Automatic driving local motion planning method based on dynamic safety topological structure

Abstract

The invention relates to an automatic driving local motion planning method based on a dynamic safety topological structure, and belongs to the technical field of automatic driving, and the method comprises the following steps: S1, defining a detectable range of a host vehicle HV; s2, detecting, positioning and determining the state of an obstacle around the host vehicle HV, and obtaining the information of a lane marking; s3, according to a behavior decision-making principle, local motion behaviors are planned, and the local motion behaviors comprise following, lane changing and overtaking tasks; and S4, setting a maximum safety range, and performing motion planning on possible dangerous conditions. According to the method, the driving decision task can be decomposed and dynamically solved, safe, stable and high-adaptability driving decisions are ensured, and therefore dynamic and various driving scenes are effectively processed.

Description

Technical Field

[0001] The present invention belongs to the technical field of autonomous driving, and relates to an autonomous driving local motion planning method based on a dynamic safety topology structure. Background Art

[0002] Local motion planning generally includes creating a decision-making module that enables the host vehicle (HV) to adaptively make local motion decisions based on road conditions and the behavior of other vehicles (OVs). These decisions are mainly divided into rule-based methods and learning-based methods.

[0003] Rule-based methods have the characteristics of clear logic, strong model interpretability, and good scalability. However, they rely on limited rule parameters to make decisions in a dynamic environment, so their adaptability to complex driving scenarios is limited. Rule-based methods usually divide continuous decision-making behaviors into discrete driving behaviors by adopting predefined rules. Then, the vehicle state and environmental characteristics are evaluated against these rules to obtain the final driving decision. The finite state machine (FSM) is one of the most representative rule-based decision-making methods. The key components of the finite state machine include states, events, transitions, and actions. Each state responds to a specific event, executes the corresponding action, and transitions to a new state. Miller et al. developed the Skynet autonomous vehicle, which integrated traffic rules and environmental data in a three-layer planning system including driving behavior, planning strategy, and low-level operation control. Patz et al. developed the Knight Rider autonomous vehicle, which adopted a hierarchical driver model and divided the driving task into three levels: strategic, tactical, and operational. The strategic level sets sub-goals, the tactical level generates continuous optimal strategies, and the operational level outputs control signals. Ziegler et al. proposed a parallel decision-making system that includes four independent and parallel modules: route planning, goal analysis, traffic light management, and takeover management. This parallel structure ensures fast and flexible response by isolating each task into a separate module. Despite these advances, rule-based methods still face challenges under complex driving conditions. Traversing a large number of states may result in significant computational overhead. In addition, state partitioning and resolving conflicts between states are still key obstacles to improving the efficiency and reliability of these methods.

[0004] Learning-based decision-making methods are known for their self-learning ability, high decision-making flexibility, and in-depth exploration ability in specific scenarios. Compared with rule-based methods, learning-based methods can also simplify the scale of decision-making modules by leveraging network structures. Currently, learning-based decision-making methods mainly rely on deep learning and reinforcement learning (RL). For example, Wang et al. proposed a deep RL method for lane-changing speed control in a mixed traffic environment. This study utilized natural driving trajectory data to simulate the behavior of other vehicles and designed reward functions and collision avoidance strategies to train autonomous vehicles for safe and efficient lane changes. Similarly, Alagumuthukrishnan et al. developed a deep RL method that combines vehicle-to-vehicle (V2V) communication and sensor data to optimize the lane-changing behavior of connected autonomous vehicles on highways, thereby improving their safety and efficiency.

[0005] However, learning-based methods face challenges when operating in scenarios outside of the training data, which limits their safety in various scenarios. To address this issue, researchers have explored hybrid methods that combine rule-based and learning-based methods to enhance decision-making safety through risk constraints and takeover mechanisms. Yang et al. proposed a game-theory-based safety reinforcement learning framework. This framework combines a reinforcement learning network and a safety constraint module. The safety module ensures risk assessment and driving safety during and after training. It also takes over decision-making when reinforcement learning poses a driving risk. Similarly, Xu et al. proposed an integrated decision-making framework for highway automation. This framework combines reinforcement learning with imitation learning and uses rule-based corrections to address flawed decisions and improve safety. Despite these advancements, learning-based methods still suffer from poor interpretability. Even when combined with rule-based aids, the decision-making process of learning-based algorithms often remains opaque, presenting challenges for verification and deployment in safety-critical scenarios. Summary of the Invention

[0006] In view of this, the purpose of the present invention is to provide an autonomous driving local motion planning method based on a dynamic safety topological structure (DSTS), where DSTS uses a hierarchical topological structure to establish a basic decision-making framework.

[0007] To achieve the above objective, the present invention provides the following technical solutions:

[0008] An autonomous driving local motion planning method based on a dynamic safety topological structure, comprising the following steps:

[0009] S1: Define the detectable range of the host vehicle HV;

[0010] S2: Detect, locate and determine the status of obstacles around the host vehicle HV, and obtain lane marking information;

[0011] S3: According to the behavioral decision-making principle, local motion behaviors are planned, including following, lane changing, and overtaking tasks;

[0012] S4: Set the maximum safety range and perform motion planning for possible dangerous situations.

[0013] Further, in step S1, based on the following factors, with the host vehicle HV as the center, D max For the detection radius, set the detectable range of the host vehicle:

[0014] The maximum effective detection range of vehicle-mounted sensors in actual applications;

[0015] The need to limit the amount of computation in the surrounding environment in order to improve computational efficiency.

