driving data of the driver over several weeks can be collected first. The data includes the vehicle's speed, acceleration, jerk, headway to the vehicle in front, relative speed, collision time, accelerator pedal opening, brake pedal opening, and steering angular velocity. This raw data can be continuously recorded at a frequency of 10 Hz and aggregated and stored via the vehicle or the cloud.

[0198] Subsequently, the original time series data of each driving object is segmented, for example, by using a sliding window slice with a window length of 120 seconds and a step size of 60 seconds. Multiple statistical features are calculated in each window, including average headway, acceleration standard deviation, maximum braking deceleration, 5th percentile of collision time, frequency of rapid acceleration and deceleration, and proportion of high-speed driving time, thereby constructing a driving description vector that can reflect the characteristics of individual driving behavior.

[0199] After the driving description vectors are constructed, unsupervised learning algorithms such as K-Means are used to cluster the driving description vectors of all driving objects.

[0200] In this embodiment, driving subjects can be automatically divided into several groups, such as aggressive, conservative, economical, and normal types. Aggressive driving subjects typically have shorter headway, larger acceleration fluctuations, and higher frequency of abrupt changes; conservative driving subjects exhibit longer headway, lower acceleration changes, and milder braking behavior; economical driving subjects show a high ratio of throttle opening to acceleration and a low proportion of high-speed driving; normal types fall between the above types. Each of these categories can be represented using its corresponding cluster center vector, that is, the cluster center of each cluster is used as the profile description vector.

[0201] Furthermore, clustering results can be evaluated using metrics such as silhouette coefficient to ensure that the groups have significant discriminative power and stability.

[0202] In this embodiment of the application, based on the clustering results, a second adjustment vector can be further constructed for adjusting the acceleration state parameters. The second adjustment vector includes an aggressive adjustment coefficient. Patience adjustment coefficient Risk propensity adjustment coefficient and economic adjustment coefficient .

[0203] Aggression adjustment coefficient This can be reflected by the normalized values ​​of the acceleration standard deviation and the jerk frequency; patience adjustment coefficient. It can be measured by the negative correlation between average headway and collision time quantiles; risk propensity adjustment coefficient. The assessment can be based on a combination of maximum braking deceleration and the proportion of short-term collision events; economic adjustment coefficient. This is quantified using throttle efficiency and the percentage of high-speed duration. After normalization, each dimension forms a second adjustment vector used to adjust the acceleration state parameters. .

[0204] After determining the aforementioned second adjustment vector The following section further explains the second adjustment operation that adjusts at least one acceleration state parameter in the acceleration state parameter set.

[0205] First, the first reference distance parameter can be adjusted. This first reference distance parameter is obtained by adjusting the first distance parameter according to at least one first adjustment coefficient.

[0206] Specifically, adjustments can be made in the following ways:

[0207]

[0208] in, The patience adjustment coefficient is the second reference distance parameter obtained after the first adjustment operation. The risk propensity adjustment coefficient is used to characterize driver patience and is obtained from the second adjustment vector. Its value range is [0,1], with 0 representing extreme impatience and 1 representing extreme patience; The risk tolerance of the driver is represented by the second adjustment vector. The value range is [0,1], where the endpoint 0 represents risk aversion (very cautious) and the endpoint 1 represents risk preference (willing to take risks). and Risk propensity adjustment coefficient Patience adjustment coefficient Each has its corresponding weighting coefficient.

[0209] Next, the second reference distance parameter can be adjusted. The second reference distance parameter is obtained by adjusting the second distance parameter according to at least one first adjustment coefficient.

[0210] Specifically, adjustments can be made in the following ways:

[0211]

[0212] in, The patience adjustment coefficient is the second reference distance parameter obtained after the first adjustment operation. Used to characterize driver patience; aggression adjustment coefficient The value used to characterize the driver's aggression is obtained from the second adjustment vector. The value range is [0,1], with the endpoint 0 representing extremely mild and the endpoint 1 representing extremely aggressive. and Adjust the patience level coefficient and the aggression adjustment coefficient Each has its corresponding weighting coefficient.

[0213] Furthermore, the reference deceleration parameter can be adjusted. The reference deceleration parameter is obtained by adjusting the deceleration parameter according to at least one first adjustment coefficient.

[0214] Specifically, adjustments can be made in the following ways:

[0215]

[0216] in The reference deceleration parameter obtained after the first adjustment operation, and the risk propensity adjustment coefficient. Used to characterize the driver's risk tolerance; economic adjustment coefficient It is used to characterize the driver's economic preference and is obtained from the second adjustment vector. The value range is [0,1], where the endpoint 0 indicates limited performance and the endpoint 1 indicates economic priority. and Risk propensity adjustment coefficient and economic adjustment coefficient Each has its corresponding weighting coefficient.

[0217] Finally, the reference acceleration parameters can be adjusted. These reference acceleration parameters are obtained by adjusting the maximum acceleration parameters according to at least one first adjustment coefficient.

[0218] Specifically, adjustments can be made in the following ways:

[0219]

[0220] in The reference acceleration parameter obtained after the first adjustment operation, and the aggressive adjustment coefficient. Used to characterize driver aggression; economic adjustment coefficient Used to characterize a driver's economical tendencies. and Adjustment coefficient for aggression and economic adjustment coefficient Each has its corresponding weighting coefficient.

