Vehicle trajectory planning method based on interaction intention and related device

Abstract

The application discloses a vehicle trajectory planning method based on interaction intention and related devices. The method comprises the following steps: obtaining state information of a target ego vehicle and obstacles and road environment information; identifying obstacles with interaction risks with the target ego vehicle; performing interaction intention evaluation; if not the first time, based on a historical deduction trajectory, using Bayesian filtering to fuse current observation to update belief probabilities of multiple possible trajectories; if the first time, generating multiple deduction trajectories based on a semantic decision tree, and giving initial belief probabilities through a comprehensive cost function; selecting a high-probability trajectory, introducing time-growing timing noise to the high-probability trajectory to construct a dynamic conflict area; based on the state information of the target ego vehicle, the road environment information, N obstacle deduction trajectories and the corresponding dynamic conflict areas, obtaining an optimal trajectory of the target ego vehicle. The application enhances the stability, accuracy and smoothness of decision planning, and improves driving safety.

Description

Technical Field

[0001] This application relates to the field of autonomous driving technology, specifically to a vehicle trajectory planning method and related apparatus based on interactive intent. Background Technology

[0002] With the gradual implementation and widespread application of unmanned delivery scenarios, the decision-making and planning of autonomous vehicles are facing increasingly severe challenges. In autonomous driving systems, when dealing with situations where the predicted trajectory of an obstacle intersects with the vehicle's collision box, how to assess whether to decelerate or accelerate is a key issue.

[0003] Several strategies are commonly used in the industry to address this type of decision-making problem. One is a safety distance-based strategy, which determines the vehicle's behavioral decision by assessing the distance between the obstacle and the vehicle. Another is a trajectory optimization-based strategy, which uses cost functions (such as the energy cost of acceleration / deceleration, collision risk cost, etc.) to evaluate the cost of different decisions, thereby selecting the optimal behavior.

[0004] However, while distance-based methods are simple to implement, their limitation lies in relying solely on distance thresholds, which can easily lead to decision-making lag (e.g., frequent switching of decisions near threshold boundaries). Trajectory optimization-based strategies, on the other hand, consider more cost factors and perform better in balancing trajectory smoothness and safety. However, these methods rely too heavily on the obstacle's current single predicted trajectory and do not adequately consider the obstacle's potential intentions based on historical behavior and future interactions. Therefore, in scenarios with ambiguous intentions or abrupt behavioral changes, the planned trajectory may jump between consecutive frames, affecting driving comfort and safety.

[0005] Therefore, existing technologies lack a decision-making and planning method that can stably and smoothly process obstacle interaction intentions and generate safe and reasonable vehicle trajectories accordingly. Summary of the Invention

[0006] This application provides a vehicle trajectory planning method and related apparatus based on interactive intent, which can effectively evaluate the interactive intent of obstacles when the obstacle trajectory intersects with the vehicle's collision box, and generate a decision plan for a safe and reasonable vehicle trajectory accordingly, thereby enhancing the stability, accuracy and smoothness of the decision plan and improving driving safety.

[0007] The technical solution is as follows: Firstly, a vehicle trajectory planning method based on interactive intent is provided, including: Acquire the status information of the target vehicle in the current driving scenario, the status information of at least one obstacle, and the road environment information; Based on the state information of the target vehicle, the state information of at least one obstacle, and the road environment information, obstacles that potentially intersect with the collision frame of the target vehicle are identified as obstacles to be evaluated. The interaction intent of the obstacle to be evaluated is evaluated to obtain multiple obstacle deduction trajectories with corresponding belief probabilities; From multiple obstacle deduction trajectories, select N obstacle deduction trajectories whose belief probabilities satisfy a first set threshold, where N is a positive integer; Introduce time-varying temporal noise into each of the N obstacle projection trajectories to construct a dynamic conflict region that reflects the uncertainty of the prediction; Based on the target vehicle's state information, the road environment information, the N obstacle projection trajectories and their corresponding dynamic conflict areas, the optimal planned trajectory of the target vehicle is obtained.

[0008] In one possible implementation, the interaction intent of the obstacle to be evaluated is assessed to obtain multiple obstacle deduction trajectories with corresponding belief probabilities, specifically including: If the obstacle to be evaluated is observed for the first time, then according to the rational driving assumption, multiple semantic behavior sequences of the obstacle to be evaluated are generated, and a physically feasible trajectory is deduced for each semantic behavior sequence; a comprehensive cost function is constructed to evaluate each trajectory, and the comprehensive cost function considers at least safety cost, efficiency cost, comfort cost and navigation cost; based on the evaluation results of the comprehensive cost function, an initial belief probability is determined for each trajectory. If the obstacle to be evaluated is not observed for the first time, then multiple historical projection trajectories corresponding to the obstacle to be evaluated and their historical belief probabilities, recorded at the previous moment, are obtained; based on the latest state information of the obstacle to be evaluated observed at the current moment, the historical belief probabilities of each of the historical projection trajectories are updated using Bayesian filtering to obtain the latest belief probability.

[0009] In one possible implementation, after obtaining the latest belief probability, the method further includes: Determine whether the latest belief probabilities after updating all historical projection trajectories at the current moment are all less than the second set threshold; If so, a replanning is triggered, generating multiple new deduction trajectories for the obstacle to be evaluated, and transferring the belief probability corresponding to the semantic behavior sequence of the original historical deduction trajectory to the newly generated deduction trajectory with the same semantic behavior sequence.

[0010] In one possible implementation, multiple semantic behavior sequences of the obstacle to be evaluated are generated, specifically including: Construct a semantic decision tree, wherein the nodes of the semantic decision tree represent combinations of lateral and vertical actions, wherein lateral actions include detouring to the left, detouring to the right, and maintaining position, and vertical actions include accelerating, decelerating, and maintaining a constant speed. Each node in the semantic decision tree is used to generate a semantic behavior sequence with coherent actions within a predetermined time range, according to the tree structure.

[0011] In one possible implementation, the dynamic conflict region includes an absolute conflict region and a relative conflict region; the absolute conflict region is the geometric intersection region between the planned trajectory of the target vehicle and the deduced trajectory of the obstacle, and the relative conflict region is related to the introduced temporal noise.