[0016] Further, step S2 specifically includes the following steps:

[0017] Establish a topological range with the main vehicle HV as the center and define the range parameter d lane-obs as follows:

[0018]

[0019] β1·D safe ≤d lane-obs ≤β2·D max

[0020] stTTC t =TTC t (HV,OV t )if OV t else TTC safe

[0021] TTC t ≤TTC safe

[0022] Among them, TTC t Indicates the host vehicle HV and the nearest vehicle OV in the HV's intended lane t The collision time TTC between the two max N other vehicles within OV, OV t|1≤t≤N Indicates the vehicle closest to the HV in the HV's intended lane; D safe represents the radius of the attention area, which depends on the selected driving style; v HV is the speed of HV, wpt HV is the route planning point of HV, β1 and β2 are hyperparameters;

[0023] Parameter d lane-obs For balancing the following factors: (1) When N OVs i|1≤i≤N coexist, accurately calculate TTC i ; (2) As v HV increases, expand the decision-making range; (3) When the curvature of the planned route of the host vehicle HV is high, reduce the possibility of lane-changing decisions;

[0024] For the road environment within the range of d lane-obs , it is divided into two parts with the longitudinal axis of HV as the dividing line: the front calculation layer and the rear calculation layer;

[0025] The front calculation layer is responsible for calculating the traffic information in front of the host vehicle, including the front, the left front, and the right front; the rear calculation layer is responsible for calculating the traffic information behind the host vehicle, including the rear, the left rear, and the right rear;

[0026] For front detection, first collect the speed data v of the vehicles traveling in the same direction in the lane set ov and the distance data d obs , and represent the collected data as a combined array:

[0027]

[0028] wherein, x lane-marking represents the lateral distance between HV and the lane marking; OV state represents the presence of a vehicle, which is true when a vehicle is present; when OV state is false, d obs is used to calculate the space occupancy to ensure that the overall safety topological space around HV remains complete and does not collapse.

[0029] Furthermore, in step S2, for the driving scenario where lane markings overlap and multiple IDs lane are confused with each other, a regular hexagon detection structure is proposed, which is complementary to the range of d lane-obs ;

[0030] For this complementary structure, the latitude axis is used as a fixed dividing line, and the area around HV is divided into six parts with a fixed angle: the front, the left front, the right front, the rear, the left rear, and the right rear; then the surrounding vehicles are partitioned and positioned for calculation according to the defined areas.

[0031] Furthermore, for the lane-changing and overtaking operations based on the local path, the following conditions need to be satisfied simultaneously:

[0032] (1) If there is already a global planned path, the lane-changing operation Event based on the global path planningglobal must have a higher priority than the lane change operation Event based on local path planning local ; If a local lane change operation conflicts with the global path planning, the vehicle will abandon the lane change and switch to the safe following mode. The specific operations are as follows:

[0033]

[0034] (2) There should be a time interval between two consecutive lane change operations. The lane change interval time t gap is adaptively adjusted according to v HV and is set as follows:

[0035]

[0036] where v max is the maximum speed of the host vehicle HV, and T is a predefined constant and T belongs to positive integers;

[0037] (3) In terms of longitudinal positioning, the distance d between the HV and the vehicle in front in the same lane longitude must be lower than the threshold d turn , as follows:

[0038] (d longitude < d turn ) ∧ (TTC(HV, OV longitude ) < TTC th )

[0039] ∧ (OV longitude is True)

[0040] where the setting of TTC th is affected by the driving style selection. The more aggressive the driving style, the smaller the value of TTC th ; d turn is an adjustable threshold used to determine whether a lane change operation should be performed; the value of d turn is defined as

[0041] d turn = (w1 - w2Curvature(wpt HV )) · d lane-obs

[0042] where 0 < w1 < 1, Curvature({wpt HV}) represents the route curvature estimated by a quadratic polynomial based on multiple planned path points {wpt HV} near the HV;

[0043] (4) In terms of lateral positioning, the distances to the leading vehicle and the trailing vehicle in the target lane are obtained, and the time to collision TTC is estimated; the lateral distance d latitude and the TTC value must comply with a preset threshold range; if the OV of the vehicle state is False, the distance and condition are automatically considered satisfied, and the specific description is as follows:

[0044]

[0045] where bj ∈ {front, rear}, d safe is the safe lane change distance, and its definition is as follows:

[0046]

[0047] where 0 < β3 < β1, β3 > 1, v desired is the desired speed of the host vehicle HV, is the minimum braking distance.

[0048] Furthermore, for a lane change operation based on local motion, it is necessary to satisfy

[0049] (d longitude < d turn ) ∧ (TTC(HV, OV longitude ) < TTC th )

[0050] ∧ (OV longitude is True)

[0051] For a lane change operation based on the global path, in addition to satisfying the above formula, it also satisfies the safety condition for the vehicle ahead, and this condition is described as:

[0052] (TTC(HV, OV longitude ) < TTC th ) ∨ (OV longitude is False)

[0053] When performing a lane change operation, the desired lane change speed v of the host vehicle HV lc is set as follows:

[0054]

[0055] s.t. v lc = max(v lc , v min )

[0056] where 0 < σ1 < σ2, v min is the minimum speed limit of the host vehicle;

[0057] When the lane change condition of the host vehicle HV is not met, the host vehicle performs lane keeping or following operations; the frequency of lane changes depends on the driving style of the HV. The more aggressive the driving style, the higher the frequency of lane changes; when it is determined that the host vehicle HV needs to stay in the current lane, its driving speed is automatically adjusted according to the surrounding environmental conditions.

[0058] Furthermore, the environmental conditions are divided into three categories: (I) structured roads, i.e., straight roads and curves, (II) unstructured roads, i.e., intersections, (III) staying in the lane and waiting for a lane change to meet the global path planning; where the priority of (III) is higher than that of (I) and (II);

[0059] Condition (I): For scenarios where lane change situations do not need to be considered, a piecewise dynamic speed planning method is used; the dynamic safety constraint is used as the confidence range for speed adjustment. Within this range, a speed control strategy is executed to enable the host vehicle HV to adapt to the surrounding environment, while ensuring safety through dynamic piecewise constraints. The speed of the HV is set as follows:

[0060]

[0061] where 0 < μ1 < 1 < μ2, 0 < γ1 < 1, γ2 > 1, γ3v max > v desired , η ≥ 1, η ∝ μ2, TTC t represents the time required for the host vehicle HV to collide with other vehicles OV in the target lane;

[0062] Condition (II): In the intersection scenario, the speed of the host vehicle HV is adjusted according to the speeds of the surrounding vehicles. The speed of the host vehicle HV is set as follows:

[0063]

[0064] where 0 < γ4 < 1, 0 < γ5 < 1;

[0065] Condition (III): If, based on the global path planning, the host vehicle HV needs to change lanes, then during the period of waiting for the lane change condition to be met, the speed of the HV will decrease. The speed of the HV is set as follows:

[0066]

[0067] where 0 < μ3 < 1, 0 < μ4 < 1, 0 < γ6 < 1, 0 < γ7 < 1.