[0221] In conventional methods of determining acceleration, the relevant acceleration state parameters are fixed, which deviates significantly from actual traffic observations and fails to reflect the differences in driving styles such as aggressive, conservative, and economical driving. Furthermore, in existing simulation technologies, the behavior of all vehicles in traffic flow tends to be similar, failing to simulate the diversity of driving behaviors in the real world. Finally, existing technologies cannot meet the personalized preferences and comfort needs of different users in autonomous driving applications.

[0222] Through the above-described implementation method of this application, a four-dimensional driver profile vector (aggression, patience, risk tolerance, and economic preference) is established. Personalized features are mapped to the spatial transformation of IDM core parameters using a linear weighting function. Based on personalized coefficients calibrated from actual driving data, customized adjustments to parameters such as safe distance, acceleration characteristics, and braking behavior are achieved. This solves the problem of behavioral uniformity caused by neglecting individual differences in existing acceleration determination methods, significantly improving the realism of traffic simulation.

[0223] After adjusting the acceleration state parameters through the above implementation method, the acceleration corresponding to each trajectory point can be determined in the following specific way:

[0224]

[0225] in, This represents the vehicle's current speed. For the desired speed of the vehicle, The acceleration index is used to adjust the smoothness of acceleration behavior. This represents the actual headway between your vehicle and the vehicle in front.

[0226] In the above formula The acceleration state parameter set after the first and second adjustment operations can be determined in the following way:

[0227]

[0228] By further introducing a dynamic compensation term for the acceleration of the preceding vehicle through the above-described embodiments of this application, the vehicle can not only respond to changes in relative speed but also anticipate the acceleration and deceleration trends of the preceding vehicle. By calculating the influence coefficient of the preceding vehicle's acceleration on the safe distance in real time and dynamically adjusting the expected following distance, the collision risk prediction is elevated from the speed level to the acceleration level, fundamentally solving the problem of acceleration response delay.

[0229] It is understood that the acceleration determined in the above embodiments of this application can be used in the trajectory planning process during autonomous driving. In the above embodiments of this application, the driving state corresponding to different state nodes in the predicted trajectory can be determined at a certain period, and then the vehicle motion state corresponding to multiple simulation sub-nodes in different state nodes can be determined based on the state cost.

[0230] The following combination Figure 6 A complete process of this application is described.

[0231] S602, the decision tree begins to expand; marking the official start of the search process for this planning cycle.

[0232] S604, Initialize the root node N_root of the decision tree; establish the root node based on the current vehicle state and environmental perception information, which serves as the starting point of the entire tree.

[0233] S606, Select the node to be expanded, N_Current; Select a node to be expanded from the existing leaf nodes according to a strategy (such as lowest cost priority).

[0234] S608, query the state machine to obtain the current state S_M; read the lateral state (lane, lane change phase, etc.) and longitudinal state (speed, acceleration, etc.) of the state machine at time N_Current for subsequent action legality judgment.

[0235] S610, generate a candidate action set, and based on S_M, fit the action subset A_Legal; filter out legal horizontal-vertical action combinations from the complete action space according to the current state machine constraints (such as not being able to change lanes twice consecutively, not being able to cross the road boundary), and form A_legal.

[0236] S612 extends the current N_Current node based on A_Legal; generates new child nodes for each action in A_legal, preparing for parallel simulation.

[0237] S614: Obtain the initial state of the current N_Current node, including coordinates, heading angle, velocity, and acceleration; extract the end pose and kinematics of the parent node as initial values ​​for the forward simulation of the child node.

[0238] S616 Sorts all targets by proximity to the moving object using ObjListSort; sorts surrounding vehicles in the perception list by their longitudinal distance from the vehicle to the nearest, ensuring that the most relevant traffic bodies are processed first.

[0239] S618, sequentially retrieve the targets in ObjListSort; cyclically retrieve the vehicles according to the sorting results, and incorporate them into the interactive simulation one by one.

[0240] S620 simulates forward dynamics model simulation; based on the candidate actions of the self-vehicle and the motion assumptions of the target vehicle, it runs a bicycle or double integral dynamics model to estimate the initial trajectory draft for the next few seconds.

[0241] S622, a multi-dimensional perception adaptive intelligent IDM vehicle-following simulation; based on dynamic trajectory, it adopts an improved IDM to calculate following acceleration, and integrates multi-target distance, relative speed, congestion description parameters and driver profile in real time to output a more realistic longitudinal acceleration correction value.

[0242] S624, update the state of the next simulation point, coordinates / heading angle / velocity / acceleration; using the acceleration obtained in the previous step, advance the vehicle's pose and kinematic state in 0.15s steps and write it into the new node.

[0243] S626: Has the forward simulation of all targets in ObjListSort been completed? If yes, proceed to S628: Check if the interactive simulation of all neighboring target vehicles has been completed. If not, return to S618 to continue to the next target.

[0244] S628 checks if each node (15 steps, step size 0.15) has completed the current N_Current node; the maximum number of simulation steps per node is 15 steps (approximately 2.25 seconds). If the completion condition is not met even after reaching the required number of steps, the branch is forcibly terminated, and S614 is executed to prevent infinite expansion. If completed, S630 is executed.

[0245] S630 determines the actual cost of the simulated trajectory; based on the completed trajectory, it calculates the actual costs in multiple dimensions such as navigation deviation, traffic efficiency, comfort, and safety, which are then used for subsequent ranking.