[0012] In one possible implementation, obtaining the optimal planned trajectory of the target vehicle based on the state information of the target vehicle, the road environment information, the N obstacle projection trajectories and their corresponding dynamic conflict areas includes: Based on the state information of the target vehicle, the road environment information, the N obstacle projection trajectories, and their corresponding dynamic conflict regions, a strain-based trajectory optimization objective function is constructed. The objective function is then solved based on set constraints to generate the optimal planned trajectory for the target vehicle. The constructed strain-based trajectory optimization objective function is: Where i represents the number of projected trajectories, the The state information of the target vehicle at time k, representing the i-th projected trajectory in response to the obstacle. For a rough explanation, the above The control information for the target vehicle's predicted trajectory for dealing with the obstacle at time k is the i-th path. and Q and R are process weight matrices derived from the dynamic conflict region corresponding to the i-th deduced trajectory of the obstacle, respectively; The objective function must satisfy at least the following constraints: Among them, the For a safety margin distance, the and These are the front and rear boundaries of the conflict area of ​​the target vehicle, respectively.

[0013] Secondly, a vehicle trajectory planning device based on interactive intent is provided, comprising: The acquisition module is used to acquire the status information of the target vehicle, the status information of at least one obstacle, and the road environment information in the current driving scenario; The identification module is used to identify obstacles that potentially intersect with the collision frame of the target vehicle as obstacles to be evaluated, based on the state information of the target vehicle, the state information of at least one obstacle, and the road environment information. The evaluation module is used to evaluate the interaction intent of the obstacle to be evaluated and obtain multiple obstacle inference trajectories with corresponding belief probabilities. The selection module is used to select N obstacle deduction trajectories from multiple obstacle deduction trajectories whose belief probabilities satisfy a first set threshold, where N is a positive integer; A construction module is used to introduce time-varying temporal noise into each of the N obstacle projection trajectories to construct a dynamic conflict region that reflects the uncertainty of prediction. The generation module is used to obtain the optimal planned trajectory of the target vehicle based on the state information of the target vehicle, the road environment information, the N obstacle projection trajectories and their corresponding dynamic conflict areas.

[0014] Thirdly, an electronic device is provided, comprising: At least one processor; and A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the methods described above and any possible implementations.

[0015] Fourthly, a computer program product / computer-readable storage medium is provided. The computer program product includes a computer program that, when executed by a processor, implements the vehicle trajectory planning method based on interactive intent; and / or, The computer-readable storage medium stores computer-executable instructions, which, when executed by a processor, are used to implement the vehicle trajectory planning method based on interactive intent.

[0016] Fifthly, an unmanned delivery vehicle is provided, including the electronic equipment described above.

[0017] The beneficial effects of the technical solution provided in this application include at least the following: As can be seen from the above technical solution, the embodiments of this application can obtain vehicle state information, obstacle state information, and road environment information, identify obstacles that have a risk of intersecting with the collision box of the target vehicle, and perform interaction intent evaluation for the identified obstacles. If the obstacle is appearing for the first time, multiple inferred trajectories are generated based on the constructed multiple semantic behavior sequences, and an initial belief probability is determined for each inferred trajectory using a cost function. If the obstacle is not appearing for the first time, its historical inferred trajectories and belief probability results are obtained, and the belief probability of each inferred trajectory is updated using Bayesian filtering in combination with the observation data of the current frame. Several obstacle inferred trajectories with the highest probability are selected from the multiple obstacle inferred trajectories, and temporal noise that increases over time is introduced to construct a dynamic conflict region. Based on the state information of the target vehicle, the road environment information, the N obstacle inferred trajectories, and their corresponding dynamic conflict regions, the optimal planning trajectory of the target vehicle is obtained. This application can effectively assess the interaction intent of obstacles when they intersect with the collision box of the vehicle, and generate a decision plan for a safe and reasonable vehicle trajectory accordingly, thereby enhancing the stability, accuracy and smoothness of the decision plan and improving driving safety.

[0018] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of this application, nor is it intended to limit the scope of this application. Other features of this application will become readily apparent from the following description. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 This is a schematic diagram illustrating the steps of a vehicle trajectory planning method based on interactive intent provided in an embodiment of this application.

[0021] Figure 2 These are schematic diagrams illustrating two typical scenarios involved in the embodiments of this application.

[0022] Figure 3 This is a schematic diagram of a semantic decision tree provided in one embodiment of this application; Figure 4 This is a schematic diagram of the obstacle prediction trajectory after noise reconstruction provided in one embodiment of this application.

[0023] Figure 5 This is a schematic diagram of the dynamic conflict area formed by the vehicle and other vehicles provided in the embodiments of this application.

[0024] Figure 6 This is a schematic diagram of the vehicle trajectory planning process based on interactive intent provided in the embodiments of this application.

[0025] Figure 7 This is a structural block diagram of a vehicle trajectory planning device based on interactive intent provided in another embodiment of this application.

[0026] Figure 8 This is a block diagram of an electronic device provided in an embodiment of this application. Detailed Implementation

[0027] The following description, in conjunction with the accompanying drawings, illustrates exemplary embodiments of this application, including various details to aid understanding. These embodiments should be considered merely exemplary. Therefore, those skilled in the art will recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of this application. Similarly, for clarity and brevity, descriptions of well-known functions and structures are omitted in the following description.

[0028] Obviously, the described embodiments are only some, not all, of the embodiments in this application. All other embodiments obtained by those skilled in the art based on the embodiments in this application without inventive effort are within the scope of protection of this application.

[0029] It should be noted that the terminal devices involved in the embodiments of this application may include, but are not limited to, smart devices such as mobile phones, personal digital assistants (PDAs), wireless handheld devices, and tablet computers; the display devices may include, but are not limited to, personal computers, televisions, and other devices with display functions.

[0030] Furthermore, the term "and / or" in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. Additionally, the character " / " in this article generally indicates that the preceding and following related objects have an "or" relationship.