[0068] Furthermore, in step S4, a maximum safety range is set, i.e., the radius D of the area of concern safe ;

[0069] For the target lane of the HV, if the target lanes are adjacent, when d(HV, OV front ) < D safe the HV activates a braking action; if the target lane is the lane where the HV itself is located, when d(HV, OV front ) < D safe and d(HV, OV front ) < d brake the HV will activate a braking action;

[0070] Introduce the dynamic parameter d brake , adjust the boundaries of safe driving behavior, thereby expanding the operating range while maintaining safety;

[0071] If the rear OV is too close to the HV, it may cause a rear-end collision. The HV will activate a lane-changing action. To change lanes to avoid a rear-end collision, the longitudinal positioning conditions must be met:

[0072] (d longitude < d turn ) ∧ (TTC(HV, OV longitude ) < TTC th )

[0073] ∧ (OV longitude is True)

[0074] The potential danger judgment criteria for the rear vehicle to the HV must also be met:

[0075] (TTC(HV, OV rear ) < TTC th )

[0076]

[0077] where d brake is defined as:

[0078]

[0079] where is the braking distance threshold, set according to the driving style of the host vehicle HV; Set according to the state of other surrounding vehicles OV. When other vehicles are stationary, assign a very small value.

[0080] The beneficial effects of the present invention are as follows: By incorporating dynamic environment parameters, each layer of the DSTS of the present invention can dynamically adapt to the surrounding driving environment and the driving styles of different autonomous vehicles. This structure can decompose driving decision-making tasks and solve them dynamically, ensuring safe, stable, and highly adaptable driving decisions, thereby effectively handling dynamic and multi-class driving scenarios.

[0081] Other advantages, objectives, and features of the present invention will, to some extent, be elaborated in the subsequent description, and to some extent, will be obvious to those skilled in the art based on the examination of the following text, or can be learned from the practice of the present invention. The objectives and other advantages of the present invention can be achieved and obtained through the following description. Brief Description of the Drawings

[0082] In order to make the objectives, technical solutions, and advantages of the present invention clearer, the present invention will be described in detail preferably with reference to the accompanying drawings, where:

[0083] Figure 1 It is a schematic diagram of DSTS;

[0084] Figure 2 They are four sub-structural layers of DSTS;

[0085] Figure 3 It is a schematic diagram of the detection layer;

[0086] Figure 4 It is a schematic diagram of the obstacle and lane marking calculation layer;

[0087] Figure 5 It is a supplement to the obstacle and lane marking calculation layer;

[0088] Figure 6 It is a schematic diagram of the behavior decision-making layer;

[0089] Figure 7 It is a schematic diagram of the safety layer;

[0090] Figure 8 It is a driving scenario; where (a) is scenario 1, (b) is scenario 2, (c) is scenario 3, and (d) is scenario 4;

[0091] Figure 9 They are the performance scores of DSTS in straight and curved scenarios, where (a) is scenario 1 and (b) is scenario 2;

[0092] Figure 10 They are the performance scores of DSTS in scenario 3 and scenario 4, where (c) is scenario 3 and (d) is scenario 4;

[0093] Figure 11 They are the visualized trajectories in scenario 1, where (a) is the static scenario and (b) is the dynamic scenario;

[0094] Figure 12 They are the visualized trajectories in scenario 2, where (a) is the static scenario and (b) is the dynamic scenario;

[0095] Figure 13For the visualization trajectories in Scenario 3, where (a) represents a cautious driving style, (b) represents a steady driving style, and (c) represents an aggressive driving style;

[0096] Figure 14 For the visualization trajectories in Scenario 4, where (a) represents a cautious driving style, (b) represents a steady driving style, and (c) represents an aggressive driving style;

[0097] Figure 15 For the time steps of Scenario 1 and Scenario 2, where (a) is Scenario 1 and (b) is Scenario 2;

[0098] Figure 16 For the time steps of Scenario 3 and Scenario 4, where (c) is Scenario 3 and (d) is Scenario 4. Detailed implementation manners

[0099] The following uses specific specific examples to illustrate the implementation manners of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific implementation manners. The details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that the diagrams provided in the following embodiments only illustrate the basic concept of the present invention in a schematic manner. Without conflict, the following embodiments and the features in the embodiments can be combined with each other.

[0100] It should be noted that the diagrams provided in the following embodiments only illustrate the basic concept of the present invention in a schematic manner. Therefore, only the components related to the present invention are shown in the diagrams, rather than being drawn according to the number, shape, and size of the components in actual implementation. The type, quantity, and proportion of each component in its actual implementation can be arbitrarily changed, and the component layout type may also be more complex.

[0101] In the following description, a large number of details are explored to provide a more thorough explanation of the embodiments of the present invention. However, it is obvious to those skilled in the art that the embodiments of the present invention can be implemented without these specific details. In other embodiments, well-known structures and devices are shown in the form of block diagrams rather than in detail to avoid making the embodiments of the present invention difficult to understand.

