A data processing method for a planning module

Abstract

An embodiment of the present invention relates to a data processing method for a planning module. The method includes: the planning module receives a multi-objective prediction trajectory set output by an upstream prediction module at the start time of the current working cycle; and obtains the prediction tree and policy tree of the previous working cycle as the corresponding first prediction tree and first policy tree; constructs the prediction tree of the current working cycle according to the multi-objective prediction trajectory set to generate the corresponding second prediction tree; identifies similar nodes of the prediction trees for the first and second prediction trees to obtain a plurality of first similar node pairs; constructs the second policy tree corresponding to the current working cycle according to the topological structure of the second prediction tree, and initializes the second policy tree according to the first policy tree and all the first similar node pairs. Through the present invention, not only can the repetitive calculation amount of each working cycle of the planning module be reduced, but also the redundant occupation of storage resources by the planning module can be reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and in particular to a data processing method for a planning module. Background Art

[0002] The upstream modules of the planning module of a vehicle autonomous driving system include a prediction module and a perception module. The prediction module predicts multiple possible motion trajectories of each target within a specified future time period based on the pose information of each obstacle target around the host vehicle obtained by the perception module, and sends the set of multi-target prediction trajectories obtained by the prediction to the planning module; the planning module then plans and calculates the driving trajectories of the host vehicle in various possible multi-target motion scenarios corresponding to the output of the prediction module within the specified future time period. When performing the planning calculation, the traditional approach of the planning module is to perform a full-time period trajectory calculation on multiple possible driving trajectories of the host vehicle within the specified future time period based on the set of multi-target prediction trajectories received each time at the start of each working cycle, so as to obtain a set of planning trajectories composed of multiple planning trajectories; in addition, the planning module also stores the set of multi-target prediction trajectories and the set of planning trajectories in a traditional data sequence storage manner.

[0003] However, in practical applications, we found that this traditional approach has the following problems: 1) The duration of the planning trajectory output by the planning module, that is, the duration of the specified future time period, is much greater than the working cycle duration of the planning module. Therefore, a large part of the set of planning trajectories output in adjacent two working cycles is basically overlapped. Obviously, the traditional approach of recalculating the trajectories within the specified future time period in each working cycle has a large amount of redundant calculation problems; 2) There will inevitably be overlapping trajectory segments between the prediction trajectories of any target in the set of multi-target prediction trajectories, and there will also inevitably be overlapping trajectory segments between the planning trajectories in the set of planning trajectories. Obviously, storing the trajectories in the two sets in the traditional data sequence storage manner will cause waste of storage resources brought by the overlapping trajectory segments. Summary of the Invention

[0004] The purpose of the present invention is to provide a data processing method, an electronic device and a computer-readable storage medium for a planning module in view of the deficiencies of the prior art. Through the present invention, not only can the redundant calculation amount of each working cycle of the planning module be reduced, but also the redundant occupation of storage resources by the planning module can be reduced.

[0005] To achieve the above object, in the first aspect of the embodiments of the present invention, a data processing method for a planning module is provided, and the method includes:

[0006] The planning module receives a multi-objective prediction trajectory set output by the upstream prediction module at the start of the current working cycle; and obtains the prediction tree and policy tree of the previous working cycle as the corresponding first prediction tree and first policy tree;

[0007] Construct a corresponding second prediction tree for the prediction tree of the current working cycle according to the multi-objective prediction trajectory set;

[0008] Identify similar nodes of the first and second prediction trees to obtain multiple pairs of first similar nodes;

[0009] Construct a second policy tree corresponding to the current working cycle according to the topological structure of the second prediction tree, and initialize the second policy tree according to the first policy tree and all the pairs of first similar nodes.

[0010] Preferably, the multi-objective prediction trajectory set includes multiple first target trajectory groups \(R_i\), where \(i\) is the target index, \(1\leq i\leq n\), and \(n\) is the number of targets; each of the first target trajectory groups \(R_i\) includes multiple first prediction trajectories \(p_{ij}\), where \(j\) is the prediction trajectory index of target \(i\), \(1\leq j\leq m\), and \(m\) is the number of prediction trajectories of target \(i\); in each of the first target trajectory groups \(R_i\), there is a section of trajectory overlap between any two of the first prediction trajectories \(p_{ij}\) starting from the trajectory start time; the start time of the multi-objective prediction trajectory set is denoted as \(t_0\) and the end time is denoted as \(t\), then the trajectory period of each of the first prediction trajectories \(p_{ij}\) in the set is \([t_0,t]\); i , where \(i\) is the target index, \(1\leq i\leq n\), and \(n\) is the number of targets; each of the first target trajectory groups \(R_i\) i includes multiple first prediction trajectories \(p_{ij}\) i,j , where \(j\) is the prediction trajectory index of target \(i\), \(1\leq j\leq m\) i and \(m\) i is the number of prediction trajectories of target \(i\); in each of the first target trajectory groups \(R_i\), i for any two of the first prediction trajectories \(p_{ij}\) i,j there is a section of trajectory overlap starting from the trajectory start time; the start time of the multi-objective prediction trajectory set is denoted as \(t_0\) and the end time is denoted as \(t\) end , then the trajectory period of each of the first prediction trajectories \(p_{ij}\) in the set is \([t_0,t]\); i,j end ;

[0011] The first prediction tree is composed of a first root node and multiple first leaf nodes. The first root node is the parent node of the subordinate leaf nodes connected to it. Each of the first leaf nodes is the child node of the superior root node or superior leaf node connected to it and the parent node of the subordinate leaf nodes connected to it; the first root node or each of the first leaf nodes each includes a set (first node period, first node trajectory set, and first node depth); the first node trajectory set includes the target number \(n\) of first target sub-trajectories \(\tau_{i}\), and the first target sub-trajectory \(\tau_{i}\) is the predicted trajectory segment of the corresponding target \(i\) within the current first node period; the first node depth is the total number of nodes on the node path from the first root node to the current node; in the first prediction tree, for any pair of parent and child nodes, the two first target sub-trajectories \(\tau_{i}\) corresponding to the same target \(i\) 1,i , the first target sub-trajectory \(\tau_{i}\) 1,i is the predicted trajectory segment of the corresponding target \(i\) within the current first node period; the first node depth is the total number of nodes on the node path from the first root node to the current node; in the first prediction tree, for any pair of parent and child nodes, the two first target sub-trajectories \(\tau_{i}\) corresponding to the same target \(i\) 1,i ​It should be a continuous trajectory relationship;

[0012] The first policy tree is composed of a second root node and multiple second leaf nodes. The second root node is the parent node of the subordinate leaf nodes connected to it. Each of the second leaf nodes is the child node of the superior root node or superior leaf node connected to it and is the parent node of the subordinate leaf nodes connected to it. Each of the second root node or the second leaf nodes includes a set (second node time period, second node trajectory set, and second node depth). The second node trajectory set is composed of a first planned trajectory τ c1 which is a planned trajectory segment of the vehicle itself within the current second node time period. The second node depth is the total number of nodes on the node path from the second root node to the current node. In the first policy tree, any pair of parent and child nodes' two first planned trajectories τ c1 should be in a continuous trajectory relationship. The topological structure of the first policy tree corresponding to the same working cycle is consistent with that of the first prediction tree: the root nodes correspond to each other, the leaf nodes correspond to each other, and the connection relationships of parent and child nodes at all levels are consistent. c1 Preferably, constructing a corresponding second prediction tree for the prediction tree of the current working cycle according to the multi-objective prediction trajectory set specifically includes:

[0013] Extracting the end time of the overlapping trajectory between any two first prediction trajectories p

[0014] in each first target trajectory group R i in the multi-objective prediction trajectory set starting from the trajectory start time to generate corresponding first branch times; and performing duplicate time filtering on all the obtained first branch times, and sorting the remaining first branch times after filtering in chronological order to obtain a first branch time sequence. The first branch time sequence includes multiple first branch times t i,j , where a is the first branch time index, 1 ≤ a ≤ A, and A is the number of first branch times; a Using each of the first branch times t

[0015] as a segmentation node, dividing the time period [t0, t a corresponding to the multi-objective prediction trajectory set into A + 1 sub-time periods S end ; k is the sub-time period index, 1 ≤ k ≤ A + 1. The sub-time period S k is denoted as the time period [t0, t k＝1 corresponding to the root time period. The sub-time period S a＝1 corresponds to the time period [t k＝2 , t a＝1 , and the sub-time period S a＝2 corresponds to the time period [t k＝3 corresponds to the time period [t a＝2,t a＝3 , and so on, for the sub - time period S k＝A+1 corresponding to the time period [t a＝A ,t end ;

[0016] Create a corresponding third root node for the root node in the second prediction tree corresponding to the root time period [t0, t a＝1 according to the multi - target prediction trajectory set; the third root node includes a group (third - node time period, third - node trajectory set, and third - node depth);

[0017] Using the third root node with the third - node depth of 1 as the current parent node and the sub - time period S k＝2 [t a＝1 ,t a＝2 as the current sub - time period, create corresponding multiple third - leaf nodes for each lower - level child node connected to the current parent node in the second prediction tree and corresponding to the current sub - time period according to the multi - target prediction trajectory set; the third - leaf node includes a group (third - node time period, third - node trajectory set, and third - node depth);

[0018] Using each third - leaf node with the third - node depth of 2 as the current parent node and the sub - time period S k＝3 [t a＝2 ,t a＝3 as the current sub - time period, create corresponding multiple third - leaf nodes for each lower - level child node connected to the current parent node in the second prediction tree and corresponding to the current sub - time period according to the multi - target prediction trajectory set; and so on, until all the third - leaf nodes corresponding to the sub - time period S k＝A+1 [t a＝A ,t end are created.