[0031] Existing trajectory decision-making and planning schemes based on safe distance rely too heavily on the current single predicted trajectory of obstacles, failing to fully consider the potential intentions of obstacles based on historical behavior and future interactions. Therefore, in scenarios with ambiguous intentions or abrupt behavior changes, the planned trajectory may jump between consecutive frames, affecting driving comfort and safety. In view of this, this application proposes a vehicle trajectory planning scheme based on interaction intention. Its main inventive concept is as follows: acquiring vehicle state information, obstacle state information, and road environment information; identifying obstacles that have a risk of intersecting with the target vehicle's collision box; for the identified obstacles, performing interaction intention evaluation; if the obstacle is appearing for the first time, generating multiple inferred trajectories based on constructed multiple semantic behavior sequences, and using a cost function to determine the initial belief probability for each inferred trajectory; if the obstacle is not appearing for the first time, acquiring its historical inferred trajectories and belief probability results, and combining them with the observation data of the current frame, using Bayesian filtering to update the belief probability of each inferred trajectory. This application selects several obstacle projection trajectories with the highest probabilities from the belief probability results, introduces temporal noise that increases over time, constructs a dynamic conflict region, and obtains the optimal planned trajectory for the target vehicle based on the target vehicle's state information, the road environment information, the N obstacle projection trajectories, and their corresponding dynamic conflict regions. This application can effectively evaluate obstacle interaction intentions based on the intersection of obstacle trajectories and vehicle collision boxes, and generate a safe and reasonable vehicle trajectory decision plan accordingly, enhancing the stability, accuracy, and smoothness of the decision plan and improving driving safety.

[0032] The solutions involved in this application will be described in detail below through specific embodiments.

[0033] Reference Figure 1 The diagram shown illustrates the steps of a vehicle trajectory planning method based on interactive intent, as provided in an embodiment of this application. The method may include the following steps: Step 102: Obtain the status information of the target vehicle, the status information of at least one obstacle, and the road environment information in the current driving scenario.

[0034] In this application, the target vehicle's status information includes at least: vehicle shape, position, speed, and planned path; the obstacle's status information includes at least: obstacle shape, position, and speed; and the road environment information includes at least: other traffic participants, road signs and markings.

[0035] Step 104: Based on the state information of the target vehicle, the state information of at least one obstacle, and the road environment information, identify obstacles that potentially intersect with the collision frame of the target vehicle as obstacles to be evaluated.

[0036] Reference Figure 2The diagram illustrates two typical scenarios involved in the embodiments of this application. First, the target vehicle brakes suddenly when faced with an obstacle, but due to insufficient braking or unstable decision-making, the target vehicle fails to maintain yielding outside the conflict zone. Second, the target vehicle and the obstacle start moving simultaneously. Both scenarios will result in… Figure 2 In the dilemma shown, the target vehicle needs to judge the obstacle's intention to go around or give way, and then make a reasonable decision to respond.

[0037] It should be understood that the specific implementation of step 104 is not limited in this application. Any method in the prior art that can identify obstacles that may intersect with the target vehicle can be used as the specific implementation of this step in this application.

[0038] Step 106: Perform interactive intent assessment on the obstacle to be evaluated to obtain multiple obstacle deduction trajectories with corresponding belief probabilities.

[0039] Optionally, when evaluating the interaction intent of the obstacle to be evaluated and obtaining multiple obstacle deduction trajectories with corresponding belief probabilities, if the obstacle to be evaluated is being observed for the first time, multiple semantic behavior sequences of the obstacle to be evaluated are generated according to the rational driving assumption, and a physically feasible deduction trajectory is deduced for each semantic behavior sequence; a comprehensive cost function is constructed to evaluate each deduction trajectory, the comprehensive cost function considering at least safety cost, efficiency cost, comfort cost, and navigation cost; based on the evaluation result of the comprehensive cost function, an initial belief probability is determined for each deduction trajectory; if the obstacle to be evaluated is not being observed for the first time, multiple historical deduction trajectories corresponding to the obstacle to be evaluated recorded at the previous moment and their historical belief probabilities are obtained; based on the latest state information of the obstacle to be evaluated observed at the current moment, the historical belief probabilities of each of the historical deduction trajectories are updated using Bayesian filtering to obtain the latest belief probability.

[0040] Furthermore, after obtaining the latest belief probability, it can be determined whether the latest belief probability of all historical deduction trajectories updated at the current moment is less than the second set threshold; if so, replanning is triggered, multiple new deduction trajectories are regenerated for the obstacle to be evaluated, and the belief probability corresponding to the semantic behavior sequence of the original historical deduction trajectory is transferred to the newly generated new deduction trajectory with the same semantic behavior sequence.

[0041] In this application, multiple semantic behavior sequences of the obstacle to be evaluated can be generated in the following way: a semantic decision tree is constructed, wherein the nodes of the semantic decision tree represent combinations of lateral and longitudinal actions, wherein lateral actions include detouring to the left, detouring to the right, and maintaining position, and longitudinal actions include accelerating, decelerating, and maintaining a constant speed; and the nodes contained in each branch of the semantic decision tree are used to generate a semantic behavior sequence with coherent actions within a predetermined time range according to the tree structure.

[0042] Specifically, this step mainly includes four parts: determining the set of projected trajectories, inferring initial beliefs, determining continuous belief estimates, and trajectory replanning.

[0043] The determination of the set of projected trajectories can be achieved by following these steps: Step 1: Design the detour route.

[0044] According to the rational driving assumption, when an obstacle is bypassed from in front of the target vehicle, the target vehicle's response process needs to be appropriately considered, and the target vehicle applies maximum braking. Decelerate to stopping distance: Deceleration to a stop time: The obstacle is in front of the target vehicle (see reference). Figure 2 ) Around the boundary, expected lateral position: The obstacle is behind the target vehicle (see reference). Figure 2 Desired lateral position around the boundary: This yields the expected detour position, i.e., the lateral offset position. Then, the predicted trajectory of the obstacle is processed and shifted to both sides to obtain the lateral reference line.

[0045] Step 2, semantic decision-making.

[0046] Considering that in the above interaction scenario, the obstacle will not first move left and then right, or vice versa, in a short period of time, therefore, the following can be formed: Figure 3 The semantic decision tree is shown. Similarly, obstacles do not have frequent acceleration or deceleration actions in the vertical direction; that is, an obstacle performs only one vertical semantic action in a semantic decision sequence. Thus, the set of semantic action sequences can be represented as: A: L, L, L, ..., L M: L, L, L, ..., L D: L, L, L, ..., L A: K, L, L, ..., L M: K, L, L, ..., L D: K, L, L, ..., L The lateral movements are: leftward movement (L), rightward movement (R), and holding (K); the longitudinal movements are: acceleration (A), deceleration (D), and constant speed (M).

[0047] Step 3: Calculate the trajectory.

[0048] Based on the above detour location and the current predicted trajectory of the obstacle, three reference lines can be obtained: proceed along the current predicted trajectory ( ), detour around the target vehicle from the front ( ), and circle around the target vehicle from behind ( To proceed along the current predicted trajectory, the current predicted trajectory needs to be discretized. Assume the distance between the discrete points is... The current predicted trajectory is denoted as: trajectory points Including location ,direction and cumulative mileage The cumulative mileage starts at 0, so the cumulative mileage is: Among them, the starting point of the planned trajectory .