[0102] Such as Figure 1As shown, in order to accurately evaluate the risks related to the driving state and make correct decisions, the present invention proposes an autonomous driving local motion planning method based on a dynamic safety topology structure (DSTS). In terms of design, DSTS divides different driving conditions and establishes basic driving behavior logics to ensure the interpretability of decisions while avoiding overly complex rule-based calculations. In addition, by introducing flexible structural parameter settings, in the learning-based method, the dynamic decision-making parameters of the system can be configured according to the continuously changing behavior parameters of the host vehicle (HV) and surrounding vehicles (OV). This enables the system to output flexible and environment-adaptive dynamic decision-making controls. DSTS is mainly used for decision-making on structured roads, and its output results are processed by lateral and longitudinal PID controllers. In addition, it also provides a supplementary solution to handle scenarios of overlapping or separated roads, thereby ensuring robust performance in more complex road layouts. Specifically, by analyzing road conditions and surrounding vehicles, DSTS uses lane positioning, distance calculation, and time-to-collision (TTC) analysis to establish a safe topological range. This local safety detection framework supports key driving decisions such as following and lane changing, ensuring safe and adaptive navigation in dynamic and complex scenarios.

[0103] As Figure 2 shown, DSTS includes four sub-structural layers: a detection layer, an obstacle and lane line calculation layer, a behavior decision layer, and a safety layer. The parameters and functions of each layer are described in detail as follows.

[0104] a) Detection layer. The detection layer defines the detectable range of the host vehicle (HV). As Figure 3 shown, its range parameter D max sets the boundary of the detection circle centered on the HV. The value of D max is determined by two factors: the maximum effective detection range of on-vehicle sensors in practical applications and the requirement for limiting the computational amount of the surrounding environment to improve computational efficiency.

[0105] b) Obstacle and lane marking calculation layer. As Figure 4 shown, the obstacle and lane marking calculation layer is responsible for detecting, positioning, determining the state of obstacles around the host vehicle, and obtaining information on lane markings. This layer establishes a topological range centered on the HV, which is defined by the formula of the range parameter d lane-obs obtained through experience. The expression of d lane-obs determined through experiments and data analysis is as follows:

[0106]

[0107] where TTC t represents the TTC between the HV and the nearest OV t in the intended lane of the HV. For those within the range Dmax Among the N OVs within, OV t|1≤t≤N represents the vehicle closest to the HV in the HV intended lane. D safe represents the radius of the attention area, which depends on the selected driving style. v HV is the speed of the HV, wpt HV is the route planning point of the HV, and β1 and β2 are hyperparameters. Parameter d lane-obs is designed to balance multiple factors: (1) accurately calculate the TTC when N OVs i|1≤i≤N coexist; (2) expand the decision-making range as v i increases; (3) reduce the likelihood of lane-changing decisions when the curvature of the planned route of the host vehicle HV is high to maintain safety and stability. HV

[0108] For the road environment within the range of d lane-obs this layer is divided into two parts with the longitudinal axis of the HV as the dividing line: the front calculation layer and the rear calculation layer. These two layers are responsible for calculating the traffic information in the front (including the front, left front, and right front) and the rear (including the rear, left rear, and right rear) respectively. The obstacle and lane marking calculation layer gives priority to processing the surrounding moving vehicles, followed by the lane marking information. Taking the front detection as an example, this layer first collects the speed data v of the vehicles traveling in the same direction in the lane set ov and the distance data d obs . The collected data is then represented as a combined array:

[0109]

[0110] where, x lane-marking represents the lateral distance between the HV and the lane marking. OV state represents the presence of a vehicle, which is true when a vehicle is present. When OV state is false, d obs is mainly used to calculate the space occupancy to ensure that the overall safety topological space around the HV remains complete and does not collapse.

[0111] For vehicles with the same ID lane the surrounding vehicles closest to the HV are filtered out according to their distance d obs from the HV. Then the speed information of these vehicles is obtained. If there is no vehicle in a certain lane within the distance d lane-obs the focus is shifted to the lane marking of that lane.

[0112] As Figure 5 shown, for overlapping lane markings and multiple IDs laneConfusing driving scenarios, such as those at intersections and on-ramp areas, present a structure that is complementary to the obstacle and lane marking calculation layers. This structure is designed based on regular hexagons. Compared with the previous structure, the main difference lies in the way of positioning surrounding vehicles. The previous structure relied on the latitude axis of the host vehicle (HV) and the IDs of surrounding vehicles lane of the latitude axis for position calculation. This complementary structure uses the latitude axis as a fixed dividing line and divides the area around the HV into six parts at fixed angles: front, front left, front right, rear, rear left, and rear right. Then, the surrounding vehicles are partitioned and positioned based on these defined areas.

[0113] c) Behavior decision layer. As Figure 6 shown, the behavior decision layer is responsible for local behavior planning, including tasks such as following, lane changing, and overtaking.

[0114] 1) Principles.

[0115] The behavior decision principles can be mainly divided into three categories: lane changing and overtaking based on local paths, lane changing based on global paths, and lane changing to avoid following too closely. This layer ensures that the vehicle can make safe and efficient decisions according to the dynamic traffic conditions and optimize local and global driving strategies.

[0116] For lane changing and overtaking operations based on local paths, the following conditions must be met simultaneously:

[0117] (1) The priority of the lane change operation Event global based on global path planning must be higher than the lane change operation Event local based on local path planning. If there is a conflict between the local lane change and the global path, the vehicle will abandon the lane change and switch to the safe following mode, as shown in the following operations:

[0118]

[0119] (2) To prevent continuous rapid lane changes and ensure safety, a period of time is required between two consecutive lane change operations. The lane change interval time t gap is adaptively adjusted according to v HV and is set as follows:

[0120]

[0121] where v max is the maximum speed of the host vehicle (HV), and T is a predefined constant and T belongs to the set of positive integers.

[0122] (3) In terms of longitudinal positioning, the distance d longitude between the HV and the vehicle in front in the same lane must be lower than the threshold d turn, as shown in the following logical operation:

[0123] (d longitude <d turn ) ∧ (TTC(HV, OV longitude ) < TTC th ) ∧ (OV longitude is True) (5)

[0124] Among them, the setting of TTC th is affected by the driving style selection. The more aggressive the driving style, the smaller the value of TTC th . d turn is an adjustable threshold used to determine whether a lane change operation should be performed. The value of d turn is defined as

[0125] d turn = (w1 - w2Curvature(wpt HV )) · d lane-obs (6)

[0126] Among them, 0 < w1 < 1, Curvature({wpt HV}) represents the route curvature estimated by a quadratic polynomial based on multiple planned path points {wpt HV} near HV. By introducing the calculation of Curvature({wpt HV}), when the route curvature is large, the probability of the host vehicle (HV) changing lanes is reduced, making the driving operation more in line with human expectations.