[0246] S632, determine the heuristic future cost; use a lightweight heuristic function (such as remaining distance × reciprocal of expected velocity) to quickly estimate the potential cost from the end of the trajectory to the target, and add it to the actual cost to form the total score.

[0247] S634, determine whether the current N_Current node has been expanded; prevent repeated expansion; if the node has already completed the generation of child nodes, skip it.

[0248] S636 evaluates the cost of each action trajectory based on the fusion cost and selects the TOP-K as the effective nodes for the next cycle; it sorts all child nodes in ascending order of total cost, retains only the K with the best cost to enter the next round of expansion, and prunes the rest to ensure that computing resources are controllable.

[0249] S638 determines whether the decision tree has reached its maximum depth or whether the computation time has exceeded the threshold. If the tree depth or CPU time reaches the preset upper limit, the loop terminates and enters the final stage; otherwise, S606 is executed.

[0250] S640, decision tree expansion complete; output the retained candidate trajectory set for final fine simulation and optimal trajectory selection, and record the corresponding state machine state as the initial value for the next cycle. This concludes the planning cycle expansion process.

[0251] The following combination Figure 3 Another specific embodiment of this application will be described.

[0252] This implementation provides a forward simulation example based on a five-layer decision tree to explain the complete execution process of the joint horizontal and vertical search in a typical high-speed scenario. The decision tree is set to five layers with a 1-second time interval between layers, for a total simulation duration of 4 seconds. Within each 1-second interval, fine-grained forward inference is performed in 0.1-second increments. The horizontal action space includes {KeepLane, LeftChange, RightChange}, and the vertical action space includes {A1, A2, A3}, corresponding to three types of vertical control intensities: "mild acceleration / deceleration," "moderate acceleration / deceleration," and "strong acceleration / deceleration," respectively.

[0253] The initial state of the scenario is as follows: The driver is traveling at a constant speed of 100 km / h in the middle lane. Directly in front of the driver is a slow-moving target vehicle traveling at 90 km / h. To the left rear of the driver is another vehicle rapidly approaching at a speed significantly higher than the driver's. In terms of road geometry, the left lane line is a dashed line, and the right lane line is a solid line; therefore, the state machine prohibits any attempt to change lanes to the right.

[0254] The algorithm execution flow is as follows:

[0255] S1, initialize the root node N0;

[0256] Upon receiving the new planning cycle instruction, a root node N0 is created, which internally records the vehicle's current pose, velocity, and a list of environmental targets, such as... Figure 3 As shown, this serves as the starting point for the entire decision tree.

[0257] S2, first-level extension and state machine filtering;

[0258] When performing a combined horizontal and vertical expansion on N0, the state machine reports the current horizontal state as KeepLane, allowing horizontal actions {KeepLane, LeftChange, RightChange}. Since the right side is a solid line, the state machine immediately marks RightChange as invalid; the three right lane change branches generated as a result are prematurely removed, leaving only the KeepLane and LeftChange branches. Vertically, each of the two branches combines with {A1, A2, A3} to form six first-level child nodes: {KeepLane, A1}, {KeepLane, A2}, {KeepLane, A3}, {LeftChange, A1}, {LeftChange, A2}, and {LeftChange, A3}.

[0259] S3, First layer forward simulation and pruning;

[0260] A 1-second forward simulation was run for each of the six nodes. Due to the continuous pressure from the slower vehicle ahead at 90 km / h, the Time-to-Collision (TTC) of the KeepLane branch rapidly decreased towards the end of the simulation, causing a sharp increase in collision risk. Based on this, the heuristic evaluation module marked {KeepLane,A1}, {KeepLane,A2}, and {KeepLane,A3} as high-risk and pruned them. After the first layer expansion was completed, only the three nodes {LeftChange,A1}, {LeftChange,A2}, and {LeftChange,A3} were retained for the next round.

[0261] S4, second-level extension;

[0262] According to the state machine, the current horizontal state is in the LeftChange phase. Based on state machine constraints, the horizontal actions of child nodes can only continue with LeftChange. Vertically, they continue to combine with {A1, A2, A3}, theoretically forming a total of 9 second-level child nodes (for simplicity, the diagram is incomplete). Figure 3 (Only 3 are shown in the image). Each node advances by 1 second in 0.1-second increments to generate the corresponding trajectory.

[0263] S5, second-layer simulation and secondary pruning;

[0264] In the second-level simulation results, {LeftChange,A3} exhibited excessive longitudinal acceleration, resulting in a significantly higher lateral speed for the vehicle compared to the oncoming vehicle approaching from the left rear. This created a noticeable speed difference, causing a sharp decrease in the following vehicle's TTC (Traffic Time Control), and the overall safety cost exceeded the threshold. The system then pruned this node. The remaining nodes were evaluated and retained.

[0265] S6, progressive expansion from the third to the fifth layer;

[0266] Following the same logic, the algorithm sequentially completes node expansion, forward simulation, and cost evaluation for the third, fourth, and fifth layers. Each layer maintains the LeftChange state under state machine constraints, with the longitudinal direction optimized within the range {A1, A2}. As the number of layers increases, the vehicle gradually completes lateral displacement and establishes a stable following relationship with the vehicle in front in the left lane, balancing the costs of TTC, comfort, and traffic efficiency. Finally, the decision tree outputs several complete trajectories at the end of the fifth layer. The system selects the trajectory with the lowest total cost as the optimal trajectory and sends its corresponding acceleration sequence and lane-changing command to the execution layer.