[0019] Preferably, creating a corresponding third root node for the root node in the second prediction tree corresponding to the root time period [t0, t a＝1 according to the multi - target prediction trajectory set specifically includes:

[0020] Create a root node in the second prediction tree as the corresponding third root node;

[0021] Set the third - node time period of the third root node as the root time period [t0, t a＝1 ;

[0022] Set the third - node depth of the third root node as 1;

[0023] Among each of the first target trajectory groups R in the multi-target predicted trajectory set i for any first predicted trajectory p within the group i,j extract the predicted trajectory segment within the root time period [t0, t a＝1 as the second target sub-trajectory τ of the corresponding target i within the root time period [t0, t a＝1 ; and the second target sub-trajectories τ of all targets 2,i constitute the third node trajectory set of the third root node 2,i .

[0024] Preferably, creating the respective lower-level child nodes corresponding to the current sub-time period and connected to the current parent node in the second prediction tree according to the multi-target predicted trajectory set to generate corresponding multiple third leaf nodes specifically includes:

[0025] Denote the second target sub-trajectory τ corresponding to each target i in the current parent node as the previous time period trajectory pr of each target i 2,i ; i ;

[0026] Among each of the first target trajectory groups R in the multi-target predicted trajectory set i extract the predicted trajectory segments of each first predicted trajectory p within the group i,j in the current sub-time period to generate corresponding first segments, and form a first segment set from the obtained multiple first segments; and delete the first segments in the first segment set that do not have a continuous trajectory relationship with the corresponding previous time period trajectory pr i ; then delete the first segments with redundant segment trajectories in the first segment set; then count the remaining number of first segments in the first segment set to generate a corresponding first quantity d i ; and respectively use the remaining first segments as a target branch trajectory ps of the corresponding target i in the current first target trajectory group R i in the current sub-time period, and the target branch trajectories ps i,h constitute a corresponding sub-time period target trajectory set O i,h ; each sub-time period target trajectory set O i corresponds to a target i, and the sub-time period target trajectory set O i includes the target branch trajectories ps i with the first quantity d i , where h is the branch trajectory index of target i, 1 ≤ h ≤ d i,h ; i ;

[0027] According to the obtained first quantity d of the target quantity n i Calculate the number of child nodes d s , d s = d i＝1 ×d i＝2 …×d i …×d i＝n ; And in the second prediction tree, create child nodes with the number of child nodes d s under the current parent node, connect the current parent node with each child node, and regard each child node as a corresponding third leaf node;

[0028] From each of the target branch trajectory sets O i in the sub-period, randomly select one of the target branch trajectories ps i,h for combination to obtain a corresponding sub-period multi-target trajectory set X z ; z is the index of the sub-period multi-target trajectory set, 1 ≤ z ≤ d s ; The sub-period multi-target trajectory set X z includes the target branch trajectory ps of the target quantity n i,h , and the combination relationships of the target branch trajectories in each sub-period multi-target trajectory set X z are not repeated;

[0029] Use the obtained sub-period multi-target trajectory set X of the number of child nodes d s to set the third node trajectory sets of the third leaf nodes with the number of child nodes d created this time one by one; z s s Set the third node time periods of the third leaf nodes with the number of child nodes d created this time to the current sub-period;

[0030] Add 1 to the third node depth of the current parent node to obtain the corresponding child node depth, and set the third node depths of the third leaf nodes with the number of child nodes d created this time to the child node depth. s 1,i

[0031] Preferably, the prediction tree similar node recognition for the first and second prediction trees to obtain multiple first similar node pairs specifically includes: s

[0033] ​​​In the second prediction tree, extract the third node period of the third root node as the current root node period, and extract the third node trajectory set of the third root node as the current root node trajectory set; and in the first prediction tree, mark the first root node or the first leaf node whose first node period contains or is equal to the current root node period as the corresponding first node, so as to obtain one or more of the first nodes; and calculate the set trajectory similarity between the first node trajectory set of each of the first nodes and the current root node trajectory set to obtain one or more similarity calculation results, and take the maximum value among them as the corresponding first maximum similarity; if the first maximum similarity is greater than the preset similarity threshold, then form a first similar node pair by the third root node and the first root node or the first leaf node corresponding to the first maximum similarity.

[0034] In the first prediction tree, take the branch prediction tree structure under the first root node or the first leaf node corresponding to the first maximum similarity as the corresponding first sub-prediction tree.

[0035] Traverse each of the third leaf nodes of the second prediction tree one by one; during the traversal, mark the currently traversed third leaf node as the current leaf node, and extract the third node period of the current leaf node as the current leaf node period, and extract the third node trajectory set of the current leaf node as the current leaf node trajectory set; and in the first sub-prediction tree, mark the first leaf node whose first node period is equal to the current leaf node period as the corresponding second node, so as to obtain one or more of the second nodes; and calculate the set trajectory similarity between the first node trajectory set of each of the second nodes and the current leaf node trajectory set to obtain one or more similarity calculation results, and take the maximum value among them as the corresponding second maximum similarity; if the second maximum similarity is greater than the similarity threshold, then form a first similar node pair by the third leaf node corresponding to the current leaf node and the first leaf node corresponding to the second maximum similarity.

[0036] Preferably, constructing the second policy tree corresponding to the current working cycle according to the topological structure of the second prediction tree, and initializing the second policy tree according to the first policy tree and all the first similar node pairs specifically includes:

[0037] Generate the tree structure of the second policy tree by copying the tree structure of the second prediction tree; the second policy tree includes a fourth root node and multiple fourth leaf nodes; the fourth root node is the parent node of the subordinate leaf nodes connected thereto, and each of the fourth leaf nodes is the child node of the superior root node or superior leaf node connected thereto and is the parent node of the subordinate leaf nodes connected thereto; each of the fourth root node or the fourth leaf nodes includes a set (fourth node time period, fourth node trajectory set, and fourth node depth); the topological structure of the second policy tree is consistent with that of the second prediction tree corresponding to the same working cycle: the root nodes correspond to each other, the leaf nodes correspond to each other, and the connection relationships of the parent and child nodes at all levels are consistent;

[0038] Poll all the first similar node pairs, and denote the currently polled first similar node pair as the current similar node pair; and in the current similar node pair, denote the node corresponding to the first prediction tree as the first prediction tree node, and the node corresponding to the second prediction tree as the second prediction tree node; and based on the topological correspondence between the first prediction tree and the first policy tree, denote the second root node or the second leaf node in the first policy tree corresponding to the first prediction tree node as the first policy tree node; and based on the topological correspondence between the second prediction tree and the second policy tree, denote the second root node or the second leaf node in the second policy tree corresponding to the second prediction tree node as the second policy tree node; and initialize the fourth node time period of the second policy tree node to the second node time period of the first policy tree node, initialize the fourth node trajectory set of the second policy tree node to the second node trajectory set of the first policy tree node, and initialize the fourth node depth of the second policy tree node to the second node depth of the first policy tree node.

[0039] A second aspect of an embodiment of the present invention provides an electronic device, including: a memory, a processor, and a transceiver;

[0040] The processor is used to be coupled with the memory, read and execute the instructions in the memory to implement the method steps described in the first aspect above;

[0041] The transceiver is coupled with the processor, and the processor controls the transceiver to perform message sending and receiving.

[0042] A third aspect of an embodiment of the present invention provides a computer-readable storage medium, and the computer-readable storage medium stores computer instructions, and when the computer instructions are executed by a computer, the computer is caused to execute the instructions of the method described in the first aspect above.

[0043] An embodiment of the present invention provides a data processing method, an electronic device, and a computer-readable storage medium for a planning module. First, at the start time of each working cycle, a pair of tree-shaped data objects (a prediction tree and a policy tree) with the same topological structure are constructed for data storage. The prediction tree is used to store a multi-objective prediction trajectory set, and the policy tree is used to store a planned trajectory set. Node time periods and node trajectory sets are assigned to the root nodes and each leaf node of the prediction tree and the policy tree, and parent-child node relationships are established between upper and lower nodes in the prediction tree and the policy tree. In this way, only one parent node can store the overlapping trajectory segments of multiple trajectories corresponding to multiple child nodes. Secondly, on the premise that the topological structures of the prediction tree and the policy tree are the same in the same cycle, based on the similar nodes of the prediction tree in the previous working cycle and the current working cycle, the similar nodes of the policy tree in the two working cycles before and after are located, and the similar nodes of the policy tree in the current working cycle are initialized using the policy tree in the previous working cycle. In this way, the computational workload of the planning module in the current working cycle can be reduced by data reuse. Through the present invention, the repeated calculation amount of each working cycle of the planning module is reduced, and the redundant occupation of storage resources by the planning module is reduced. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 FIG. is a schematic diagram of a data processing method for a planning module provided in Embodiment 1 of the present invention;

[0045] Figure 2a FIG. is a schematic diagram of a set of prediction tree and policy tree provided in Embodiment 1 of the present invention;

[0046] Figure 2b FIG. is a schematic diagram of the predicted trajectories of vehicles 1 and 2 provided in Embodiment 1 of the present invention;

[0047] Figure 2c FIG. is a schematic diagram of a second prediction tree provided in Embodiment 1 of the present invention;

[0048] Figure 3 FIG. is a schematic diagram of the structure of an electronic device provided in Embodiment 2 of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0049] In order to make the objectives, technical solutions, and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0050] Embodiment 1 of the present invention provides a data processing method for a planning module, as Figure 1As shown in the schematic diagram of the data processing method of a planning module provided in the first embodiment of the present invention, the method mainly includes the following steps:

[0051] Step 1, the planning module receives the multi-objective prediction trajectory set output by the upstream prediction module at the start time of the current working cycle; and obtains the prediction tree and policy tree of the previous working cycle as the corresponding first prediction tree and first policy tree.