[0049] The predicted trajectory above is reconstructed based on cumulative mileage. Then we have: Define the set of reference points for reconstructing the current predicted trajectory as follows: The reconstruction process is as follows: a) Traverse the accumulated mileage set Get the cumulative mileage query ; b) Traverse the predicted trajectory set Find the point ,satisfy: and If the above problems cannot be solved, the reconstruction process is complete.

[0050] c, Reconstruction point It should be located on a line segment Above, there is: Among them, point direction Reconstruction points Add to the trajectory reconstruction reference line point set Then, jump back to step a. After obtaining the set of reference points for the current trajectory reconstruction, determine the points based on their direction and lateral offset distance. This will give you the set of reference points for the detour around the target vehicle. and the set of reference points for circling around the target vehicle from behind. The process is as follows: After obtaining the lateral reference line, kinematic deduction is performed. The intelligent driver model (IDM) is used longitudinally, and the pure tracking model (PP) is used laterally. The process is as follows: a. Front wheel steering angle Forward exploration distance S: ,in, This represents the vehicle's current speed.

[0051] Find the set of reference points Use the projection point of the forward exploration distance S as the aiming point and obtain its coordinates (x, y). The Euclidean distance from the aiming point is: .

[0052] The vector angle of the current state pointing to the aiming point state : .

[0053] Current front wheel steering angle compensation: ,Will Input the PP model and calculate the front wheel steering angle (steer): ,in, The vehicle's current speed. Here are the vehicle's current position coordinates, and W is the wheelbase.

[0054] b. Forward expected velocity In this application scenario, only the trajectory of the obstacle is simulated, considering its motion properties. Other obstacles besides the one of interest are not considered, thus reducing the time required for environmental modeling. Therefore, the IDM model assumes that the obstacle following the vehicle in front of the first obstacle is infinitely far away and has an infinite velocity. Then, the forward acceleration is calculated based on the IDM model: in, At the current speed, For the desired speed, For expected acceleration (expected velocity and expected acceleration are related to longitudinal semantic actions). To achieve the desired deceleration, This represents the speed difference between the vehicle and the vehicle in front. The forward speed is: ,in, This represents the execution time of the current action.

[0055] c. Constraints Current front wheel steering angle: ,in, The current steering angle of the vehicle's front wheels. Current curvature. Processing control speed. And control the front wheel steering angle: in, For the maximum front wheel steering angle constraint, longitudinal control speed Execution must satisfy maximum jump and acceleration / deceleration constraints, which are applied by bottom-up integration: Expected jerk: Constraints are imposed on the integral: Controlling speed: .

[0056] in, For longitudinal control of acceleration, At the current speed, To control the jump vertically, For the current acceleration, For the maximum longitudinal deceleration, For maximum longitudinal jerk, For the maximum longitudinal deceleration, This represents the maximum longitudinal acceleration. Similarly, the steer function for lateral control of the wheel steering angle must satisfy the maximum jump and acceleration / deceleration constraints, which are applied by integrating from bottom to top: Expected jerk: Constraints are imposed on the integral: Control the front wheel steering angle: ,in, To control lateral acceleration, For the current lateral acceleration, To control the jump laterally, The maximum lateral acceleration, This represents the maximum lateral jerk.

[0057] Considering the maximum front wheel angular velocity constraint of the vehicle: Expected front wheel steering angle: ,in, For the desired angular velocity, This represents the maximum front wheel angular velocity. The desired acceleration is... and expected angular velocity They are as follows: Transitional differential equation: According to the above differential equation, by forward integration, we have: in, For the target state, For the target curvature, The target longitudinal acceleration is defined. Different constraint parameters are set for different types of vehicles; for example, for large trucks, the maximum front wheel angular velocity is... It should be smaller than a regular passenger car.

[0058] Next, semantic labels can be defined for different deduction trajectories. Each semantic label corresponds to a specific semantic action sequence to facilitate subsequent belief inference and belief updating. Labels 1, 2, 3, ... are set to correspond to obstacle deduction trajectories. , , , ……

[0059] Next, we infer the initial beliefs.

[0060] Initial belief inference involves assigning an initial intention to a newly encountered obstacle. This inference requires the assumption that the obstacle is boundedly rational, and a cost function can be designed to evaluate this initial intention. The cost function primarily considers safety, efficiency, comfort, and navigation, and includes: a. security Assuming the obstacle interaction is a target vehicle moving at a constant speed in the current state, the set of trajectory points for this constant speed movement is defined as follows: Traverse obstacles by timestamp to deduce trajectory and If a collision occurs, a larger loss is applied. .

[0061] b. Efficiency Efficiency depends on the longitudinal travel distance of the obstacle vehicle. ,have: in, The adjustment parameter in the efficiency cost function is taken as an empirical value of 0.1. for The cumulative mileage of the last simulation point.

[0062] c. Comfort Since the longitudinal motion design has been considered during the simulation, scenarios of acceleration followed by deceleration or deceleration followed by acceleration have been ruled out. Therefore, comfort primarily considers the acceleration state of the obstacle, which includes: in, The adjustment parameter in the comfort cost function is taken as an empirical value of 0.1. for The acceleration at the k-th projection point.

[0063] d. Navigation Navigation primarily reflects congestion ahead on the driving trajectory. Congestion ahead (other road users / road signs and markings that cannot be crossed / static debris) will affect traffic flow. This perception is often reflected in distance; the closer the distance, the more obvious the impact of congestion. For a projected trajectory with surrounding congestion, the following applies: in, This is an adjustment parameter in the navigation cost function, where d is the distance from the obstacle to the congestion location. The costs are sorted in ascending order, and a threshold range is selected. The cost is less than The belief level for the projected trajectory is set to 0, and the belief probabilities for other trajectories are set according to their costs, as follows: It should be noted that the denominator of the above equation corresponds to... The deduction trajectory does not include those where the belief has been set to 0.

[0064] Then, the continuous belief estimate is determined.

[0065] (1) State The vehicle state consists of position, velocity, and direction of motion. The vehicle state at time t is defined as follows: in, For vehicle location, For vehicle speed, This indicates the direction of vehicle travel.