[0267] Through the aforementioned five-layer, 4-second, 50-step forward deduction and layer-by-layer pruning, the algorithm automatically eliminates high-risk right lane changes and excessive longitudinal maneuvers under multiple constraints, including solid line constraints, slow vehicle suppression, and rapid approach from following vehicles. It prioritizes retaining smooth trajectories of "left lane change + moderate acceleration / deceleration," achieving a comprehensive optimization of safety, comfort, and traffic efficiency. It should be noted that this process is completed entirely within a single cycle of 100 milliseconds. The pruning mechanism reduces the node size from the theoretical 243 nodes to less than 20, significantly reducing the computational load.

[0268] The above-described implementation of this application provides a refined five-state lane-changing state machine, whose rules are deeply embedded in the expansion process of the decision tree. The state machine strictly constrains the generation logic of lateral actions at nodes, fundamentally solving the problems of behavioral sequence rationality defects and safety risks, and ensuring that all candidate behaviors conform to driving logic.

[0269] Furthermore, forward simulation of the actions between nodes is performed. The simulation results evaluate the costs already incurred by the current node, while also assessing the potential costs of future nodes from the perspectives of decision-making flexibility and situational quality. The heuristic function intelligently guesses "how much more cost it will take to reach the destination from here" based on the current node's state. This avoids pruning potential optimal solutions due to "short-sighted" decisions, making the pruning process more forward-looking. It can more accurately identify nodes with potentially higher short-term costs but significant long-term potential, significantly improving the overall quality of the final decision scheme (safer, more efficient, and more reasonable), especially in complex merging and lane-changing interaction scenarios. Before each node expansion, the Top-K nodes with the lowest overall cost are retained for subsequent expansion, reducing computational complexity from exponential to linear while maintaining decision-making quality.

[0270] Furthermore, a multi-level adaptive following state determination algorithm that integrates scene recognition, forward vehicle state perception, and driver profile is proposed. Specifically, it introduces a forward vehicle acceleration dynamic compensation term, constructs a hierarchical road scene recognition and parameter mapping mechanism, and establishes a four-dimensional driver profile vector (aggression, patience, risk tolerance, and economic tendency), thereby significantly improving the realism of traffic simulation.

[0271] Furthermore, in the forward simulation stage of the decision tree, an innovative vehicle dynamics model incorporating tire slip angle, understeer gradient, and powertrain delay is introduced. By calculating the centroid slip angle and velocity-related curvature, the actual motion posture of the vehicle under medium-to-high speed and large lateral acceleration conditions is accurately simulated. More accurate simulation trajectories are used for cost evaluation, making pruning decisions more reliable. This ensures from the outset that the final planned trajectory conforms to vehicle dynamics constraints, avoiding the risk of "simulation feasible, but not executable in real vehicles," thus improving system safety. This overcomes the problem of simulation trajectories becoming disconnected from real vehicle dynamics caused by the use of simplified kinematic models in traditional decision planning, improving the accuracy of forward simulation predictions.

[0272] In the above implementation, a combination of hard constraints and soft pruning is proposed. The state machine first ensures the safety and rationality of the behavior, and heuristic cost pruning performs soft pruning. On this basis, efficient and optimal solutions are further selected. Finally, fine simulation and optimal solution output are completed from a very small number of candidate trajectories, achieving a balance between computational efficiency and decision quality.

[0273] It should be noted that, for the sake of simplicity, the foregoing method embodiments are all described as a series of actions. However, those skilled in the art should understand that this application is not limited to the described order of actions, as some steps may be performed in other orders or simultaneously according to this application. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are preferred embodiments, and the actions and modules involved are not necessarily essential to this application.

[0274] According to another aspect of the embodiments of this application, a device for determining a driving state for implementing the above-described method for determining a driving state is also provided. For example... Figure 7 As shown, the device includes:

[0275] The first determining unit 702 is used to determine a set of candidate driving states that match the first driving state, and to determine a flexibility cost based on the number of candidate states of the second driving state included in the set of candidate driving states. The first driving state is used to indicate the driving state of the vehicle object at the first time node, and the set of candidate driving states includes a second driving state that allows the vehicle object to switch directly from the first driving state.

[0276] The second determining unit 704 is used to determine the collision risk value of the vehicle object at the second time node based on the operating state of at least one road object included in the road environment that matches the first driving state.

[0277] The third determining unit 706 is used to determine the trajectory value based on the driving trajectory of the above-mentioned vehicle object when it switches from the current driving state to the first driving state.

[0278] The fourth determining unit 708 is used to determine a state node in the state simulation tree based on the first driving state, provided that the weighted sum of the above-mentioned flexibility value, the above-mentioned collision risk value and the above-mentioned trajectory value satisfies the target condition. The state simulation tree includes multiple state nodes, and the state nodes describe the driving state of the vehicle object at at least one time node.

[0279] Optionally, the first determining unit 702 is configured to: obtain a target time interval matching the road environment, wherein each of the plurality of environment identifiers corresponds to a time interval; determine the second time node after the first time node based on the first time node and the target time interval; and determine the candidate driving state set based on at least one of the second driving states that the vehicle object is allowed to switch to at the second time node.