[0052] Among them, the first prediction tree is composed of a first root node and multiple first leaf nodes. The first root node is the parent node of the subordinate leaf nodes connected thereto, and each first leaf node is the child node of the superior root node or superior leaf node connected thereto and the parent node of the subordinate leaf nodes connected thereto; the first root node or each first leaf node each includes a set (first node time period, first node trajectory set, and first node depth); the first node trajectory set includes the first target sub-trajectories τ of the target number n 1,i , the first target sub-trajectory τ 1,i is the predicted trajectory segment of the corresponding target i within the current first node time period; the first node depth is the total number of nodes on the node path from the first root node to the current node; in the first prediction tree, any pair of parent and child nodes corresponding to the same target i's two first target sub-trajectories τ 1,i should be in a continuous trajectory relationship;

[0053] The first policy tree is composed of a second root node and multiple second leaf nodes. The second root node is the parent node of the subordinate leaf nodes connected thereto, and each second leaf node is the child node of the superior root node or superior leaf node connected thereto and the parent node of the subordinate leaf nodes connected thereto; the second root node or each second leaf node each includes a set (second node time period, second node trajectory set, and second node depth); the second node trajectory set is composed of a first planned trajectory τ c1 constitutes, and the first planned trajectory τ c1 is the planned trajectory segment of the vehicle itself within the current second node time period; the second node depth is the total number of nodes on the node path from the second root node to the current node; in the first policy tree, any pair of parent and child nodes' two first planned trajectories τ c1 should be in a continuous trajectory relationship; the topological structure of the first policy tree and the first prediction tree corresponding to the same working cycle is kept consistent: the root nodes correspond to each other, the leaf nodes correspond to each other, and the connection relationships of each level of parent and child nodes are kept consistent.

[0054] Here, in the embodiments of the present invention, a pair of tree-shaped data objects (prediction tree and policy tree) with the same topological structure will be constructed in each working cycle for data storage. The prediction tree is used to store the multi-objective prediction trajectory set, and the policy tree is used to store the planned trajectory set. The first prediction tree and the first policy tree are the prediction tree and policy tree generated in the previous working cycle.

[0055] In the embodiments of the present invention, the structures of any pair of prediction trees and policy trees are basically the same: they are both composed of a root node and multiple leaf nodes; each node (root node or leaf node) is the parent node of its subordinate nodes (leaf nodes), and each node (leaf node) is the child node of its superior node (root node or leaf node); each node corresponds to a group (node time period, node trajectory set, and node depth); the node time periods of each node (root node or leaf node) are used to identify the trajectory time period corresponding to the current node, and the node depth of each node (root node or leaf node) is the total number of nodes on the node path from the root node to the current node (root node or leaf node).

[0056] The difference between the prediction tree and the policy tree in the embodiments of the present invention is that: the node trajectory set of each node (root node or leaf node) of the prediction tree is composed of a prediction trajectory segment of multiple targets within the current node time period, that is to say, the node trajectory set of each node (root node or leaf node) of the prediction tree actually identifies a multi-target motion scenario that may occur within the current node time period; while the node trajectory set of each node (root node or leaf node) of the policy tree only includes a planned trajectory segment of the host vehicle within the current node time period, that is to say, the node trajectory set of each node (root node or leaf node) of the policy tree actually identifies a host vehicle motion scenario that may occur within the current node time period.

[0057] Since the host vehicle motion scenario and the multi-target motion scenario are in one-to-one correspondence, the topological structures of the prediction tree and the policy tree in the embodiments of the present invention are the same; in addition, there is temporal continuity between the two multi-target motion scenarios corresponding to the parent and child nodes in the prediction tree, so the target sub-trajectories of the two time periods corresponding to the same target i in the two node trajectory sets of the parent and child nodes should be in a continuous trajectory relationship; similarly, there is temporal continuity between the two host vehicle motion scenarios corresponding to the parent and child nodes in the policy tree, so the planned trajectories of the two time periods in the two node trajectory sets of the parent and child nodes should be in a continuous trajectory relationship.

[0058] The following takes Figure 2a a set of schematic diagrams of the prediction tree and the policy tree provided in Embodiment 1 of the present invention as an example to further illustrate the tree structures of the prediction tree and the policy tree.

[0059] Figure 2aAmong them, the node depth of the root node N1 of the prediction tree is 1; taking the root node N1 as the parent node, there are two child nodes, namely leaf nodes N11 and N12, and the node depths of leaf nodes N11 and N12 are both 2; taking leaf node N11 as the parent node, there are two child nodes, namely leaf nodes N111 and N112, taking leaf node N12 as the parent node, there are two child nodes, namely leaf nodes N121 and N122, and the node depths of leaf nodes N111, N112, N121, and N122 are all 3;

[0060] Suppose the node time period of the root node N1 is [ts0, ts1], then the node trajectory set of the root node N1 identifies 1 multi-target motion scenario in the time period [ts0, ts1], and this node trajectory set is {τ1, τ2... τ i ... τ n} which is composed of the target prediction trajectories τ i of n targets;

[0061] The two child nodes (leaf nodes N11 and N12) of the root node N1 actually have 2 possible scenario changes at time ts1, and thus 2 multi-target motion scenarios are generated; the node time periods of leaf nodes N11 and N12 are the same, set as [ts1, ts2]; the 2 node trajectory sets of leaf nodes N11 and N12 identify 2 multi-target motion scenarios in the time period [ts1, ts2]; because these 2 multi-target motion scenarios are changed from the multi-target motion scenario of their parent node (root node N1), they have temporal continuity before and after. That is to say, in the two node trajectory sets of leaf node N11 (or leaf node N12) and root node N1, the two segments of target prediction trajectories τ i corresponding to the same target i before and after are continuous, and this continuity can be verified through the end trajectory pose of the previous segment of target prediction trajectory τ i and the start trajectory pose of the next segment of target prediction trajectory τ i as well as the target motion state relationship between the front and rear trajectories;

[0062] The two child nodes (leaf nodes N111 and N112) of leaf node N11 actually have 2 possible scenario changes at time ts2, and thus 2 multi-target motion scenarios are generated; the node time periods of leaf nodes N111 and N112 are the same, set as [ts2, ts3], and the 2 node trajectory sets of leaf nodes N111 and N112 identify 2 multi-target motion scenarios in the time period [ts2, ts3]; because these 2 multi-target motion scenarios are changed from the multi-target motion scenario of their parent node (leaf node N11), they have temporal continuity before and after. That is to say, in the two node trajectory sets of leaf node N111 (or leaf node N112) and leaf node N11, the two segments of target prediction trajectories τ iis continuous;

[0063] The two child nodes (leaf nodes N121 and N122) of leaf node N12 actually represent two possible scenario changes at time ts2, resulting in two multi-target motion scenarios. The node time periods of leaf nodes N121 and N122 are the same as those of leaf nodes N111 and N112 with a node depth of 3, i.e., [ts2, ts3]. The two node trajectory sets of leaf nodes N121 and N122 identify two multi-target motion scenarios within the time period [ts2, ts3]. Since these two multi-target motion scenarios are evolved from the multi-target motion scenario of their parent node (leaf node N12), they are temporally continuous. That is to say, in the two node trajectory sets of leaf node N121 (or leaf node N122) and leaf node N12, for the same target i, the two consecutive target prediction trajectories τ i are continuous.

[0064] Figure 2a In, the tree topology structure of the policy tree is exactly the same as that of the prediction tree; the node depth of the root node N2 of the policy tree is 1; with the root node N2 as the parent node, there are two child nodes, namely leaf nodes N21 and N22, and the node depths of leaf nodes N21 and N22 are both 2; with leaf node N21 as the parent node, there are two child nodes, namely leaf nodes N211 and N212, and with leaf node N22 as the parent node, there are two child nodes, namely leaf nodes N221 and N222, and the node depths of leaf nodes N211, N212, N221, and N222 are all 3;

[0065] Let the node time period of the root node N2 be [ts0, ts1 ’ , then the node trajectory set of the root node N2 identifies one ego-vehicle motion scenario within the time period [ts0, ts1 ’ , and this node trajectory set consists of a planned trajectory τ c of the ego-vehicle;

[0066] The two child nodes (leaf nodes N21 and N22) of the root node N2 actually represent two possible scenario changes at time ts1 ’ , resulting in two ego-vehicle motion scenarios. Here, ts1 ’ should not be earlier than ts1 of the prediction tree; the node time periods of leaf nodes N21 and N22 are the same, set as [ts1 ’ , ts ’ 2]; the two node trajectory sets of leaf nodes N21 and N22 identify two ego-vehicle motion scenarios within the time period [ts1 ’ , ts ’The self-vehicle motion scenarios within [2]; Since these two self-vehicle motion scenarios are changed from the self-vehicle motion scenario of their parent node (root node N2), they have temporal continuity before and after. That is to say, in the two-node trajectory sets of leaf node N21 (or leaf node N22) and root node N2, the two successive planned trajectories τ c are continuous. This continuity can be verified by the end trajectory pose of the previous planned trajectory τ c , the start trajectory pose of the subsequent planned trajectory τ c , and the relationship of the self-vehicle motion state between the front and rear trajectories;

[0067] The two child nodes (leaf nodes N211, N212) of leaf node N21 actually have two possible scenario changes at time ts2 ’ , and the two resulting self-vehicle motion scenarios. Here, ts2 ’ should not be earlier than ts2 of the prediction tree; The node time periods of leaf nodes N211, N212 are the same and set to [ts ’ 2, ts3]; The two-node trajectory sets of leaf nodes N211, N212 identify two self-vehicle motion scenarios within the time period [ts ’ 2, ts3]; Since these two self-vehicle motion scenarios are changed from the self-vehicle motion scenario of their parent node (leaf node N21), they have temporal continuity before and after. That is to say, in the two-node trajectory sets of leaf node N211 (or leaf node N212) and leaf node N21, the two successive planned trajectories τ c are continuous;

[0068] The two child nodes (leaf nodes N221, N222) of leaf node N22 actually have two possible scenario changes at time ts2 ’ , and the two resulting self-vehicle motion scenarios; The node time periods of leaf nodes N221, N222 are the same as those of leaf nodes N211, N212 with node depth 3, i.e., [ts2 ’ , ts3]; The two-node trajectory sets of leaf nodes N221, N222 identify two self-vehicle motion scenarios within the time period [ts2 ’ , ts3]; Since these two self-vehicle motion scenarios are changed from the self-vehicle motion scenario of their parent node (leaf node N22), they have temporal continuity before and after. That is to say, in the two-node trajectory sets of leaf node N221 (or leaf node N222) and leaf node N22, the two successive planned trajectories τ c are continuous.