[0127] (4) In terms of lateral positioning, it is necessary to obtain the distances and estimate the time to collision (TTC) of the vehicle in front and the vehicle behind in the target lane. The lateral distance d latitude and the TTC value must meet the preset threshold range. If the OV state of the vehicle is False, the distance and condition are automatically considered satisfied, and the specific description is as follows:

[0128]

[0129] Among them, bj ∈ {front, rear}, d safe is the safe lane change distance, and its definition is as follows:

[0130]

[0131] Among them, 0 < β3 < β1, β3 > 1, v desired is the expected speed of the host vehicle (HV), is the shortest braking distance.

[0132] 2) Lane-changing Scheme

[0133] For lane-changing operations based on local motion, the conditional formula (5) should be satisfied. For lane-changing operations based on the global route, in addition to the conditional formula (5) in longitudinal positioning, the safety condition for the vehicle ahead must also be satisfied, which can be described as:

[0134] (TTC(HV,OV longitude ) < TTC th ) ∨ (OV longitude is False) (9)

[0135] When performing a lane-changing operation, the desired lane-changing speed v of the host vehicle (HV) lc is set as follows:

[0136]

[0137] where 0 < σ1 < σ2, v min is the minimum speed limit of the host vehicle.

[0138] 3) Lane-keeping Scheme

[0139] When the lane-changing conditions of the host vehicle (HV) are not met, the vehicle will perform lane-keeping or following operations. The frequency of lane-changing depends on the driving style of the HV. The more aggressive the driving style, the higher the frequency of lane-changing.

[0140] When it is determined that the host vehicle (HV) needs to stay in the current lane, its driving speed will be automatically adjusted according to the surrounding environmental conditions. The environmental conditions can be divided into three categories: (I) structured roads (i.e., straight roads and curves), (II) unstructured roads (i.e., intersections), and (III) staying in the lane and waiting for lane-changing to meet the global route planning. Among them, the priority of (III) is higher than that of (I) and (II).

[0141] Condition (I): For scenarios where lane-changing does not need to be considered, a piecewise dynamic speed planning module is proposed. Generally, the piecewise speed state function can provide better safety, but its adaptability is limited; the dynamic speed parameters can achieve more flexible adjustment, but the safety and stability are insufficient. Inspired by the TRPO algorithm in reinforcement learning, which updates the decision-making policy within the trust region, this module uses the dynamic safety constraint as the trust range for speed adjustment and executes the speed control policy within this range. This method enables the host vehicle (HV) to adapt to the surrounding environment while ensuring safety through dynamic piecewise constraints. The speed of the HV is set as follows:

[0142]

[0143] where 0 < μ1 < 1 < μ2, 0 < γ1 < 1, γ2 > 1, γ3vmax >v desired , η ≥ 1 and η ∝ μ2, TTC t Represents the time required for the host vehicle (HV) to collide with other vehicles (OV) in the target lane.

[0144] Condition (ii): In intersection scenarios (e.g., exit ramps, entrance ramps, and intersections), traffic conditions are usually more complex. The speed of the host vehicle (HV) is adjusted according to the speeds of surrounding vehicles to ensure that it has sufficient reaction time to respond to potential hazards. This adjustment enables the vehicle to safely pass through the intersection, adapt to the traffic flow, and provides more time for decision-making compared to typical straight or curved road scenarios. The speed of the host vehicle (HV) Is set as follows:

[0145]

[0146] Where 0 < γ4 < 1, 0 < γ5 < 1.

[0147] Condition (III): If, based on the global route plan, the host vehicle (HV) needs to change lanes, the speed of the HV will decrease during the period of waiting to meet the lane-changing conditions. This allows the vehicle to safely prepare for lane changes, ensures sufficient distance from surrounding vehicles, and minimizes the risk of sudden lane changes. The speed of the HV Is set as follows:

[0148]

[0149] Where 0 < μ3 < 1, 0 < μ4 < 1, 0 < γ6 < 1 and 0 < γ7 < 1.

[0150] d) Safety layer. As Figure 7 Shown, the safety layer is the basic safety guarantee of the HV. The parameter D safe Represents the maximum safety range, i.e., the radius of the area of concern. For the target lane of the HV, if the target lane is adjacent, when d(HV, OV front ) < D safe The HV will initiate a braking action. If the target lane is the lane where the HV itself is located, when d(HV, OV front ) < D safe And d(HV, OV front ) < d brake The HV will initiate a braking action. By introducing the dynamic parameter d brake, HV can adjust the boundaries of safe driving behavior, thus expanding the operating range while maintaining safety. At the same time, if the following OV is too close to HV, which may lead to a rear-end collision, HV will initiate a lane-changing maneuver. To avoid a rear-end collision and change lanes, the longitudinal positioning conditional expression (5) must be satisfied. In addition, the potential danger judgment criteria for the following vehicle with respect to HV must also be met.

[0151]

[0152] where d brake is defined as:

[0153]

[0154] where is the braking distance threshold, which is set according to the driving style of the host vehicle (HV). It should be noted that where is set according to the state of other surrounding vehicles (OV). When other vehicles are stationary, a smaller value is assigned.