[0280] Optionally, the first determining unit 702 is used to: determine the product between the number of candidate states and the flexibility weight as the flexibility cost; and, in the case where at least one reference driving state set is determined based on the candidate driving state set, determine the flexibility cost based on the weighted sum between the number of candidate states and the number of at least one reference state, wherein the first reference driving state in the first reference driving state set of at least one of the reference driving state sets is obtained by simulation deduction based on the second driving state, and the i-th reference driving state in the i-th reference driving state set of at least one of the reference driving state sets is obtained by simulation deduction based on the (i-1)-th reference driving state.

[0281] Optionally, the second determining unit 704 is configured to: determine the current road object from at least one of the road objects; obtain the time change rate of the estimated collision time between the vehicle object and the current road object; if the time change rate is greater than 0, use the negative value of the time change rate as the object risk cost value matched with the current road object; if the time change rate is less than or equal to 0, use 0 as the object risk cost value matched with the current road object; and if the object risk cost value corresponding to each of the at least one road object is determined, determine the collision risk cost value based on the weighted summation result of the at least one object risk cost value.

[0282] Optionally, the third determining unit 706 is configured to: obtain lane offset cost values ​​corresponding to multiple trajectory points in the driving trajectory; determine a navigation offset cost value matching the driving trajectory based on a weighted sum of the multiple lane offset cost values, wherein the trajectory cost value is determined based on the navigation offset cost value, the lane offset cost value of the trajectory point is determined based on the ratio between the trajectory point offset and the reference offset, and the trajectory point offset is the distance between the trajectory point position and the lane centerline of the target lane; and determine the lane offset cost value based on the ratio between the average trajectory speed corresponding to the driving trajectory and the reference speed. The following steps are taken: First, determine the speed cost; second, determine the distance cost based on the ratio between the first distance indicated by the endpoint of the trajectory and the reference distance; third, determine the traffic efficiency cost of the matched trajectory based on the weighted sum of the speed cost and the distance cost, wherein the trajectory cost is determined based on the traffic efficiency cost; fourth, obtain the acceleration and rate of change of acceleration corresponding to each of the multiple trajectory points in the trajectory, and obtain the maximum trajectory curvature; fifth, determine the acceleration cost based on the acceleration corresponding to each of the multiple trajectory points; and sixth, determine the traffic efficiency cost based on the rate of change of acceleration corresponding to each of the multiple trajectory points. Acceleration change cost; based on the above acceleration cost, the weighted sum of the above acceleration change cost and the above maximum trajectory curvature, determine the comfort cost, wherein the above trajectory cost is determined based on the above comfort cost; obtain the trajectory curvature and trajectory curvature change rate corresponding to each of the multiple trajectory points in the above driving trajectory; based on the above trajectory curvature corresponding to each of the multiple trajectory points, determine the curvature cost; based on the above trajectory curvature change rate corresponding to each of the multiple trajectory points, determine the curvature change rate; based on the weighted sum of the above curvature cost and the above trajectory curvature change rate, determine the curvature continuity cost, wherein the above trajectory The cost value is determined based on the curvature continuity cost value mentioned above; the collision time cost value and relative distance cost value corresponding to each of the multiple trajectory points in the driving trajectory are obtained; based on the exponential decay summation result of the multiple collision time cost values ​​and the exponential decay summation result of the multiple relative distance cost values, the safety cost value matching the driving trajectory is determined, wherein the trajectory cost value is determined based on the safety cost value mentioned above, the collision time cost value is determined based on the exponential decay value between the estimated collision time of the trajectory point and the collision time safety threshold, and the relative distance cost value is determined based on the exponential decay value between the relative distance of the trajectory point and the distance safety threshold.

[0283] Optionally, the above-mentioned driving state determination device further includes: a decision unit, configured to determine that the first driving state satisfies the above-mentioned target condition when the weighted sum of the above-mentioned flexibility cost, the above-mentioned collision risk cost, and the above-mentioned trajectory cost is less than or equal to a cost threshold; and to obtain a reference driving state from a set of reference driving states as the next vehicle driving state to be simulated when the weighted sum of the above-mentioned flexibility cost, the above-mentioned collision risk cost, and the above-mentioned trajectory cost is greater than the above-mentioned cost threshold, wherein the set of reference driving states includes a fourth driving state that allows the vehicle object to directly switch from the above-mentioned third driving state, and the above-mentioned third driving state is the vehicle driving state corresponding to the previous time node.

[0284] Optionally, the decision-making unit is further configured to: acquire a target state machine diagram, wherein the target state machine diagram includes multiple state nodes, the state nodes are used to indicate a vehicle driving state, and the directed connections between the state nodes are used to indicate that a state switch is allowed between two vehicle driving states; determine at least one target state node pointed to by the current state node in the target state machine diagram that matches the third driving state; and determine the reference driving state set based on at least one of the target state nodes.

[0285] Optionally, the above-mentioned driving state determination device further includes: an acceleration determination unit, used to determine the vehicle acceleration corresponding to multiple timestamps according to a target step size when the trajectory starting point of the driving trajectory and the first driving state are determined; wherein the vehicle acceleration is determined based on a target acceleration state set and the previous vehicle speed, and the target acceleration state set is obtained by adjusting at least one acceleration state parameter in the acceleration state parameter set according to a first adjustment vector matching the current road environment and a second adjustment vector matching the object profile of the vehicle object.