[0069] Here, the tree structure characteristics of the first prediction tree and the first policy tree obtained in the current step 1, as well as the second prediction tree and the second policy tree involved in the subsequent steps, are all the same as the above example Figure 2aThe tree - shaped structure features of the prediction tree and the strategy tree mentioned are the same. The only differences lie in the specific settings of the number of nodes at each level and the specific settings of the node depth, node time period, and node trajectory set of each node.

[0070] The multi - target prediction trajectory set includes multiple first - target trajectory groups R i , where i is the target index, 1 ≤ i ≤ n, and n is the number of targets; each first - target trajectory group R i includes multiple first - prediction trajectories p i,j , where j is the prediction - trajectory index of target i, 1 ≤ j ≤ m i and m i is the number of prediction trajectories of target i; in each first - target trajectory group R i , for any two first - prediction trajectories p i,j , there is a section of trajectory overlap starting from the trajectory start time; the start time of the multi - target prediction trajectory set is denoted as t0 and the end time is denoted as t end , then the trajectory time period of each first - prediction trajectory p i,j in the set is [t0, t end .

[0071] Here, the multi - target prediction trajectory set is the complete set of various possible motion trajectories of all obstacles (targets) in the traffic environment where the host vehicle is located in the future time period [t0, t end ; each first - target trajectory group R i corresponds to a target i, and each first - prediction trajectory p i,j in the trajectory group corresponds to a possible motion trajectory of this target i in the future time period [t0, t end ; in the embodiments of the present invention, the time points of various possible trajectory switches (mutations) of each target in the future time period [t0, t end are all after the start time point t0. Therefore, for the multiple first - prediction trajectories p i,j corresponding to the same target i, there must be a section of overlapping trajectory at their head positions. So, for any two first - prediction trajectories p i in each first - target trajectory group R i,j , there is an overlapping trajectory starting from the starting position.

[0072] Taking Figure 2b as an example of the prediction trajectory schematic diagrams of vehicles 1 and 2 provided in Embodiment 1 of the present invention, there are two obstacle targets, vehicle 1 and vehicle 2, in front of the host vehicle during driving. The upstream prediction module predicts that vehicle 1 has two possible motions (going straight, turning right) at time t1 in the future time period [t0, t end , and in the future time period [t0, t endAt time t2 (t1 < t2) in [0], vehicle 2 has two possible movements (going straight or turning right); then, the target number n = 2, and the number of trajectories m for target i = 1 i＝1 = 2, and the number of trajectories m for target i = 2 i＝2 = 2, that is, the multi-target predicted trajectory set includes two first target trajectory groups R1 and R2. R1 has two first predicted trajectories p 1,1 、p 1,2 , and R2 has two first predicted trajectories p 2,1 、p 2,2 ; In R1, the predicted trajectories of p 1,1 、p 1,2 in the segment [t0, t1] are coincident, and the predicted trajectories of p 2,1 、p 2,2 in R2 in the segment [t0, t2] are coincident.

[0073] Step 2, construct a corresponding second prediction tree for the prediction tree of the current working cycle according to the multi-target predicted trajectory set;

[0074] Specifically, it includes: Step 21, in the multi-target predicted trajectory set, for any two first predicted trajectories p i in each first target trajectory group R i,j , extract the end time of the coincident trajectory from the starting time of the trajectory to generate the corresponding first branch time; and perform duplicate time filtering on all the obtained first branch times, and sort the remaining first branch times after filtering in chronological order to obtain the first branch time sequence;

[0075] Among them, the first branch time sequence includes multiple first branch times t a , a is the first branch time index, 1 ≤ a ≤ A, and A is the number of first branch times;

[0076] Here, any trajectory involved in the embodiments of the present invention is composed of multiple trajectory points arranged in chronological order, and each trajectory point corresponds to a pose and time; the coincident trajectory is actually a completely identical predicted trajectory segment between two first predicted trajectories p i,j starting from the starting time of the trajectory, i.e., t0, and the end time of the coincident trajectory is the time of the last trajectory point on this completely identical predicted trajectory segment; because each first target trajectory group R i corresponds to a target i, then by extracting the end time of the coincident trajectory between any two first predicted trajectories p i in each first target trajectory group R i,j , all the trajectory branch times of target i can be obtained, and the number of trajectory branch times of target i is actually the number of target i in the future time period [t0, t endThe possible number of movements in it; after obtaining all the first branch times of all the targets, there may be multiple first branch times with the same time inside (such as the trajectory branches of two different targets occurring at the same time). At this time, the redundant first branch times need to be deleted through duplicate time filtering; in this way, the finally obtained first branch time sequence can include all the time points that will generate trajectory branches in the multi-target prediction trajectory set;

[0077] For example, as Figure 2b shown, there are two obstacle targets, vehicle 1 and vehicle 2, in front of the ego vehicle. The multi-target prediction trajectory set includes 2 first target trajectory groups R1 and R2. Among them, R1 includes 2 first prediction trajectories p 1,1 , p 1,2 , and R2 includes 2 first prediction trajectories p 2,1 , p 2,2 ; the predicted trajectories of p 1,1 , p 1,2 in R1 are coincident in the [t0, t1] section, and the predicted trajectories of p 2,1 , p 2,2 in R2 are coincident in the [t0, t2] section, and t1 < t2;

[0078] Then, the first branch time obtained by extracting the end time of the coincident trajectory of p 1,1 , p 1,2 in the first target trajectory group R1 is t1, and the first branch time obtained by extracting the end time of the coincident trajectory of p 2,1 , p 2,2 in the first target trajectory group R2 is t2, so as to obtain 2 first branch times t1 and t2; because t1 ≠ t2, so when filtering the duplicate times of t1 and t2, neither of them will be deleted, and the remaining first branch times after filtering are still composed of t1 and t2; and because t1 < t2, the sorted first branch time sequence is {t a＝1 = t1, t a＝2 = t2}, A = 2;

[0079] Step 22, using each first branch time t a as a segmentation node, divide the time period [t0, t end corresponding to the multi-target prediction trajectory set into A + 1 sub-time periods S k ;

[0080] Among them, k is the sub-time period index, 1 ≤ k ≤ A + 1, and the sub-time period S k＝1 is recorded as the time period corresponding to the root time period [t0, t a＝1 , and the sub-time period S k＝2 corresponds to the time period [t a＝1 , ta＝2 , sub - time period S k＝3 corresponding time period [t a＝2 , t a＝3 , and so on, sub - time period S k＝A+1 corresponding time period [t a＝A , t end ;

[0081] For example, the first branch time series is {t1, t2}, and A = 2; then, A + 1 = 3 sub - time periods S can be obtained k : S k＝1 [t0, t1], S k＝2 [t1, t2], S k＝3 [t2, t end ;

[0082] After obtaining A + 1 sub - time periods S k , the embodiments of the present invention can construct a second prediction tree with a maximum node depth of A + 1 through the subsequent steps 23 - 24. For example, A + 1 = 3, which means that the maximum node depth of the second prediction tree constructed this time is 3, that is, the node level of the second prediction tree constructed this time is 3 - level;

[0083] Step 23: Create a corresponding third root node for the root node in the second prediction tree corresponding to the root time period [t0, t a＝1 according to the multi - objective prediction trajectory set;

[0084] Among them, the third root node includes a group (the third node time period, the third node trajectory set, and the third node depth);

[0085] Specifically, it includes: Step 231: Create a root node in the second prediction tree as the corresponding third root node;

[0086] Here, as Figure 2c shown in the schematic diagram of the second prediction tree provided by Embodiment 1 of the present invention, the created third root node is the third root node P1;

[0087] Step 232: Set the third node time period of the third root node as the root time period [t0, t a＝1 ;

[0088] For example, it is known that the sub - time periods S k corresponding to the current multi - objective prediction trajectory set are: S k＝1 [t0, t1] (root time period), S k＝2 [t1, t2], S k＝3 [t2, t end ; then Figure 2c the third node time period of the third root node P1 in is the root time period [t0, t1];

[0089] Step 233, set the third node depth of the third root node to 1;

[0090] Here, there is only 1 node, i.e., the third root node itself, on the node path from the third root node to the third root node itself. Therefore, the third node depth of the third root node should be set to 1;

[0091] Step 234, in each first target trajectory group R i in the multi-target prediction trajectory set, extract the prediction trajectory segment of any first prediction trajectory p i,j in the root time period [t0, t a＝1 as the second target sub-trajectory τ a＝1 of the corresponding target i in the root time period [t0, t 2,i ; and form the third node trajectory set of the third root node from the second target sub-trajectories τ 2,i of all targets;

[0092] Here, because no target has had a trajectory branch within the root time period [t0, t a＝1 , that is, all first prediction trajectories p i in each first target trajectory group R i,j in the multi-target prediction trajectory set are the same within the root time period [t0, t a＝1 ; then, when determining the prediction trajectory segment of each target i in the root time period [t0, t a＝1 , which is also the second target sub-trajectory τ 2,i , only need to select any first prediction trajectory p i from the first target trajectory group R i,j corresponding to each target i and extract the prediction trajectory segment for the time period [t0, t a＝1 ;

[0093] For example, as Figure 2b shown, there are two obstacle targets, vehicle 1 and vehicle 2, in front of the ego vehicle's travel, that is, target i = 1 is vehicle 1, target i = 2 is vehicle 2. The corresponding multi-target prediction trajectory set includes 2 first target trajectory groups R1 and R2. R1 includes 2 first prediction trajectories p 1,1 and p 1,2 , R2 includes 2 first prediction trajectories p 2,1 and p 2,2 . The prediction trajectories of p 1,1 and p 1,2 in R1 within the [t0, t1] segment are coincident. The prediction trajectories of p 2,1 and p 2,2 in R2 within the [t0, t2] segment are coincident. Since t1 < t2, naturally, p2,1 , p 2,2 The predicted trajectories in the [t0, t1] segment also coincide;