[0155] Four scenarios, including static and dynamic conditions, were tested using the method of the present invention. In the dynamic scenario, a car-following strategy was used as a comparison benchmark, and the autopilot of CARLA was used in combination with the settings defined by formula (11). The driving style parameters of HV in the benchmark were kept consistent with those in the DSTS configuration. Other vehicles (OV) in the dynamic scenario were also controlled using the autopilot of CARLA. Table 1 provides the specific configurations of each scenario. Figure 8 Figures (a)-(d) therein show the driving scenarios and the routes to be completed. The red route represents the route to be completed, while the blue route indicates the section where the high-voltage vehicle must take action to successfully drive off the ramp. It should be noted that in Scenario 4, HV uses the VI algorithm to drive off the ramp, and if HV deviates from the original planned route, the algorithm will re-plan the route.

[0156] The driving performance was evaluated using both qualitative and quantitative methods, covering a variety of scenarios. Qualitative evaluation included scoring the driving performance and visualizing the driving trajectory to evaluate the behavior and effects. Quantitative evaluation involved recording the time steps to measure the efficiency and making a side-by-side comparison of the scenario settings and their corresponding performance metrics to demonstrate their superiority. The performance score is defined as follows:

[0157] R = r ttc + r collision + r speed + r time + r target (16)

[0158] The definition of each sub-part follows the piecewise reward function design outlined in the literature and is established as follows:

[0159]

[0160] r time =-1.1-0.001·t (20)

[0161]

[0162] in r ttc and r collision is the safety assessment parameter, r speed is the stability evaluation parameter, r time is the efficiency evaluation parameter, t is the current time step of the host vehicle (HV), r target It’s a reward for reaching a goal.

[0163] Table 1

[0164]

[0165] Experimental results based on driving performance scores show that DSTS performs well in both dynamic and static scenarios, works well with the VI algorithm, and can adapt to various road types and driving style parameters. Figure 9 (a) and (b) show the performance scores of DSTS in straight and curved scenarios. In dynamic scenarios, DSTS significantly outperforms the baseline model. This shows that DSTS is able to improve driving performance through operations such as lane changing and overtaking while ensuring driving safety. In addition, the performance score of DSTS in dynamic scenarios is higher than that in static scenarios. This improvement is attributed to the dynamic decision space provided by traffic flow in dynamic scenarios, which enables HVs to take advantage of environmental changes to optimize the timing of operations. In contrast, in static scenarios, due to the fixed position of the OV, the decision space is limited, which limits the ability of HVs to optimize decisions. Figure 10 (a) and (b) show the performance scores of DSTS in scenarios 3 and 4. The images are smoothed for a clearer presentation. The results show that DSTS outperforms the baseline model in a variety of driving styles, demonstrating adaptability to different parameters such as speed, safety distance, and reaction time. Figure 10The results also show that in Scenario 4, the combination of the VI algorithm and the DSTS algorithm has proven to be highly effective, ensuring that the HV maintains driving efficiency, stability, and safety while executing a successful route plan. The synergy between these algorithms is largely attributed to the fast computational ability of VI, which enables timely decision-making and re-planning of routes according to the dynamic changes in the environment. In addition, DSTS's focus on safety and stability helps the HV navigate in complex and dynamic environments by ensuring that critical decisions (such as lane changes, overtaking, and adapting to traffic flow) are made in a controlled and safe manner. By analyzing the driving trajectories of the HV in different scenarios, the results show that the DSTS algorithm has strong driving stability and safety.

[0166] Figure 11 (a), (b) and Figure 12 (a), (b) in show the test driving trajectories of the HV in Scenario 1 and Scenario 2. These figures show that the HV has successfully completed the driving tasks in basic dynamic and static scenarios, including driving on turning and straight roads. Specifically, as Figure 11 (a) and Figure 12 (a) shows, in the static scenario, the driving trajectory of the HV remains stable and maintains a safe distance from surrounding vehicles. In addition, in the static scenario, the curvature of the lane change trajectory of the HV is higher than that in the dynamic scenario. This difference is because in the dynamic scenario, the HV can maintain a certain distance from surrounding vehicles, enabling a smoother lane change. On the contrary, in the static scenario, the surrounding vehicles are stationary, resulting in a gradually decreasing relative distance between the HV and the surrounding vehicles. To ensure safety, the HV must complete the lane change more promptly and quickly.

[0167] Figure 13 (a)-(c) and Figure 14 (a)-(c) respectively show the driving trajectories of the HV under different driving styles in Scenario 3 and Scenario 4. During the continuous scenario transitions, the HV always maintains a smooth and coherent driving trajectory. The HV shows stability in maintaining the lane and lane change operations. These observations indicate that the Dynamic Driving Task System (DSTS) can effectively support stable and safe driving performance in different driving styles and continuously changing scenarios, ensuring adaptability and reliability in complex driving environments.

[0168] In the statistical results of time steps, the driving efficiency of the DSTS-based HV is significantly better than the benchmark in all scenarios. As Figure 15 (a), (b) show, in both static and dynamic scenarios, DSTS enables the HV to complete the driving tasks with fewer time steps. This result highlights DSTS's ability to maintain excellent passing efficiency in different scenario states. In addition, as Figure 16As shown in (a) and (b) below, under the same driving style, DSTS can enable the HV to further improve passing efficiency. This indicates that DSTS can not only optimize passing efficiency under different driving styles, but also handle relatively complex driving tasks, ensuring adaptability and effectiveness in various scenarios. Appropriate lane-changing and overtaking operations are crucial for improving driving efficiency. The experimental statistical results highlight the effectiveness of these operations. In the static scenario, the autonomous vehicle made 1352 lane-changes on straight roads and 1192 lane-changes on curved roads. In the dynamic scenario, the autonomous vehicle completed 788 lane-changes on straight roads and 748 lane-changes on curved roads. Different driving styles were tested in Scenario 3 and Scenario 4. In Scenario 3, the cautious autonomous vehicle completed 1242 lane-changes, the normal autonomous vehicle completed 1476 lane-changes, and the aggressive autonomous vehicle completed 1483 lane-changes. In Scenario 4, the cautious autonomous vehicle completed 1118 lane-changes, the normal autonomous vehicle completed 1288 lane-changes, and the aggressive autonomous vehicle completed 1274 lane-changes. It is worth noting that no collisions occurred in all scenario tests, indicating that its operations are highly safe. In addition, the number of lane-changes under the normal and aggressive driving styles is similar. This is mainly due to other surrounding vehicles, especially in high-density traffic scenarios. In such cases, the ability of the autonomous vehicle to reach the desired speed is limited, reducing overtaking opportunities. However, the time-step statistics show that the average speed of the aggressive autonomous vehicle is higher than that of the conservative autonomous vehicle.