[0066] (2) Prior distribution Based on the vehicle's past trajectories and cost functions, the prior probabilities of each current trajectory can be calculated. The vehicle's state at current time t is... The known observation data are The prior probability of each trajectory is defined as follows: If the obstacle is a newly appeared obstacle, then If the obstacle is not a newly appeared obstacle, then the recursive update can proceed normally. Furthermore, since the above trajectory is a continuous trajectory derived from derivation, the influence of state transition noise is not considered, and we have: (3) Likelihood function New observational data is generated as the vehicle moves. The likelihood function is calculated based on the observation model, representing the probability of the observed data given a particular state. Assume the observed data is... The vehicle state at time t+1 is Then the likelihood function is expressed as The likelihood function value is related to the degree of deviation between the observed trajectory and the inferred trajectory, and is defined as: in, This is the weight coefficient matrix, with empirical values ​​of [0.5, 0.5, 0.1].

[0067] (4) Posterior distribution Based on Bayes' theorem, combined with the prior distribution and the likelihood function, the posterior probability distribution of the vehicle state is obtained. (Known observation data) Vehicle status The probability is: in, To normalize the constants and ensure that the sum of the posterior probabilities is 1, we obtain the following by integrating all possible states: (5) Recursive update The vehicle state distribution changes continuously with each observation update. At time t+1, the posterior distribution updated from time t is used. As a priori, and through a new round of observations Update the posterior distribution.

[0068] Finally, you can choose to replan the trajectory.

[0069] When there is a significant discrepancy between the actual trajectory of an obstacle and the projected trajectory, replanning should be performed. The criterion for replanning is that, at time t, for All of the following are available: in, This refers to the trajectory point state at time t. Let be the observed state at time t. For the set of all projected trajectories, This represents the probability threshold. In this application, replanning mainly refers to the replanning of the inferred trajectory, and the intention and belief need to be inherited into the new inferred trajectory. Specifically, the intention and belief values ​​should be transmitted according to the semantic tags mentioned above.

[0070] Step 108: Select N obstacle deduction trajectories from the belief estimation results whose belief probabilities satisfy the first set threshold, where N is a positive integer.

[0071] In this application, the first set threshold can be a belief threshold determined based on empirical values.

[0072] Step 110: Introduce time-varying temporal noise into each of the N obstacle projection trajectories to construct a dynamic conflict region that reflects the uncertainty of the prediction.

[0073] The dynamic conflict region includes an absolute conflict region and a relative conflict region; the absolute conflict region is the geometric intersection region of the planned trajectory of the target vehicle and the deduced trajectory of the obstacle, and the relative conflict region is related to the introduced temporal noise.

[0074] In this application, the time-varying temporal noise is represented by the following time-varying Gaussian noise model: Among them, noise variance As time t increases, Let V be the initial variance of the noise. As a growth factor, This represents the upper limit of variance.

[0075] The possible trajectory and interaction intent of an obstacle can be obtained through the obstacle interaction intent evaluation process. In this application, considering the gradually increasing temporal uncertainty of obstacle movement, noise reconstruction is introduced to predict the trajectory, defined as follows: in, Furthermore, we define the process by which prediction uncertainty changes over time as follows: in, It is the initial variance of the noise. This is an adjustment factor representing the rate at which noise increases over time. The obstacle prediction trajectory after noise reconstruction is illustrated below. Figure 4 The diagram shown is a schematic representation of the predicted trajectory noise provided in an embodiment of this application. The obstacle's trajectory follows a Gaussian distribution along the predicted trajectory axis, and the noise rate is... Initial noise variance .

[0076] like Figure 5 The diagram shown is a schematic representation of the dynamic conflict zone formed by the vehicle and another vehicle according to an embodiment of this application. By introducing Gaussian noise, the conflict zone formed between the target vehicle and the obstacle should include both absolute and relative conflict zones, with the relative conflict zone being related to the noise.

[0077] Step 112: Based on the state information of the target vehicle, the road environment information, the N obstacle projection trajectories and their corresponding dynamic conflict areas, obtain the optimal planned trajectory of the target vehicle.

[0078] Optionally, based on the state information of the target vehicle, the road environment information, the N obstacle projection trajectories, and their corresponding dynamic conflict areas, a strain-based trajectory optimization objective function is constructed. The objective function is then solved based on set constraints to generate the optimal planned trajectory for the target vehicle. The constructed strain-based trajectory optimization objective function is: Where i represents the number of projected trajectories, the The state information of the target vehicle at time k, representing the i-th projected trajectory in response to the obstacle. For a rough explanation, the above The control information for the target vehicle's predicted trajectory for dealing with the obstacle at time k is the i-th path. and Q and R are process weight matrices derived from the dynamic conflict region corresponding to the i-th deduced trajectory of the obstacle, respectively; The objective function must satisfy at least the following constraints: Among them, the For a safety margin distance, the and These are the front and rear boundaries of the conflict area of ​​the target vehicle, respectively.

[0079] This step employs strain programming to construct an optimization problem, comprehensively considering both the efficiency and safety of the autonomous delivery vehicle to plan the target vehicle's trajectory. Specifically, the optimization problem can be constructed as follows: Where n is the number of obstacles considered in the predicted trajectory. and The above-mentioned response prediction trajectories are respectively A summary of equality and inequality constraints. Add strain constraints as follows: Specifically, the state transition function is defined as: Where X is the state vector, U is the control vector, and A and B are coefficient matrices, defined as follows: The target vehicle's current status is Assume the multimodal predicted trajectories for obstacles are as follows: in, Let the obstacle state vector be defined as: Generally, the objective function is defined as: in, and The target is the vehicle's trajectory prediction for dealing with obstacles. The response status and actions, For a coarse solution, we can assume that the target vehicle is moving at a constant linear speed.

[0080] Constraint functions include equality constraints and inequality constraints. Equality constraints include initial state constraints and state transition constraints, defined as follows: Inequality constraints mainly refer to collision constraints, which include two types: a. soft constraints, which mainly refer to conflicts at the noise points of the projected trajectory; b. hard constraints, which mainly refer to the absolute conflicts between the target vehicle's planned trajectory and the obstacle's projected trajectory.

[0081] Assume the front and rear boundaries of the obstacle collision zone are located at... and The front and rear boundaries of the target vehicle conflict area are respectively and We define: Hard constraints must satisfy the following conditions: Summarized as follows: in, and This is the weight matrix. For a safety margin distance, an empirical value of 0.1 is used.