[0094] Then, the second target sub-trajectory τ of target 1 in the root time period [t0, t a＝1 is 2,1 p 1,1 and p 1,2 in the [t0, t1] segment of the coincident predicted trajectory segmentation. The τ of target 2 in the root time period [t0, t a＝1 is 2,2 p 2,1 and p 2,2 in the [t0, t1] segment of the coincident predicted trajectory segmentation. Figure 2c The third node trajectory set of the third root node P1 in 2,1 τ 2,2 in the above [t0, t1] segment;

[0095] Step 24: Using the third root node with the third node depth of 1 as the current parent node and the sub-time period S k＝2 [t a＝1 , t a＝2 as the current sub-time period, create corresponding multiple third leaf nodes for each lower-level sub-node connected to the current parent node and corresponding to the current sub-time period in the second prediction tree according to the multi-target predicted trajectory set;

[0096] Among them, the third leaf node includes a group (third node time period, third node trajectory set, and third node depth);

[0097] Here, after completing the creation of the third root node, create the lower-level third leaf nodes with the node depth of 2 under the third root node and corresponding to the sub-time period S k＝2 [t a＝1 , t a＝2 ;

[0098] Specifically, it includes: Step 241: Denote the second target sub-trajectory τ of each target i in the current parent node as the previous time period trajectory pr 2,i of each target i; i ;

[0099] Taking Figure 2b , 2c as an example, the current parent node is the third root node P1. Since the second target sub-trajectory τ of target 1 in the third root node P1 is 2,1 which is the predicted trajectory segment of vehicle 1 from t0 to t1 in Figure 2b , so the previous time period trajectory pr1 of target 1 is the predicted trajectory segment of vehicle 1 in the time period [t0, t1]; Since the second target sub-trajectory τ of target 2 in the third root node P1 is​​2,2 For Figure 2b the predicted trajectory segment of vehicle 2 from t0 to t1, so the previous time segment trajectory pr2 of target 2 is the predicted trajectory segment of vehicle 2 in the time segment [t0, t1];

[0100] Step 242, in each first target trajectory group R i in the multi-target predicted trajectory set, extract the predicted trajectory segments of each first predicted trajectory p i,j in the current sub-time segment to generate corresponding first segments, and form a first segment set from the obtained multiple first segments; and delete the first segments in the first segment set that do not have a continuous trajectory relationship with the corresponding previous time segment trajectory pr i ; then delete the first segments with redundant segment trajectories in the first segment set; then count the remaining number of first segments in the first segment set to generate a corresponding first quantity d i ; and use each of the remaining first segments as a target branch trajectory ps i of the corresponding target i in the current sub-time segment i,h , and form a corresponding sub-time segment target trajectory set O i,h from all the target branch trajectories ps i ;

[0101] Among them, each sub-time segment target trajectory set O i corresponds to a target i, and the sub-time segment target trajectory set O i includes a first quantity d i of target branch trajectories ps i,h , h is the branch trajectory index of target i, 1 ≤ h ≤ d i ;

[0102] Taking Figure 2b and 2c as an example, the multi-target predicted trajectory set corresponding to Figure 2b includes 2 first target trajectory groups R1 and R2. R1 includes 2 first predicted trajectories p 1,1 and p 1,2 , and R2 includes 2 first predicted trajectories p 2,1 and p 2,2 ;

[0103] Then, in the first target trajectory group R1, extract the predicted trajectory segments of each first predicted trajectory p 1,1 and p 1,2 in the current sub-time segment [t1, t2] to obtain a first segment set 1 composed of 2 first segments p 1,1 [t1, t2] and p 1,2 [t1, t2], p 1,1[t1,t2] is the predicted trajectory segment of target 1 (vehicle 1) from time t1 to t2 along the motion direction of the first predicted trajectory p 1,1 ; p 1,2 [t1,t2] is the predicted trajectory segment of target 1 (vehicle 1) from time t1 to t2 along the motion direction of the first predicted trajectory p 1,2 ; p in the first segment set 1 1,1 [t1,t2], p 1,2 [t1,t2] are both continuous with the previous segment trajectory pr1 of target 1, so both are retained; p in the first segment set 1 1,1 [t1,t2], p 1,2 The segment trajectories of [t1,t2], p 1,1 [t1,t2], p 1,2 [t1,t2] do not overlap with each other, so both are retained; then by counting the remaining quantity of the first segments in the first segment set 1, the corresponding first quantity d1 = 2 can be obtained; then the remaining first segments p 1,1 [t1,t2], p 1,2 [t1,t2] are used as the 2 target branch trajectories ps of target 1 (vehicle 1) corresponding to the current first target trajectory group R1 in the current sub - time period [t1,t2] 1,1 , ps 1,2 , and the sub - time period target trajectory set O1{ps 1,1 , ps 1,2} is formed;

[0104] Similarly, in the first target trajectory group R2, a first segment set 2 composed of 2 first segments p 2,1 [t1,t2], p 2,2 [t1,t2] can be obtained, p 2,1 [t1,t2] is the predicted trajectory segment of target 2 (vehicle 2) from time t1 to t2 along the motion direction of the first predicted trajectory p 2,1 ; p 2,2 [t1,t2] is the predicted trajectory segment of target 2 (vehicle 2) from time t1 to t2 along the motion direction of the first predicted trajectory p 2,2 ; p in the first segment set 2 2,1 [t1,t2], p 2,2 [t1,t2] are both continuous with the previous segment trajectory pr2 of target 2, so both are retained; for p in the first segment set 2 2,1 [t1,t2], p 2,2 [t1,t2], the segment trajectories overlap, then only one of them is retained and the other is regarded as a first segment with redundant segment trajectory and is deleted. Suppose p 2,1If it is [t1, t2], then only p remains in the first segment set 2,1 [t1, t2]; Then, by counting the remaining quantity of the first segment in the first segment set 2, the corresponding first quantity d2 = 1 can be obtained; Then, the remaining first segment p 2,1 [t1, t2] is used as 1 target branch trajectory ps of the target 2 (vehicle 2) corresponding to the current first target trajectory group R2 in the current sub-period [t1, t2] 2,1 , and the target branch trajectory ps 2,1 forms the corresponding sub-period target trajectory set O2{ps 2,1};

[0105] Step 243, according to the obtained first quantity d of the target quantity n i Calculate the number of child nodes d s , d s = d i＝1 ×d i＝2 …×d i …×d i＝n ; And in the second prediction tree, create child nodes with the number of child nodes d s under the current parent node, connect the current parent node with each child node, and regard each child node as a corresponding third leaf node;

[0106] Taking Figure 2b , 2c as an example, the target quantity n = 2, and 2 first quantities d1 = 2, d2 = 1 are obtained. Then, the number of child nodes d s = d1×d2 = 2×1 = 2. Since the current parent node is the third root node P1, that is, create child nodes with the number of child nodes d s = 2 third leaf nodes, namely the third leaf nodes P11 and P12, and connect the third root node P1 with the third leaf node P11 and the third root node P1 with the third leaf node P12 respectively;

[0107] Step 244, from each sub-period target trajectory set O i , select any one target branch trajectory ps i,h for combination to obtain a corresponding sub-period multi-target trajectory set X z ;

[0108] Among them, z is the sub-period multi-target trajectory set index, 1 ≤ z ≤ d s ; The sub-period multi-target trajectory set X z includes the target branch trajectories ps of the target quantity n i,h , and the target branch trajectory combination relationships of each sub-period multi-target trajectory set X z are not repeated;

[0109] Taking Figure 2b and 2c as an example, the sub-period target trajectory set O1{ps 1,1 , ps 1,2} corresponding to the target 1 (vehicle 1) is obtained from step 242, and the sub-period target trajectory set O2{ps 2,1} corresponding to the target 2 (vehicle 2); Selecting any one of the target branch trajectories from O1{ps 1,1 , ps 1,2} and O2{ps 2,1} can obtain 2 sub-period multi-target trajectory sets X1{ps 1,1 , ps 2,1} and X2{ps 1,2 , ps 2,1};

[0110] Step 245, using the obtained number d s of sub-period multi-target trajectory sets X z , the third node trajectory sets of the third leaf nodes with the number d s created this time are set one by one;

[0111] Taking Figure 2b and 2c as an example, after obtaining 2 sub-period multi-target trajectory sets X1{ps 1,1 , ps 2,1} and X2{ps 1,2 , ps 2,1}, the third node trajectory set of the third leaf node P11 is set using the sub-period multi-target trajectory set X1{ps 1,1 , ps 2,1}. The τ 2,1 in the third node trajectory set corresponds to the ps 1,1 in X1{ps 2,1}, and the τ 1,1 corresponds to the ps 2,2 in X1{ps 1,1 , ps 2,1}; The third node trajectory set of the third leaf node P12 is set using the sub-period multi-target trajectory set X2{ps 2,1}, and the τ 1,2 in the third node trajectory set corresponds to the ps 2,1 in X2{ps 2,1}, and the τ 1,2 corresponds to the ps 2,1 in X2{ps 1,2 , ps 2,2}, and the τ 1,2 corresponds to the ps 2,1 in X2{ps2,1 ;

[0112] It is not difficult to see from this that Figure 2c the third leaf nodes P11 and P12 in are actually two multi-target motion scenarios generated at time t1 due to two possible motions (going straight and turning right) of target 1 (vehicle 1);

[0113] Step 246, set the third node time periods of the third leaf nodes with the number of child nodes d s created this time to the current sub-time period;

[0114] As Figure 2c shown, if the number of child nodes d s created this time is 2, and the third leaf nodes are P11 and P12, then the current step is to set the third node time periods of the third leaf nodes P11 and P12 to the current sub-time period [t1, t2];

[0115] Step 247, add 1 to the third node depth of the current parent node to obtain the corresponding child node depth, and set the third node depths of the third leaf nodes with the number of child nodes d s created this time to the child node depth;

[0116] As Figure 2c shown, if the number of child nodes d s created this time is 2, and the third leaf nodes are P11 and P12, and the current parent node is the third root node P1, and the third node depth of the third root node P1 is 1, then the child node depth = 1 + 1 = 2. Then, the current step is to set the third node depths of the third leaf nodes P11 and P12 to 2;

[0117] Step 25, use each third leaf node with a third node depth of 2 as the current parent node, use the sub-time period S k＝3 [t a＝2 , t a＝3 as the current sub-time period, and create multiple corresponding third leaf nodes for each lower-level child node connected to the current parent node in the second prediction tree according to the multi-target prediction trajectory set; and so on, until all the third leaf nodes corresponding to the sub-time period S k＝A+1 [t a＝A , t end are created;

[0118] Here, after completing the creation of the third leaf nodes with a node depth of 2, use a processing step similar to that in Step 24 to process the nodes with a node depth of 3 under each third leaf node with a node depth of 2 and corresponding to the sub-time period S k＝3 [t a＝2 , t a＝3Create the lower-level third leaf nodes; and so on until all the last-level third leaf nodes corresponding to the sub-period S k＝A+1 [t a＝A ,t end are all created.