[0169] In the above embodiments, the reference in the specification to "this embodiment" means that the specific features, structures, or characteristics described in connection with the embodiment are included in at least some embodiments, but not necessarily all embodiments. Multiple occurrences of "this embodiment" do not necessarily all refer to the same embodiment.

[0170] In the above embodiments, although the present invention has been described in connection with specific embodiments of the present invention, many substitutions, modifications, and variations of these embodiments will be apparent to those of ordinary skill in the art based on the foregoing description. For example, other storage structures (e.g., dynamic RAM (DRAM)) may be used in the embodiments discussed. Embodiments of the present invention are intended to cover all such substitutions, modifications, and variations that fall within the broad scope of the appended claims.

[0171] This embodiment also provides a computer-readable storage medium having a computer program stored thereon, which when executed by a processor implements any one of the methods in this embodiment.

[0172] This embodiment also provides an electronic terminal, including: a processor and a memory;

[0173] The memory is used to store a computer program, and the processor is used to execute the computer program stored in the memory, so that the terminal executes any of the methods in this embodiment.

[0174] For the computer-readable storage medium in this embodiment, those of ordinary skill in the art can understand that all or part of the steps of implementing the above method embodiments can be completed by hardware related to the computer program. The foregoing computer program can be stored in a computer-readable storage medium. When the program is executed, it executes the steps including the above method embodiments; and the foregoing storage medium includes: ROM, RAM, magnetic disk or optical disc and other media that can store program codes.

[0175] The electronic terminal provided in this embodiment includes a processor, a memory, a transceiver, and a communication interface. The memory and the communication interface are connected to the processor and the transceiver and complete communication with each other. The memory is used to store a computer program, the communication interface is used for communication, and the processor and the transceiver are used to run the computer program, so that the electronic terminal executes each step of the above method.

[0176] In this embodiment, the memory may include a random access memory (Random Access Memory, abbreviated as RAM), and may also include a non-volatile memory, such as at least one disk memory.

[0177] The above-mentioned processor may be a general-purpose processor, including a central processing unit (Central Processing Unit, abbreviated as CPU), a network processor (Network Processor, abbreviated as NP), etc.; it may also be a digital signal processor (Digital Signal Processing, abbreviated as DSP), an application specific integrated circuit (Application Specific Integrated Circuit, abbreviated as ASIC), a field programmable gate array (Field-Programmable Gate Array, abbreviated as FPGA), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components.

[0178] The present invention can be used in many general-purpose or special-purpose computing system environments or configurations. For example: personal computers, server computers, handheld or portable devices, tablet devices, multi-processor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics, network PCs, minicomputers, mainframe computers, distributed computing environments including any of the above systems or devices, and so on.

[0179] The present invention may be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform particular tasks or implement particular abstract data types. The present invention may also be practiced in distributed computing environments where tasks are performed by remote processing devices connected through a communications network. In a distributed computing environment, program modules may be located in both local and remote computer storage media including storage devices.

[0180] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention may be modified or equivalently replaced without departing from the spirit and scope of the technical solutions, and they should all be covered by the scope of the claims of the present invention.

Claims

1. A local motion planning method for autonomous driving based on a dynamic safety topology structure, characterized by: The following steps are involved: S1: define the detectable range of the host vehicle HV; S2: Detect, locate and determine the status of obstacles around the host vehicle HV, and obtain lane marking information; S3: According to the behavioral decision-making principle, local motion behaviors are planned, including following, lane changing, and overtaking tasks; S4: Set the maximum safety range and perform motion planning for possible dangerous situations.

2. The method for local motion planning of autonomous driving based on dynamic safety topology structure according to claim 1, characterized in that: In step S1, based on the following factors, D max For the detection radius, set the detectable range of the host vehicle: The maximum effective detection range of vehicle-mounted sensors in actual applications; The need to limit the amount of computation in the surrounding environment in order to improve computational efficiency.

3. The method for local motion planning of autonomous driving based on dynamic safety topology structure according to claim 1, characterized in that: Step S2 specifically includes the following steps: Establish a topological range with the main vehicle HV as the center and define the range parameter d lane-obs as follows: β1·D safe ≤d lane-obs ≤β2·D max s.t.TTC t =TTC t (HV,OV t )if OV t else TTC safe TTC t ≤TTC safe Among them, TTC t Indicates the host vehicle HV and the nearest vehicle OV in the HV's intended lane t The collision time TTC between the two max N other vehicles within OV, OV t|1≤t≤N Indicates the vehicle closest to the HV in the HV's intended lane; D safe represents the radius of the attention area, which depends on the selected driving style; v HV is the speed of HV, wpt HV is the route planning point of HV, β1 and β2 are hyperparameters; Parameter d lane-obs Used to balance the following factors: (1) When N OV i|1≤i≤N Accurately calculate TTC when coexisting i ; (2) With v HV (3) when the curvature of the planned route of the host vehicle HV is high, the possibility of lane change decision is reduced; For d lane-obs The road environment within the range is divided into two parts with the longitudinal axis of the HV as the dividing line: the front calculation layer and the rear calculation layer; The front calculation layer is responsible for calculating the traffic information in front of the main vehicle, including the front, left front and right front; the rear calculation layer is responsible for calculating the traffic information behind the main vehicle, including the rear, left rear and right rear; For front detection, first collect the lane set The speed data v of vehicles traveling in the same direction ov and distance data d obs , representing the collected data as a combined array: in, x lane-marking Indicates the lateral distance between HV and the lane marking; OV state Indicates the existence of a vehicle. It is true when there is a vehicle. stare When d is false, obs It is used to calculate the space occupancy to ensure that the overall safety topology space around the HV remains intact and does not collapse.