[0082] The soft constraint takes into account the noise in the obstacle trajectory prediction, and this effect is applied as a soft constraint to the objective function, resulting in: Variable upper and lower bound constraints are defined as follows: The number of obstacle prediction trajectories is empirically set to 3, meaning that the top 3 probability obstacle intent trajectories obtained from the aforementioned selection (where N is 3, but it can also be set to other values, which are not limited in this application) are considered, and after normalization, we have: This serves as the optimal trajectory for the final decision-making and planning of the target vehicle.

[0083] Figure 6 The diagram shown is a schematic of the vehicle trajectory planning process based on interactive intent provided in an embodiment of this application.

[0084] First, obtain the target vehicle's shape, position, speed, planned path, as well as the shape, position, speed of obstacles, and road environment information.

[0085] Next, identify obstacles that intersect with the collision box of the target vehicle.

[0086] Determine if there is a historical belief estimate. If so, obtain the historical trajectory and corresponding belief estimate, and update the belief estimate based on the current obstacle state. If not, deduce the trajectory and initialize the belief estimate.

[0087] If it is an obstacle observed for the first time, a set of inferred trajectories corresponding to its various semantic behaviors (such as accelerating to the left, decelerating to the right, maintaining, etc.) is generated, and the cost of each trajectory is evaluated based on a preset cost function (taking into account safety, efficiency, comfort, and navigation), and the initial belief probability is assigned according to the cost.

[0088] If this is not the first observation, the historical inferred trajectory set of the obstacle and its belief probability are loaded. Using the new obstacle state (position, velocity, etc.) observed in the current frame, the belief probability of each historical inferred trajectory is updated through Bayesian filtering. If the belief probability of all historical trajectories is lower than the replanning threshold, replanning is triggered, a new inferred trajectory set is generated, and the belief value corresponding to the intent label is inherited.

[0089] Finally, based on the belief estimation results, multiple inferred trajectories with higher probabilities are identified, time-series noise is added and dynamic conflict regions are generated, a strain MPC optimization problem is constructed, and the optimal solution for the target vehicle's decision trajectory is generated.

[0090] Based on the updated belief distribution, the N (e.g., 3) obstacle projection trajectories with the highest probabilities are selected as multimodal prediction inputs for subsequent planning. Considering the accumulation of uncertainty in obstacle motion prediction over time, a time-varying Gaussian noise is added to each selected projection trajectory, with the noise variance increasing over time until it reaches an upper limit. Thus, each projection trajectory is no longer a line, but a probability distribution band that diverges over time. A model predictive control optimization problem is constructed. The state variable in this problem is the vehicle's state, and the control variables are the vehicle's acceleration, etc. The objective function, considering tracking and comfort, introduces the avoidance cost (soft constraint) for multiple generated relative conflict zones and the strict avoidance requirement (hard constraint) for absolute conflict zones. By solving this optimization problem, the optimal trajectory for the target vehicle in the future time domain is obtained, which is both smooth and safe, and allows for flexible responses to various possible intentions of obstacles.

[0091] 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.

[0092] In the above embodiments, 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.

[0093] Figure 7 This paper illustrates a structural block diagram of a vehicle trajectory planning device based on interactive intent, according to an embodiment of this application. Figure 7 As shown. The vehicle trajectory planning 700 based on interactive intent in this embodiment may include an acquisition module 701, an identification module 702, an evaluation module 703, a selection module 704, a construction module 705, and a generation module 706. The acquisition module 701 is used to acquire the state information of the target vehicle, the state information of at least one obstacle, and road environment information in the current driving scenario. The identification module 702 is used to identify obstacles that potentially intersect with the collision frame of the target vehicle as obstacles to be evaluated based on the state information of the target vehicle, the state information of at least one obstacle, and the road environment information. The evaluation module 703 is used to evaluate the interactive intent of the obstacles to be evaluated, obtaining multiple obstacle deduction trajectories with corresponding belief probabilities. The selection module 704 is used to select N obstacle deduction trajectories from the belief estimation results whose belief probabilities satisfy a first set threshold, where N is a positive integer. The construction module 705 is used to introduce time-varying temporal noise into each of the N obstacle deduction trajectories to construct a dynamic conflict region reflecting prediction uncertainty. The generation module 706 is used to obtain the optimal planned trajectory of the target vehicle based on the state information of the target vehicle, the road environment information, the N obstacle projection trajectories and their corresponding dynamic conflict areas.

[0094] It should be noted that some or all of the vehicle trajectory planning device based on interactive intent in this embodiment may be an application located on a local terminal, or it may be a plugin or software development kit (SDK) or other functional unit set in an application located on a local terminal, or it may be a processing engine located on a network-side server, or it may be a distributed system located on the network side, such as a processing engine or distributed system in a network-side autonomous driving platform, etc. This embodiment does not impose any particular limitations on this.

[0095] Optionally, in one possible implementation of this embodiment, when the evaluation module 703 evaluates the interaction intent of the obstacle to be evaluated and obtains multiple obstacle deduction trajectories with corresponding belief probabilities, it is specifically used as follows: if the obstacle to be evaluated is observed for the first time, it generates multiple semantic behavior sequences of the obstacle to be evaluated according to the rational driving assumption, and deduces a physically feasible deduction trajectory for each semantic behavior sequence; it constructs a comprehensive cost function to evaluate each deduction trajectory, wherein the comprehensive cost function considers at least safety cost, efficiency cost, comfort cost, and navigation cost; based on the evaluation result of the comprehensive cost function, it determines the initial belief probability for each deduction trajectory; if the obstacle to be evaluated is not observed for the first time, it obtains multiple historical deduction trajectories and their historical belief probabilities recorded at the previous moment corresponding to the obstacle to be evaluated; based on the latest state information of the obstacle to be evaluated observed at the current moment, it updates the historical belief probabilities of each of the historical deduction trajectories using Bayesian filtering to obtain the latest belief probability.

[0096] Optionally, in one possible implementation of this embodiment, after obtaining the latest belief probability, the evaluation module 703 is further used to determine whether the latest belief probability after updating all historical deduction trajectories at the current moment is less than a second set threshold; if so, replanning is triggered to regenerate multiple new deduction trajectories for the obstacle to be evaluated, and the belief probability corresponding to the semantic behavior sequence of the original historical deduction trajectory is passed to the newly generated new deduction trajectory with the same semantic behavior sequence.