[0119] Take Figure 2b 、 2c as an example. When the current parent node is the third leaf node P11, since the second target sub-trajectory τ 2,1 of target 1 in the third leaf node P11 is Figure 2b the predicted trajectory segment of vehicle 1 along the first predicted trajectory p 1,1 from t1 to t2, so the previous-period trajectory pr1 of target 1 is p 1,1 [t1,t2]; since the second target sub-trajectory τ 2,2 of target 2 in the third leaf node P11 is Figure 2b the predicted trajectory segment of vehicle 2 from t1 to t2, so the previous-period trajectory pr2 of target 2 is the predicted trajectory segment of vehicle 2 in the period [t1,t2];

[0120] Extract the predicted trajectory segments of each first predicted trajectory p 1,1 、p 1,2 in the current sub-period [t2,t end in the first target trajectory group R1 of the multi-target predicted trajectory set, and a first segment set 1 composed of 2 first segments p 1,1 [t2,t end 、p 1,2 [t2,t end can be obtained. p 1,1 [t2,t end is the predicted trajectory segment from time t2 to t 1,1 along the movement direction of the first predicted trajectory p end , and p 1,2 [t2,t end is the predicted trajectory segment from time t2 to t 1,2 along the movement direction of the first predicted trajectory p end ; since the previous-period trajectory pr1 of target 1 this time is p 1,1 [t1,t2], only p 1,1 [t2,t end in the first segment set 1 is continuous with p 1,1 [t1,t2], so delete p 1,2 [t2,t end in the first segment set 1 and only keep p 1,1 [t2,t end ; since there is only one p left in the first segment set 11,1 [t2, t end , there are no duplicate redundant segments, so no redundant segment deletion is performed; then, by counting the remaining number of the first segments in the first segment set 1, the corresponding first quantity d1 = 1 can be obtained; then, the remaining first segment p 1,1 [t2, t end is used as the target 1 (vehicle 1) corresponding to the current first target trajectory group R1 in the current sub-period [t2, t end of 1 target branch trajectory ps 1,1 , and the corresponding sub-period target trajectory set is O1{ps 1,1};

[0121] Similarly, in the first target trajectory group R2, it can be obtained that the first segment set 2 consists of 2 first segments p 2,1 [t2, t end , p 2,2 [t2, t end . p 2,1 [t2, t end is the predicted trajectory segment of target 2 (vehicle 2) from time t2 to t 2,1 according to the motion direction of the first predicted trajectory p end . p 2,2 [t2, t end is the predicted trajectory segment of target 2 (vehicle 2) from time t2 to t 2,2 according to the motion direction of the first predicted trajectory p end . Since the previous period trajectory pr2 of target 2 this time is the predicted trajectory segment of vehicle 2 in the period [t1, t2], and p 2,1 [t2, t end , p 2,2 [t2, t end in the first segment set 2 are both continuous with the previous period trajectory pr2, so both of them are retained; p 2,1 [t2, t end , p 2,2 [t2, t end in the first segment set 2 do not overlap with each other, so both of them are retained; then, by counting the remaining number of the first segments in the first segment set 2, the corresponding first quantity d2 = 2 can be obtained; then, the remaining first segments p 2,1 [t2, t end , p 2,2 [t2, t end are used as the 2 target branch trajectories ps end and ps 2,1 of the target 2 (vehicle 2) corresponding to the current first target trajectory group R2 in the current sub-period [t2, t 2,2, the corresponding sub - period target trajectory set is O2{ps 2,1 , ps 2,2};

[0122] After obtaining the first quantities d1 = 1, d2 = 2, the number of child nodes d for this time can be calculated s = d1×d2 = 1×2 = 2. Since the current parent node is the third leaf node P11, that is, create d s = 2 third leaf nodes, namely the third leaf nodes P111, P112, and connect the third leaf node P11 with the third leaf node P111 and the third leaf node P11 with the third leaf node P112 respectively;

[0123] After completing the creation of the empty nodes of the third leaf nodes P111, P112, select one target branch trajectory from the sub - period target trajectory sets O1{ps 1,1}, O2{ps 2,1 , ps 2,2} respectively for combination, and 2 sub - period multi - target trajectory sets X1{ps 1,1 , ps 2,1}, X2{ps 1,1 , ps 2,2} can be obtained; and use the sub - period multi - target trajectory set X1{ps 1,1 , ps 2,1} to set the third - node trajectory set of the third leaf node P111. The τ 2,1 in the third - node trajectory set corresponds to the ps 1,1 , ps 2,1 in X1{ps 1,1 , τ 2,2 corresponds to the ps 1,1 , ps 2,1 in X1{ps 2,1}; and use the sub - period multi - target trajectory set X2{ps 1,1 , ps 2,2} to set the third - node trajectory set of the third leaf node P112. The τ 2,1 in the third - node trajectory set corresponds to the ps 1,1 , ps 2,2 in X2{ps 1,1 , τ 2,2 corresponds to the ps 1,1 , ps 2,2 in X2{ps 2,2};

[0124] After setting the third node trajectory sets of the third leaf nodes P111 and P112, set the third node time periods of the third leaf nodes P111 and P112 to the current sub-time period [t2, t end ; Since the current parent node is the third leaf node P11 and the third node depth is 2, the depth of the current child nodes = 2 + 1 = 3. Further, set the third node depths of the third leaf nodes P111 and P112 to 3.

[0125] In addition, the process of creating the third leaf nodes P121 and P122 below the third leaf node P12 with the third leaf node P12 as the current parent node is similar to the process of creating the third leaf nodes P111 and P112 below the third leaf node P11 with the third leaf node P11 as the current parent node, and will not be elaborated here. It is not difficult to see that Figure 2c the third leaf nodes P111 and P112 are actually 2 multi-target motion scenarios generated by two possible motions (going straight, turning right) of target 2 (vehicle 2) at time t2 in the scenario where target 1 (vehicle 1) goes straight after time t1; while the third leaf nodes P121 and P122 are actually 2 multi-target motion scenarios generated by two possible motions (going straight, turning right) of target 2 (vehicle 2) at time t2 in the scenario where target 1 (vehicle 1) turns right after time t1. Because Figure 2b the corresponding sub-time period S k has only 3. After creating the four third leaf nodes P111, P112, P121, and P122 with a node depth of 3 corresponding to the sub-time period S3[t2, t end , the second prediction tree of this working cycle can be obtained.

[0126] Step 3, identify similar nodes of the first and second prediction trees to obtain multiple first similar node pairs;

[0127] Specifically, it includes: Step 31, in the second prediction tree, extract the third node time period of the third root node as the current root node time period, and extract the third node trajectory set of the third root node as the current root node trajectory set; and in the first prediction tree, record the first root node or first leaf node whose first node time period contains or is equal to the current root node time period as the corresponding first node, so as to obtain one or more first nodes; and calculate the set trajectory similarity between the first node trajectory sets of each first node and the current root node trajectory set to obtain one or more similarity calculation results, and take the maximum value among them as the corresponding first maximum similarity; if the first maximum similarity is greater than the preset similarity threshold, then form a first similar node pair by the third root node and the first root node or first leaf node corresponding to the first maximum similarity;

[0128] Here, actually, on the prediction tree (the first prediction tree) of the previous working cycle, a node (root node or leaf node) approximated to the root node (the third root node) of the prediction tree (the second prediction tree) of the current working cycle is searched for; the identification process of the approximated node is as follows:

[0129] First, multiple nodes (root nodes or leaf nodes) that match the time of the current root node period are found on the first prediction tree, and the time matching relationship here is an inclusion or equality relationship; the so-called inclusion relationship means that the corresponding time domain of the first node period should include the corresponding time domain of the current root node period. For example, if the absolute time range corresponding to the first node period is from the 1st to the 10th second, and the absolute time range corresponding to the current root node period is from the 1st to the 5th second, or from the 5th to the 10th second, or from the 2nd to the 5th second, then the first node period is regarded as including the current root node period. However, if the absolute time range corresponding to the current root node period is from the 1st to the 11th second, or from the 0.1st to the 10th second, or from the 0.1st to the 11th second, then the first node period is regarded as not including the current root node period;

[0130] After finding multiple first nodes (the first root nodes or the first leaf nodes) with time matching on the first prediction tree, the trajectory approximation degrees of each first node and the current root node (the third root node) are calculated, and the maximum value is selected as the first maximum similarity degree, and the first maximum similarity degree is compared with a preset similarity threshold; if the first maximum similarity degree cannot be greater than the preset similarity threshold, it means that the similarity of all the first nodes obtained this time to the current root node (the third root node) is not high, that is, there is no node on the first prediction tree that truly matches the current root node (the third root node). At this time, the embodiment of the present invention will stop executing the subsequent steps and give corresponding alarm prompts; if the first maximum similarity degree is greater than the preset similarity threshold, the first node (the first root node or the first leaf node) corresponding to it on the first prediction tree is used as the approximated node of the current root node (the third root node), and they are bound to form a pair of first similar node pairs;

[0131] It should be noted that there are various implementation manners for the embodiment of the present invention to calculate the set trajectory similarity between the first node trajectory set of each first node and the current root node trajectory set; one of the implementation manners is specifically as follows: taking the first node trajectory set of the first node as the first set and the current root node trajectory set as the second set, and calculating the trajectory group similarity of two groups of first target trajectory groups R1 of each target i in the first and second sets to obtain the corresponding first target similarity ρ i , and the sum of the accumulations of the n first target similarities ρ i obtained is used as the similarity calculation result corresponding to the current first node;

[0132] Step 32: In the first prediction tree, use the branch prediction tree structure below the first root node or the first leaf node corresponding to the first maximum similarity as the corresponding first sub-prediction tree.