[0286] Optionally, the acceleration determination unit is configured to: obtain the first adjustment vector, wherein the current road environment is determined based on at least one driving state of at least one other vehicle object in a first driving cycle, the first adjustment vector includes at least one first adjustment coefficient, and the at least one first adjustment coefficient is used to perform a first adjustment operation on at least one corresponding acceleration state parameter; and obtain the second adjustment vector, wherein the object profile is determined based on at least one driving operation of the own vehicle object in a second driving cycle, the second adjustment vector includes at least one second adjustment coefficient, and the at least one second adjustment coefficient is used to adjust the at least one corresponding acceleration state parameter. The parameters undergo a second adjustment operation; based on the first adjustment vector and the second adjustment vector, at least one acceleration state parameter in the acceleration state parameter set is adjusted to obtain the target acceleration state parameter set, wherein the acceleration state parameter set includes a first distance parameter, a second distance parameter, a deceleration parameter, and a maximum acceleration parameter; the first distance parameter is the safe distance between the self-vehicle and the preceding vehicle when stationary, and the second distance parameter is the safe time distance between the self-vehicle and the preceding vehicle when in motion; based on the self-vehicle speed corresponding to the previous timestamp and the target acceleration state parameter set, the self-vehicle acceleration corresponding to the current timestamp is determined.

[0287] For specific implementation examples, please refer to the examples shown in the above method for determining the driving state, which will not be repeated here.

[0288] This application also provides an electronic device, such as... Figure 8 As shown, the memory 802 can be used to store software programs and modules, such as the program instructions / modules corresponding to the method and apparatus for determining the driving state in this embodiment of the invention. The processor 804 executes various functional applications and data processing by running the software programs and modules stored in the memory 802, thereby realizing the aforementioned method for determining the driving state. The memory 802 may include high-speed random access memory, and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory 802 may further include memory remotely located relative to the processor 804, and these remote memories can be connected to the terminal via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof. Specifically, the memory 802 may be used, but is not limited to, for storing file information such as target logical files. As an example, such as Figure 8As shown, the memory 802 may include, but is not limited to, the first determining unit 702, the second determining unit 704, the third determining unit 706, and the fourth determining unit 708 in the lane change intention generation device. Furthermore, it may include, but is not limited to, other module units in the driving state determining device, which will not be elaborated upon in this example.

[0289] Optionally, the transmission device 806 described above is used to receive or send data via a network. Specific examples of the network described above may include wired networks and wireless networks. In one example, the transmission device 806 includes a Network Interface Controller (NIC), which can be connected to other network devices and a router via a network cable to communicate with the Internet or a local area network. In another example, the transmission device 806 is a Radio Frequency (RF) module, used for wireless communication with the Internet.

[0290] In addition, the above-mentioned electronic device also includes a display 808 and a connection bus 810 for connecting the various module components in the above-mentioned electronic device.

[0291] According to one aspect of this application, a computer program product is provided, comprising a computer program / instructions containing program code for performing the methods shown in the flowchart. In such embodiments, the computer program can be downloaded and installed from a network via a communication component, and / or installed from a removable medium. When the computer program is executed by a central processing unit, it performs various functions provided in embodiments of this application.

[0292] The sequence numbers of the embodiments in this application are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0293] It should be noted that the computer system of the electronic device is merely an example and should not impose any limitations on the functionality and scope of use of the embodiments of this application.

[0294] Specifically, according to embodiments of this application, the processes described in the various method flowcharts can be implemented as computer software programs. For example, embodiments of this application include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication component, and / or installed from a removable medium. When the computer program is executed by a central processing unit, it performs various functions defined in the system of this application.

[0295] According to one aspect of this application, a computer-readable storage medium is provided, wherein a processor of a computer device reads computer instructions from the computer-readable storage medium, and executes the computer instructions, causing the computer device to perform the methods provided in the various alternative implementations described above.

[0296] Optionally, in this embodiment, the computer-readable storage medium described above may be configured to store a computer program for performing the following steps:

[0297] S1, determine the set of candidate driving states that match the first driving state, and determine the flexibility cost based on the number of candidate states of the second driving state included in the candidate driving state set.

[0298] The first driving state is used to indicate the driving state of the vehicle object at the first time node, and the candidate driving state set includes a second driving state that allows the vehicle object to switch directly from the first driving state.

[0299] S2, based on the operating status of at least one road object included in the road environment that matches the first driving state, determine the collision risk value of the vehicle object at the second time node;

[0300] The object profile is determined based on at least one driving operation of the vehicle object in the second driving cycle, and the second adjustment vector includes at least one second adjustment coefficient, which is used to perform a second adjustment operation on at least one corresponding acceleration state parameter.

[0301] S3, determine the trajectory value based on the driving trajectory of the vehicle object when it switches from the current driving state to the first driving state;

[0302] The acceleration state parameter set includes a first distance parameter, a second distance parameter, a deceleration parameter, and a maximum acceleration parameter. The first distance parameter is the safe distance between the self-vehicle and the vehicle in front when they are stationary. The second distance parameter is the safe time distance between the self-vehicle and the vehicle in front when they are in motion.

[0303] S4. If the weighted sum of the value of flexibility, the value of collision risk, and the value of trajectory satisfies the target condition, determine the state nodes in the state simulation tree based on the first driving state.