[0097] Optionally, in one possible implementation of this embodiment, when generating multiple semantic behavior sequences of the obstacle to be evaluated, the evaluation module 703 is specifically used to construct a semantic decision tree. The nodes of the semantic decision tree represent combinations of lateral and longitudinal actions, wherein lateral actions include detouring to the left, detouring to the right, and maintaining position, and longitudinal actions include accelerating, decelerating, and maintaining a constant speed. The nodes contained in each branch of the semantic decision tree are used to generate a semantic behavior sequence with coherent actions within a predetermined time range according to the tree structure.

[0098] Optionally, in one possible implementation of this embodiment, the dynamic conflict region includes an absolute conflict region and a relative conflict region; the absolute conflict region is the geometric intersection region of the planned trajectory of the target vehicle and the deduced trajectory of the obstacle, and the relative conflict region is related to the introduced temporal noise.

[0099] Optionally, in one possible implementation of this embodiment, when the generation module 706 obtains the optimal planned trajectory of the target vehicle based on the target vehicle's state information, the road environment information, the N obstacle projection trajectories, and their corresponding dynamic conflict areas, it is specifically used to construct a strain-based trajectory optimization objective function based on the target vehicle's state information, the road environment information, the N obstacle projection trajectories, and their corresponding dynamic conflict areas, and solve the objective function based on set constraints to generate the optimal planned trajectory of the target vehicle. The constructed strain-based trajectory optimization objective function is: Where i represents the number of projected trajectories, the The state information of the target vehicle at time k, representing the i-th projected trajectory in response to the obstacle. For a rough explanation, the above The control information for the target vehicle's predicted trajectory for dealing with the obstacle at time k is the i-th path. and Q and R are process weight matrices derived from the dynamic conflict region corresponding to the i-th deduced trajectory of the obstacle, respectively; The objective function must satisfy at least the following constraints: Among them, the For a safety margin distance, the and These are the front and rear boundaries of the conflict area of ​​the target vehicle, respectively.

[0100] In this embodiment, vehicle state information, obstacle state information, and road environment information can be acquired to identify obstacles that pose a risk of intersecting with the target vehicle's collision box. For the identified obstacles, an interaction intent assessment is performed. If the obstacle is appearing for the first time, multiple deduced trajectories are generated based on various constructed semantic behavior sequences, and an initial belief probability is determined for each deduced trajectory using a cost function. If the obstacle is not appearing for the first time, its historical deduced trajectories and belief probability results are acquired, and the belief probabilities of each deduced trajectory are updated using Bayesian filtering in conjunction with the observation data of the current frame. Several obstacle deduced trajectories with the highest probabilities are selected from the belief probability results, and temporal noise that increases over time is introduced to construct a dynamic conflict region. Based on the target vehicle's state information, the road environment information, the N obstacle deduced trajectories, and their corresponding dynamic conflict regions, the optimal planned trajectory for the target vehicle is obtained. This application can effectively assess the obstacle interaction intent in cases where the obstacle trajectory intersects with the vehicle's collision box, and generate a decision plan for a safe and reasonable vehicle trajectory accordingly, enhancing the stability, accuracy, and smoothness of the decision plan and improving driving safety.

[0101] One embodiment of this application provides a computer program product / computer-readable storage medium. The computer program product includes a computer program that, when executed by a processor, implements the vehicle trajectory planning method based on interactive intent; and / or, the computer-readable storage medium stores computer execution instructions that, when executed by a processor, are used to implement the vehicle trajectory planning method based on interactive intent.

[0102] One embodiment of this application provides an electronic device, which includes a processor and a memory, wherein the memory stores at least one instruction, which is loaded and executed by the processor to implement the vehicle trajectory planning method based on interactive intent as described above.

[0103] One embodiment of this application provides an unmanned delivery vehicle, including the electronic equipment described above. Specifically, the unmanned delivery vehicle can be a Level 2 or higher vehicle.

[0104] The collection, storage, use, processing, transmission, provision, and disclosure of user personal information involved in the technical solution of this application all comply with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0105] Figure 8 A schematic block diagram of an example electronic device 800 that can be used to implement embodiments of this application is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device may also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the application described and / or claimed herein.

[0106] like Figure 8 As shown, the electronic device 800 includes a computing unit 801, which can perform various appropriate actions and processes based on a computer program stored in a read-only memory (ROM) 802 or a computer program loaded from a storage unit 808 into a random access memory (RAM) 803. The RAM 803 may also store various programs and data required for the operation of the electronic device 800. The computing unit 801, ROM 802, and RAM 803 are interconnected via a bus 804. An input / output (I / O) interface 805 is also connected to the bus 804.

[0107] Multiple components in electronic device 800 are connected to I / O interface 805, including: input unit 806, such as keyboard, mouse, etc.; output unit 807, such as various types of displays, speakers, etc.; storage unit 808, such as disk, optical disk, etc.; and communication unit 809, such as network card, modem, wireless transceiver, etc. Communication unit 809 allows electronic device 800 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0108] The computing unit 801 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the computing unit 801 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. The computing unit 801 performs the various methods and processes described above, such as the vehicle trajectory planning method based on interactive intent. For example, in some embodiments, the vehicle trajectory planning method based on interactive intent may be implemented as a computer software program tangibly contained in a machine-readable medium, such as storage unit 808. In some embodiments, part or all of the computer program may be loaded and / or installed on the electronic device 800 via ROM 802 and / or communication unit 809. When the computer program is loaded into RAM 803 and executed by the computing unit 801, one or more steps of the vehicle trajectory planning method based on interactive intent described above may be performed. Alternatively, in other embodiments, the computing unit 801 may be configured, by any other suitable means (e.g., by means of firmware), to perform a vehicle trajectory planning method based on interactive intent.

[0109] Various implementations of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), systems-on-a-chip (SoCs), complex programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various implementations may include: implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transferring data and instructions to the storage system, at least one input device, and at least one output device.

[0110] The program code used to implement the methods of this application may be written in any combination of one or more programming languages. This program code may be provided to a processor or controller of a general-purpose computer, special-purpose computer, or other programmable data processing device, such that when executed by the processor or controller, the functions / operations specified in the flowcharts and / or block diagrams are implemented. The program code may be executed entirely on a machine, partially on a machine, as a standalone software package partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0111] In the context of this application, a machine-readable medium can be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. Machine-readable media can be, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0112] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device for displaying information to the user (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor); and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the computer. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0113] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as a data server), or computing systems that include middleware components (e.g., an application server), or computing systems that include frontend components (e.g., a user computer with a graphical user interface or web browser through which a user can interact with implementations of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., a communication network). Examples of communication networks include local area networks (LANs), wide area networks (WANs), and the Internet.