[0133] Here, the first root node or the first leaf node corresponding to the first maximum similarity in the first prediction tree is the node that matches the third root node of the second prediction tree. That is to say, the prediction tree structures above or collateral to this node in the first prediction tree will not appear in the prediction tree of the current working cycle. In other words, the prediction tree structures above or collateral to this node in the first prediction tree will not be associated with the second prediction tree structure. Since there is no association, to further improve the screening efficiency of similar node pairs between the first and second prediction trees, specifically, use the first root node or the first leaf node corresponding to the first maximum similarity as a reference point to prune the first prediction tree, and only keep the branch prediction tree structure below the reference point as the corresponding first sub-prediction tree for subsequent steps.

[0134] Step 33: Traverse each third leaf node of the second prediction tree one by one. During the traversal, record the currently traversed third leaf node as the current leaf node, extract the third node time period of the current leaf node as the current leaf node time period, and extract the third node trajectory set of the current leaf node as the current leaf node trajectory set. And in the first sub-prediction tree, record the first leaf node whose first node time period is equal to the current leaf node time period as the corresponding second node, so as to obtain one or more second nodes. And calculate the set trajectory similarity between the first node trajectory set of each second node and the current leaf node trajectory set to obtain one or more similarity calculation results, and take the maximum value among them as the corresponding second maximum similarity. If the second maximum similarity is greater than the similarity threshold, then form a first similar node pair by the third leaf node corresponding to the current leaf node and the first leaf node corresponding to the second maximum similarity.

[0135] Here, after completing the pruning of the first prediction tree, in the first sub-prediction tree of the previous working cycle in the embodiment of the present invention, find the second leaf nodes approximate to each third leaf node of the second prediction tree in the current working cycle to form the first similar node pairs. The identification process of approximate nodes is similar to that in Step 31, and the main difference lies in the limitation of the time matching relationship. The time matching relationship in Step 31 is the inclusion or equality relationship, and the time matching relationship in the current step is the equality relationship.

[0136] Step 4: Construct the second policy tree corresponding to the current working cycle according to the topological structure of the second prediction tree, and initialize the second policy tree according to the first policy tree and all the first similar node pairs.

[0137] Specifically, it includes: Step 41: Copy the tree structure of the second prediction tree to generate the tree structure of the second policy tree.

[0138] Among them, the second policy tree includes a fourth root node and multiple fourth leaf nodes; the fourth root node is the parent node of the subordinate leaf nodes connected thereto, and each fourth leaf node is the child node of the superior root node or superior leaf node connected thereto and is the parent node of the subordinate leaf nodes connected thereto; the fourth root node or each fourth leaf node includes a set (the fourth node time period, the fourth node trajectory set, and the fourth node depth); the topological structure of the second policy tree is consistent with that of the second prediction tree corresponding to the same working cycle: the root nodes correspond to each other, the leaf nodes correspond to each other, and the connection relationships of the parent and child nodes at all levels are consistent;

[0139] Here, since the topological structures of the prediction tree and the policy tree for each working cycle are exactly the same, the initial structure of the second policy tree can be obtained by directly copying the tree structure of the second prediction tree;

[0140] Step 42, poll all the first similar node pairs, and record the currently polled first similar node pair as the current similar node pair; and in the current similar node pair, the node corresponding to the first prediction tree is recorded as the first prediction tree node, and the node corresponding to the second prediction tree is recorded as the second prediction tree node; and based on the topological correspondence between the first prediction tree and the first policy tree, the second root node or the second leaf node in the first policy tree corresponding to the first prediction tree node is recorded as the first policy tree node; and based on the topological correspondence between the second prediction tree and the second policy tree, the second root node or the second leaf node in the second policy tree corresponding to the second prediction tree node is recorded as the second policy tree node; and initialize the fourth node time period of the second policy tree node to the second node time period of the first policy tree node, initialize the fourth node trajectory set of the second policy tree node to the second node trajectory set of the first policy tree node, and initialize the fourth node depth of the second policy tree node to the second node depth of the first policy tree node.

[0141] Here, since the topological structures of the prediction tree and the policy tree for each working cycle are exactly the same, the corresponding relationship of the similar nodes between the policy trees of the previous and the next working cycles should also match the corresponding relationship of the similar nodes between the prediction trees of the previous and the next working cycles. Therefore, after obtaining multiple groups of similar node pairs (the first similar node pairs) of the first and second prediction trees, the similar node pairs of the first and second policy trees can be directly confirmed based on this, and the node parameters (the fourth node time period, the fourth node trajectory set, and the fourth node depth) of the corresponding fourth leaf nodes on the second policy tree are initialized using the node parameters (the second node time period, the second node trajectory set, and the second node depth) of the nodes (the second root node or the second leaf node) on the first policy tree based on the similar node relationship of the policy tree.

[0142] Figure 3The figure is a schematic structural diagram of an electronic device provided in the second embodiment of the present invention. The electronic device may be the aforementioned terminal device or server, or may be a terminal device or server connected to the aforementioned terminal device or server to implement the method embodiments of the present invention. As Figure 3 shown, the electronic device may include: a processor 301 (such as a CPU), a memory 302, and a transceiver 303; the transceiver 303 is coupled to the processor 301, and the processor 301 controls the transceiver operations of the transceiver 303. Various instructions may be stored in the memory 302 for completing various processing functions and implementing the processing steps described in the foregoing method embodiments. Preferably, the electronic device related to the embodiments of the present invention further includes: a power supply 304, a system bus 305, and a communication port 306. The system bus 305 is used to implement communication connections between components. The aforementioned communication port 306 is used for the electronic device to connect and communicate with other peripherals.

[0143] In Figure 3 the system bus 305 mentioned may be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. The system bus may be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 3 only a thick line is used to represent it in

[0144] but it does not mean that there is only one bus or one type of bus. The communication interface is used to implement communication between the database access device and other devices (such as clients, read-write libraries, and read-only libraries). The memory may include a Random Access Memory (RAM), and may also include a non-volatile memory, such as at least one disk memory.

[0144] The aforementioned processor may be a general-purpose processor, including a Central Processing Unit (CPU), a Network Processor (NP), a Graphics Processing Unit (GPU), etc.; it may also be a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field Programmable Gate Array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components.

[0145] It should be noted that the embodiments of the present invention further provide a computer-readable storage medium, in which instructions are stored, and when they run on a computer, the computer is made to execute the methods and processing procedures provided in the above embodiments.

[0146] The embodiments of the present invention further provide a chip for running instructions, and the chip is used to execute the processing steps described in the foregoing method embodiments.

[0147] The embodiments of the present invention provide a data processing method, an electronic device, and a computer-readable storage medium for a planning module. First, at the start time of each working cycle, a pair of tree-shaped data objects (prediction tree and policy tree) with the same topological structure are constructed for data storage. The prediction tree is used to store a set of multi-target prediction trajectories, and the policy tree is used to store a set of planned trajectories. The corresponding node time periods and node trajectory sets are assigned to the root nodes and each leaf node of the prediction tree and the policy tree, and the parent-child node relationships are established between the upper and lower nodes in the prediction tree and the policy tree. In this way, only one parent node can store the overlapping trajectory segments of multiple trajectories corresponding to multiple child nodes. Secondly, on the premise that the topological structures of the prediction tree and the policy tree are the same in the same cycle, based on the similar nodes of the prediction tree in the previous working cycle and the current working cycle, the similar nodes of the policy tree in the two working cycles before and after are located, and the similar nodes of the policy tree in the current working cycle are initialized using the policy tree in the previous working cycle. In this way, the computational workload of the planning module in the current working cycle can be reduced by data reuse. Through the present invention, the repeated calculation amount of each working cycle of the planning module is reduced, and the redundant occupation of storage resources by the planning module is reduced.

[0148] Those skilled in the art should also further realize that the units and algorithm steps of each example described in combination with the embodiments disclosed herein can be implemented by electronic hardware, computer software, or a combination of the two. To clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described according to functions in the above description. Whether these functions are executed in a hardware or software manner depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of the present invention.

[0149] The steps of the methods or algorithms described in combination with the embodiments disclosed herein can be implemented by hardware, software modules executed by a processor, or a combination of the two. The software modules can be placed in a random access memory (RAM), memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disk, removable disk, CD-ROM, or any other form of storage medium known in the technical field.