4. The method for local motion planning of autonomous driving based on dynamic safety topology structure according to claim 3, characterized in that: In step S2, for lane markings that overlap and have multiple IDs lane Confused driving scenes, a regular hexagonal detection structure is proposed, and d lane-obs Complementary scope; For this complementary structure, the latitude axis is used as a fixed dividing line, and a fixed angle is used to divide the area around the HV into six parts: front, left front, right front, rear, left rear, and right rear; then the surrounding vehicles are partitioned and positioned according to the defined areas.

5. The method for local motion planning of autonomous driving based on dynamic safety topology structure according to claim 1, characterized in that: For lane change and overtaking operations based on local paths, the following conditions must be met at the same time: (1) If there is a global planned path, the lane change operation event based on the global path planning global The priority must be higher than the lane change operation Event based on local path planning local ; If the local lane change operation conflicts with the global path planning, the vehicle will abandon the lane change and switch to the safe following mode. The specific operations are as follows: (2) There needs to be a certain interval between two consecutive lane change operations, the lane change interval time t gap According to v HV Adaptive adjustment, set as follows: where v max is the maximum speed of the host vehicle HV, T is a predefined constant and T is a positive integer; (3) In terms of longitudinal positioning, the distance d between the HV and the preceding vehicle in the same lane longitude Must be below the threshold d turn , as shown below: (d longitude <d turn )∧(TTC(HV,OV longitude )<TTC th )∧(OV longitude is True) Among them, TTC th The setting of TTC is affected by the driving style selected. The more aggressive the driving style, the higher the th The smaller the value of d turn is an adjustable threshold used to determine whether a lane change should be performed; d turn The value of is defined as d turn =(w1-w2Curvature(wpt HV ))·d lane-obs Where 0 <w1<1, Curvature({wpt HV }) represents multiple planned path points near HV {wpt HV }Route curvature estimated by quadratic polynomial; (4) In terms of lateral positioning, the distance between the front and rear vehicles in the target lane is obtained and the collision time TTC is estimated; the lateral distance d latitude and TTC values ​​must meet the preset threshold range; if the vehicle’s OV state If it is False, the distance and conditions are automatically considered to be met, as follows: where bj∈{front,rear},d safe is the safe lane change distance, which is defined as follows: Where 0<β3<β1, β3>1, v desired is the desired speed of the host vehicle HV, is the shortest braking distance.

6. The method for local motion planning of autonomous driving based on dynamic safety topology structure according to claim 5, characterized in that: For lane change operations based on local motion, (d longitude <d turn )∧(TTC(HV,OV longitude )<TTC th )∧(OV longitude is True) For the lane change operation based on the global path, in addition to satisfying the above formula, the safety condition for the vehicle in front is also satisfied, which is described as: (TTC(HV,OV longitude )<TTC th )∨(OV longitude is False) When performing a lane change operation, the desired lane change speed v of the host vehicle HV lc The settings are as follows: s.t.v lc =max(v lc ,v min ) Where 0<σ1<σ2, v min is the minimum speed limit for the host vehicle; When the lane changing conditions of the main vehicle HV are not met, the main vehicle performs lane keeping or following operations; the frequency of lane changes depends on the driving style of the HV, and the more aggressive the driving style, the higher the frequency of lane changes; when it is determined that the main vehicle HV needs to stay in the current lane, its driving speed is automatically adjusted according to the surrounding environmental conditions.

7. The method for local motion planning of autonomous driving based on dynamic safety topology structure according to claim 6, characterized in that: The environmental conditions are divided into three categories: (I) structured roads, i.e., straight roads and curves, (II) unstructured roads, i.e., intersections, and (III) keeping lanes and waiting for lane changes to satisfy global path planning; (III) has a higher priority than (I) and (II); Condition (I): For scenarios where lane change is not necessary, a segmented dynamic speed planning method is used; the dynamic safety constraint is used as the trust range for speed regulation, and the speed control strategy is executed within this range, so that the host vehicle HV can adapt to the surrounding environment. At the same time, the dynamic segmented constraint is used to ensure safety. The speed of the HV The setup is as follows: Among them 0<μ1<1<μ2, 0<γ1<1,γ2>1, γ3v max >v desired ,η≥1,η∝μ2,TTC t represents the time required for the host vehicle HV to collide with the other vehicle OV in the target lane; Condition (II): At an intersection, the speed of the host vehicle HV is adjusted according to the speed of surrounding vehicles. The settings are as follows: Among them, 0<γ4<1, 0<γ5<1; Condition (III): If the host vehicle HV needs to change lanes based on the global path planning, the speed of the HV will be reduced during the waiting period for the lane change condition to be met. The settings are as follows: Among them, 0<μ3<1, 0<μ4<1, 0<γ6<1, 0<γ7<1.

8. The method for local motion planning of autonomous driving based on dynamic safety topology structure according to claim 1, characterized in that: In step S4, the maximum safety range is set, that is, the radius D of the area of ​​interest safe ; For the target lane of HV, if the target lanes are adjacent, when d(HV,OV front ) <D safe When d(HV,OV front ) <D safe And d(HV,OV front ) <d brake When the HV starts braking action; Introducing dynamic parameter d brake , adjusting the boundaries of safe driving behavior, thereby expanding the operating range while maintaining safety; If the rear OV is too close to the HV and may cause a rear-end collision, the HV will start the lane change action. To change lanes to avoid a rear-end collision, the longitudinal positioning conditions must be met: (d longitude <d turn )∧(TTC(HV,OV longitude )<TTC th )∧(OV longitude is True) The potential hazard assessment criteria for HVs from vehicles behind must also be met: where d brake Defined as: in is the braking distance threshold, which is set according to the HV driving style of the host vehicle; According to the OV status setting of other vehicles around, a minimum value is assigned when other vehicles are stationary.

Images