[0114] Computer systems can include clients and servers. Clients and servers are generally located far apart and typically interact via communication networks. Client-server relationships are created by computer programs running on the respective computers and having a client-server relationship with each other. Servers can be cloud servers, servers in distributed systems, or servers incorporating blockchain technology.

[0115] It should be understood that the various forms of processes shown above can be used to rearrange, add, or delete steps. For example, the steps described in this application can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution disclosed in this application can be achieved, and this is not limited herein.

[0116] The specific embodiments described above do not constitute a limitation on the scope of protection of this application. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A vehicle trajectory planning method based on interactive intent, characterized in that, include: Acquire the status information of the target vehicle in the current driving scenario, the status information of at least one obstacle, and the road environment information; Based on the state information of the target vehicle, the state information of at least one obstacle, and the road environment information, obstacles that potentially intersect with the collision frame of the target vehicle are identified as obstacles to be evaluated. The interaction intent of the obstacle to be evaluated is evaluated to obtain multiple obstacle deduction trajectories with corresponding belief probabilities; From multiple obstacle deduction trajectories, select N obstacle deduction trajectories whose belief probabilities satisfy a first set threshold, where N is a positive integer; Introduce time-varying temporal noise into each of the N obstacle projection trajectories to construct a dynamic conflict region that reflects the uncertainty of the prediction; Based on the target vehicle's state information, the road environment information, the N obstacle projection trajectories and their corresponding dynamic conflict areas, the optimal planned trajectory of the target vehicle is obtained.

2. The method as described in claim 1, characterized in that, The interaction intent of the obstacle to be evaluated is assessed, resulting in multiple obstacle deduction trajectories with corresponding belief probabilities, specifically including: If the obstacle to be evaluated is observed for the first time, then according to the rational driving assumption, multiple semantic behavior sequences of the obstacle to be evaluated are generated, and a physically feasible trajectory is deduced for each semantic behavior sequence; a comprehensive cost function is constructed to evaluate each trajectory, and the comprehensive cost function considers at least safety cost, efficiency cost, comfort cost and navigation cost; based on the evaluation results of the comprehensive cost function, an initial belief probability is determined for each trajectory. If the obstacle to be evaluated is not observed for the first time, then multiple historical projection trajectories corresponding to the obstacle to be evaluated and their historical belief probabilities, recorded at the previous moment, are obtained; based on the latest state information of the obstacle to be evaluated observed at the current moment, the historical belief probabilities of each of the historical projection trajectories are updated using Bayesian filtering to obtain the latest belief probability.

3. The method as described in claim 2, characterized in that, After obtaining the latest belief probability, the method further includes: Determine whether the latest belief probabilities after updating all historical projection trajectories at the current moment are all less than the second set threshold; If so, a replanning is triggered, generating multiple new deduction trajectories for the obstacle to be evaluated, and transferring the belief probability corresponding to the semantic behavior sequence of the original historical deduction trajectory to the newly generated deduction trajectory with the same semantic behavior sequence.

4. The method as described in claim 2, characterized in that, Generate multiple semantic behavior sequences of the obstacle to be evaluated, specifically including: Construct a semantic decision tree, wherein the nodes of the semantic decision tree represent combinations of lateral and vertical actions, wherein lateral actions include detouring to the left, detouring to the right, and maintaining position, and vertical actions include accelerating, decelerating, and maintaining a constant speed. Each node in the semantic decision tree is used to generate a semantic behavior sequence with coherent actions within a predetermined time range, according to the tree structure.

5. The method as described in claim 1, characterized in that, The dynamic conflict region includes an absolute conflict region and a relative conflict region; the absolute conflict region is the geometric intersection region of the planned trajectory of the target vehicle and the deduced trajectory of the obstacle, and the relative conflict region is related to the introduced temporal noise.

6. The method as described in claim 1, characterized in that, The process of obtaining the optimal planned trajectory of the target vehicle based on the target vehicle's state information, the road environment information, the N obstacle projection trajectories and their corresponding dynamic conflict areas includes: Based on the state information of the target vehicle, the road environment information, the N obstacle projection trajectories, and their corresponding dynamic conflict regions, a strain-based trajectory optimization objective function is constructed. The objective function is then solved based on set constraints to generate the optimal planned trajectory for the target vehicle. The constructed strain-based trajectory optimization objective function is: Where i represents the number of projected trajectories, the The state information of the target vehicle at time k, representing the i-th projected trajectory in response to the obstacle. For a rough explanation, the above The control information for the target vehicle's predicted trajectory for dealing with the obstacle at time k is the i-th path. and Q and R are process weight matrices derived from the dynamic conflict region corresponding to the i-th deduced trajectory of the obstacle, respectively; The objective function must satisfy at least the following constraints: Among them, the For a safety margin distance, the and These are the front and rear boundaries of the conflict area of ​​the target vehicle, respectively.

7. A vehicle trajectory planning device based on interactive intent, characterized in that, include: The acquisition module is used to acquire the status information of the target vehicle, the status information of at least one obstacle, and the road environment information in the current driving scenario; The identification module is used to identify obstacles that potentially intersect with the collision frame of the target vehicle as obstacles to be evaluated, based on the state information of the target vehicle, the state information of at least one obstacle, and the road environment information. The evaluation module is used to evaluate the interaction intent of the obstacle to be evaluated and obtain multiple obstacle inference trajectories with corresponding belief probabilities. The selection module is used to select N obstacle deduction trajectories from multiple obstacle deduction trajectories whose belief probabilities satisfy a first set threshold, where N is a positive integer; A construction module is used to introduce time-varying temporal noise into each of the N obstacle projection trajectories to construct a dynamic conflict region that reflects the uncertainty of prediction. The generation module is used to obtain the optimal planned trajectory of the target vehicle based on the state information of the target vehicle, the road environment information, the N obstacle projection trajectories and their corresponding dynamic conflict areas.

8. An electronic device, comprising: At least one processor; as well as A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the method according to any one of claims 1-6.

9. A computer program product / computer-readable storage medium, characterized in that, The computer program product includes a computer program that, when executed by a processor, implements the method according to any one of claims 1-6; and / or, The computer-readable storage medium stores computer-executable instructions, which, when executed by a processor, are used to implement the methods described in claims 1-6.

10. An unmanned delivery vehicle, including the electronic equipment as described in claim 8.

Images