[0304] The state simulation tree includes multiple state nodes, which describe the driving state of the vehicle object at at least one time node.

[0305] Optionally, in this embodiment, those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be implemented by a program instructing related hardware of an electronic device. The program can be stored in a computer-readable storage medium, which may include: flash drive, read-only memory (ROM), random access memory (RAM), disk or optical disk, etc.

[0306] The sequence numbers of the embodiments in this application are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0307] If the integrated units in the above embodiments are implemented as software functional units and sold or used as independent products, they can be stored in the aforementioned computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause one or more computer devices (which may be personal computers, servers, or network devices, etc.) to execute all or part of the steps of the methods of the various embodiments of this application.

[0308] In the above embodiments of this application, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0309] In the several embodiments provided in this application, it should be understood that the disclosed user equipment can be implemented in other ways. The device embodiments described above are merely illustrative; for example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection of units or modules may be electrical or other forms.

[0310] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0311] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0312] The above description is only a preferred embodiment of this application. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of this application, and these improvements and modifications should also be considered within the scope of protection of this application.

Claims

1. A method for determining a driving state, characterized in that, include: A set of candidate driving states matching the first driving state is determined, and a flexibility cost is determined based on the number of candidate states of the second driving state included in the set of candidate driving states. The first driving state is used to indicate the driving state of the vehicle object at a first time node, and the set of candidate driving states includes a second driving state that allows the vehicle object to switch directly from the first driving state. Based on the operating status of at least one road object included in the road environment that matches the first driving state, the collision risk value of the vehicle object at the second time node is determined. The trajectory value is determined based on the driving trajectory of the vehicle object when it switches from the current driving state to the first driving state; If the weighted sum of the flexibility value, the collision risk value, and the trajectory value satisfies the target condition, a state node in the state simulation tree is determined based on the first driving state. The state simulation tree includes multiple state nodes, which describe the driving state of the vehicle object at at least one time node.

2. The method according to claim 1, characterized in that, Determine a set of candidate driving states that match the first driving state, including: Obtain the target time interval that matches the road environment, wherein multiple environment identifiers each correspond to a time interval; Based on the first time node and the target time interval, determine the second time node after the first time node; The candidate driving state set is determined based on at least one of the second driving states that the vehicle object is allowed to switch to at the second time node.

3. The method according to claim 2, characterized in that, The flexibility value is determined based on the number of candidate states of the second driving state included in the candidate driving state set, including one of the following: The product of the number of candidate states and the flexibility weight is determined as the flexibility cost. When at least one reference driving state set is determined based on the candidate driving state set, the flexibility value is determined based on the weighted sum between the number of candidate states and the number of at least one reference state, wherein the first reference driving state in the first reference driving state set of at least one of the reference driving state sets is obtained by simulation deduction based on the second driving state, and the i-th reference driving state in the i-th reference driving state set of at least one of the reference driving state sets is obtained by simulation deduction based on the (i-1)-th reference driving state.

4. The method according to claim 1, characterized in that, The step of determining the collision risk value of the vehicle at the first time node based on the operating status of at least one road object included in the road environment matching the first driving state includes: The current road object is determined from at least one of the road objects; Obtain the rate of change of the estimated collision time between the vehicle object and the current road object; If the time change rate is greater than 0, the negative value of the time change rate is used as the object risk cost value matching the current road object; If the rate of change over time is less than or equal to 0, then 0 is taken as the risk value of the object matching the current road object; Having determined the risk cost value corresponding to at least one of the road objects, the collision risk cost value is determined based on the weighted summation of the risk cost values ​​of at least one of the road objects.

5. The method according to claim 1, characterized in that, The step of determining the trajectory value based on the driving trajectory of the vehicle object when it switches from the current driving state to the first driving state includes at least one of the following: Obtain the lane offset cost value corresponding to each of the multiple trajectory points in the driving trajectory; determine the navigation offset cost value matching the driving trajectory based on the weighted sum of the multiple lane offset cost values, wherein the trajectory cost value is determined based on the navigation offset cost value, and the lane offset cost value of the trajectory point is determined based on the ratio between the trajectory point offset and the reference offset, wherein the trajectory point offset is the distance between the trajectory point position and the lane centerline of the target lane; The speed cost is determined based on the ratio between the average speed of the trajectory corresponding to the driving trajectory and the reference speed; the distance cost is determined based on the ratio between the first distance indicated by the endpoint of the trajectory and the reference distance; and the traffic efficiency cost of the driving trajectory is determined based on the weighted sum of the speed cost and the distance cost, wherein the trajectory cost is determined based on the traffic efficiency cost. The acceleration and rate of change of acceleration corresponding to each of multiple trajectory points in the driving trajectory are obtained, and the maximum trajectory curvature of the driving trajectory is obtained; based on the acceleration corresponding to each of the multiple trajectory points, the acceleration cost is determined; based on the rate of change of acceleration corresponding to each of the multiple trajectory points, the acceleration change cost is determined; based on the acceleration cost, the weighted sum of the acceleration change cost and the maximum trajectory curvature, the comfort cost is determined, wherein the trajectory cost is determined based on the comfort cost. The following steps are performed: Obtain the trajectory curvature and the rate of change of trajectory curvature for each of the multiple trajectory points in the driving trajectory; determine the curvature cost value based on the trajectory curvature for each of the multiple trajectory points; determine the rate of change of curvature cost value based on the rate of change of trajectory curvature for each of the multiple trajectory points; determine the curvature continuity cost value based on the weighted sum of the curvature cost value and the rate of change of trajectory curvature, wherein the trajectory cost value is determined based on the curvature continuity cost value. Obtain the collision time cost and relative distance cost corresponding to each of the multiple trajectory points in the driving trajectory; determine the safety cost matching the driving trajectory based on the exponential decay summation result of the multiple collision time cost values ​​and the exponential decay summation result of the multiple relative distance cost values, wherein the trajectory cost is determined based on the safety cost, the collision time cost is determined based on the exponential decay value between the estimated collision time of the trajectory point and the collision time safety threshold, and the relative distance cost is determined based on the exponential decay value between the relative distance of the trajectory point and the distance safety threshold.

6. The method according to claim 1, characterized in that, After determining the trajectory cost value based on the driving trajectory of the ego vehicle object switching from the current driving state to the first driving state, it further includes: When the weighted sum result among the flexibility cost value, the collision risk cost value, and the trajectory cost value is less than or equal to the cost value threshold, it is determined that the first driving state meets the target condition; When the weighted sum result among the flexibility cost value, the collision risk cost value, and the trajectory cost value is greater than the cost value threshold, a reference driving state is obtained from the reference driving state set as the next vehicle driving state to be simulated, where the reference driving state set includes the fourth driving state that allows the ego vehicle object to directly switch from the third driving state, and the third driving state is the vehicle driving state corresponding to the previous time node.

7. The method according to claim 6, characterized in that, Before obtaining a reference driving state from the reference driving state set as the next vehicle driving state to be simulated, it further includes: Obtain a target state machine diagram, where the target state machine diagram includes multiple state nodes, the state nodes are used to indicate a vehicle driving state, and the directed connections between the state nodes are used to indicate that a state switch is allowed between two vehicle driving states; According to the current state node in the target state machine diagram that matches the third driving state, determine at least one target state node pointed to by the current state node; Determine the reference driving state set according to at least one of the target state nodes.

8. The method according to claim 6, characterized in that, Before determining the trajectory cost value based on the driving trajectory of the ego vehicle object switching from the current driving state to the first driving state, it further includes: When determining the trajectory start point of the driving trajectory and the first driving state, sequentially determine the ego vehicle acceleration corresponding to multiple time stamps according to the target step size; Wherein, the ego vehicle acceleration is determined according to the target acceleration state set and the previous ego vehicle speed, and the target acceleration state set is obtained by adjusting at least one acceleration state parameter in the acceleration state parameter set according to the first adjustment vector matching the current road environment and the second adjustment vector matching the object portrait of the ego vehicle object.

9. The method according to claim 8, characterized in that, The sequentially determining the ego vehicle acceleration corresponding to multiple time stamps according to the target step size includes: Obtain the first adjustment vector, where the current road environment is determined according to at least one driving state of at least one other vehicle object in the first driving cycle, and the first adjustment vector includes at least one first adjustment coefficient, and at least one of the first adjustment coefficients is used to perform a first adjustment operation on each corresponding at least one acceleration state parameter; Obtain the second adjustment vector, where the object portrait is determined according to at least one driving operation of the ego vehicle object in the second driving cycle, and the second adjustment vector includes at least one second adjustment coefficient, and at least one of the second adjustment coefficients is used to perform a second adjustment operation on each corresponding at least one acceleration state parameter; Based on the first adjustment vector and the second adjustment vector, at least one acceleration state parameter in the acceleration state parameter set is adjusted to obtain the target acceleration state parameter set, wherein the acceleration state parameter set includes a first distance parameter, a second distance parameter, a deceleration parameter, and a maximum acceleration parameter. The first distance parameter is the safe distance between the self-vehicle object and the preceding vehicle object in a stationary state, and the second distance parameter is the safe time distance between the self-vehicle object and the preceding vehicle object in a moving state. Based on the vehicle speed corresponding to the previous timestamp and the target acceleration state parameter set, determine the vehicle acceleration corresponding to the current timestamp.

10. A device for determining a driving state, characterized in that, include: The first determining unit is configured to determine a set of candidate driving states that match the first driving state, and to determine a flexibility cost based on the number of candidate states of the second driving state included in the set of candidate driving states. The first driving state is used to indicate the driving state of the vehicle object at a first time node, and the set of candidate driving states includes a second driving state that allows the vehicle object to switch directly from the first driving state. The second determining unit is used to determine the collision risk value of the vehicle object at a second time node based on the operating status of at least one road object included in the road environment that matches the first driving state. The third determining unit is used to determine the trajectory value based on the driving trajectory of the vehicle object when it switches from the current driving state to the first driving state. The fourth determining unit is used to determine a state node in the state simulation tree based on the first driving state, provided that the weighted sum of the flexibility value, the collision risk value, and the trajectory value satisfies the target condition. The state simulation tree includes multiple state nodes, each describing the driving state of the vehicle object at at least one time node.

11. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a stored program, wherein the program is executed by an electronic device to perform the method according to any one of claims 1 to 9.

12. An electronic device comprising a memory and a processor, characterized in that, The memory stores a computer program, and the processor is configured to execute the method described in any one of claims 1 to 9 through the computer program.

Images