[0150] The specific embodiments described above further elaborate on the purpose, technical solutions, and beneficial effects of the present invention. It should be understood that the above description is only the specific embodiments of the present invention and is not used to limit the protection scope of the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A data processing method for a planning module, characterized in that The method includes: The planning module receives a multi-objective prediction trajectory set output by the upstream prediction module at the start time of the current working cycle; and obtains the prediction tree and policy tree of the previous working cycle as the corresponding first prediction tree and first policy tree; Construct the prediction tree of the current working cycle according to the multi-objective prediction trajectory set to generate the corresponding second prediction tree; Identify similar nodes of the first and second prediction trees to obtain a plurality of first similar node pairs; Construct the second policy tree corresponding to the current working cycle according to the topological structure of the second prediction tree, and initialize the second policy tree according to the first policy tree and all the first similar node pairs; The multi-target prediction trajectory set includes a plurality of first target trajectory groups R i , i is the target index, 1≤i≤n, n is the number of targets; each of the first target trajectory groups R i Including multiple first prediction trajectories p i,j , j is the predicted trajectory index of target i, 1≤j≤m i 、m i is the number of predicted trajectories of target i; each of the first target trajectory groups R i Any two of the first predicted trajectories p i,j There is a section of trajectory overlap from the start time of the trajectory; the start time of the multi-target predicted trajectory set is recorded as t0, and the end time is recorded as t end , then each of the first predicted trajectories p in the set i,j The trajectory time period is [t0,t end ]; The first prediction tree is composed of a first root node and multiple first leaf nodes. The first root node is the parent node of the subordinate leaf nodes connected to it, and each of the first leaf nodes is the child node of the superior root node or superior leaf node connected to it and is the parent node of the subordinate leaf nodes connected to it. Each of the first root node or the first leaf nodes includes a set: a first node time period, a first node trajectory set, and a first node depth. The first node trajectory set includes the first target sub-trajectories τ of the target quantity n 1,i , and the first target sub-trajectory τ 1,i is the predicted trajectory segment of the corresponding target i within the current first node time period. The first node depth is the total number of nodes on the node path from the first root node to the current node. In the first prediction tree, the two first target sub-trajectories τ corresponding to the same target i between any pair of parent and child nodes 1,i should be in a continuous trajectory relationship; The first policy tree is composed of a second root node and multiple second leaf nodes. The second root node is the parent node of the subordinate leaf nodes connected thereto, and each of the second leaf nodes is the child node of the superior root node or superior leaf node connected thereto and is the parent node of the subordinate leaf nodes connected thereto. The second root node or each of the second leaf nodes each includes a set: a second node time period, a second node trajectory set, and a second node depth. The second node trajectory set is composed of a first planned trajectory τ c1 constituted, and the first planned trajectory τ c1 is the planned trajectory segment of the ego vehicle within the current second node time period. The second node depth is the total number of nodes on the node path from the second root node to the current node. In the first policy tree, the two first planned trajectories τc1 of any pair of parent and child nodes should be in a continuous trajectory relationship. The topological structure of the first policy tree corresponding to the same working cycle is consistent with that of the first prediction tree: the root nodes correspond to each other, the leaf nodes correspond to each other, and the connection relationships of the parent and child nodes at all levels are consistent. The constructing the prediction tree of the current working cycle according to the multi-objective prediction trajectory set to generate the corresponding second prediction tree specifically includes: In the multi-object prediction trajectory set, for each of the first target trajectory groups R i any two of the first prediction trajectories p i,j extract the end time of the overlapping trajectory from the start time of the trajectory to generate a corresponding first branch time; and perform duplicate time filtering on all the obtained first branch times, and sort the remaining first branch times after filtering in chronological order to obtain a first branch time series; the first branch time series includes a plurality of first branch times t a , where a is the first branch time index, 1 ≤ a ≤ A, and A is the number of first branch times; With each of the first branch times t a as the segmentation nodes, divide the time period [t0, t end corresponding to the multi-objective prediction trajectory set into A + 1 sub-time periods S k ; k is the sub-time period index, 1 ≤ k ≤ A + 1, and the sub-time period S k＝1 is denoted as the time period corresponding to the root time period [t0, t a＝1 , the sub-time period S k＝2 corresponds to the time period [t a＝1 , t a＝2 , the sub-time period S k＝3 corresponds to the time period [t a＝2 , t a＝3 , and so on, the sub-time period S k＝A+1 corresponds to the time period [t a＝A , t end ; Create a corresponding third root node for the root node in the second prediction tree corresponding to the root time period [t0, t a＝1 according to the set of multi-object prediction trajectories; the third root node includes a group: a third node time period, a third node trajectory set, and a third node depth; Using the third root node with the third node depth of 1 as the current parent node and the sub - time period S k＝2 [t a＝1 ,t a＝2 as the current sub - time period, create a corresponding plurality of third leaf nodes for each lower - level child node connected to the current parent node in the second prediction tree and corresponding to the current sub - time period according to the multi - objective prediction trajectory set; the third leaf node includes a group: a third node time period, a third node trajectory set, and a third node depth; Using each of the third leaf nodes with the third node depth of 2 as the current parent node, and using the sub-period S k＝3 [t a＝2 ,t a＝3 as the current sub-period, create multiple corresponding third leaf nodes for each of the lower-level child nodes connected to the current parent node in the second prediction tree and corresponding to the current sub-period according to the multi-objective prediction trajectory set; and so on until all the third leaf nodes corresponding to the sub-period S k＝A+1 [t a＝A ,t end are created.

2. The data processing method of the planning module according to claim 1, characterized in that The root node corresponding to the root time period [t0, t a＝1 in the second prediction tree is created according to the multi-objective prediction trajectory set to generate a corresponding third root node, specifically including: Create a root node in the second prediction tree as the corresponding third root node; Set the third node period of the third root node as the root period [t0, t a＝1 ; Set the third node depth of the third root node to 1; In each of the first target trajectory groups R of the multi-target prediction trajectory set i For any of the first predicted trajectories p in the group i,j In the root period [t0,t a＝1 ] is extracted as the corresponding target i in the root period [t0,t a＝1 ]’s second target sub-trajectory τ 2,i ; and the second target sub-trajectory τ of all targets 2,i The third node trajectory set constituting the third root node.

3. The data processing method of the planning module according to claim 2, wherein The creating, according to the multi-objective prediction trajectory set, each lower-level child node connected to the current parent node in the second prediction tree and corresponding to the current sub-period to generate the corresponding plurality of third leaf nodes specifically includes: Denote the second target sub-trajectory τ corresponding to each target i in the current parent node 2,i as the trajectory pr of the previous time period of each target i i ; For each of the first target trajectory groups R in the multi-target predicted trajectory set i extract the predicted trajectory segments of each first predicted trajectory p in the group i,j during the current sub-period to generate corresponding first segments, and form a first segment set with the obtained multiple first segments; and for the first segments in the first segment set that do not have a continuous trajectory relationship with the corresponding previous-period trajectory pr i delete them; then delete the first segments with redundant segment trajectories in the first segment set; then count the remaining number of first segments in the first segment set to generate a corresponding first quantity d i ; and use each of the remaining first segments as a target branch trajectory ps of the corresponding target i of the current first target trajectory group R i during the current sub-period, and form a corresponding sub-period target trajectory set O with all the target branch trajectories ps i,h ; each sub-period target trajectory set O i,h corresponds to a target i, and the sub-period target trajectory set O i includes the target branch trajectories ps i with the first quantity d i , where h is the branch trajectory index of target i, and 1 ≤ h ≤ d i ; i,h ; i ; According to the obtained first quantity d of the target quantity n i Calculate the number of child nodes d s , d s = d i＝1 × d i＝2 … × d i … × d i＝n ; and in the second prediction tree, create child nodes with the number of child nodes d under the current parent node s , connect the current parent node with each child node, and take each child node as a corresponding third leaf node; From each of the sub-period target trajectory sets O i select any one of the target branch trajectories ps i,h to combine and obtain a corresponding sub-period multi-target trajectory set X z ; z is the sub-period multi-target trajectory set index, 1 ≤ z ≤ d s ; the sub-period multi-target trajectory set X z includes the target branch trajectories ps of the target number n i,h , and the target branch trajectory combination relationships of each sub-period multi-target trajectory set X z are not repeated; Using the obtained number d of child nodes s for the multi-objective trajectory set X of the sub-time periods z , set one by one the third node trajectory set of the third leaf node with the obtained number d of child nodes created this time s ; Set the third node period of the third leaf node of the number d of the child nodes created this time s to the current sub-period; Increment the third node depth of the current parent node by 1 to obtain the corresponding child node depth, and set the number d s of the third leaf nodes to the child node depth.

4. The data processing method of the planning module according to claim 1, wherein The identifying similar nodes of the first and second prediction trees to obtain a plurality of first similar node pairs specifically includes: In the second prediction tree, extract the third node period of the third root node as the current root node period, and extract the third node trajectory set of the third root node as the current root node trajectory set; and in the first prediction tree, mark the first root node or the first leaf node whose first node period contains or is equal to the current root node period as the corresponding first node, so as to obtain one or more first nodes; and calculate the set trajectory similarity between the first node trajectory set of each first node and the current root node trajectory set to obtain one or more similarity calculation results, and take the maximum value among them as the corresponding first maximum similarity; if the first maximum similarity is greater than the preset similarity threshold, then form a first similar node pair by the third root node and the first root node or the first leaf node corresponding to the first maximum similarity; Take the branch prediction tree structure under the first root node or the first leaf node corresponding to the first maximum similarity in the first prediction tree as the corresponding first sub-prediction tree; Traverse the third leaf nodes of the second prediction tree one by one; during the traversal, record the currently traversed third leaf node as the current leaf node, extract the third node time period of the current leaf node as the current leaf node time period, and extract the third node trajectory set of the current leaf node as the current leaf node trajectory set; and in the first sub-prediction tree, record the first leaf node whose first node time period is equal to the current leaf node time period as the corresponding second node, so as to obtain one or more of the second nodes; and calculate the set trajectory similarity between the first node trajectory set of each of the second nodes and the current leaf node trajectory set to obtain one or more similarity calculation results, and take the maximum value among them as the corresponding second maximum similarity; if the second maximum similarity is greater than the similarity threshold, then form a first similar node pair by the third leaf node corresponding to the current leaf node and the first leaf node corresponding to the second maximum similarity.

5. The data processing method of the planning module according to claim 1, wherein The construction of the second policy tree corresponding to the current working cycle according to the topological structure of the second prediction tree, and the initialization of the second policy tree according to the first policy tree and all the first similar node pairs specifically include: Copy the tree structure of the second prediction tree to generate the tree structure of the second policy tree; the second policy tree includes a fourth root node and multiple fourth leaf nodes; the fourth root node is the parent node of the subordinate leaf nodes connected to it, and each of the fourth leaf nodes is the child node of the superior root node or superior leaf node connected to it and the parent node of the subordinate leaf nodes connected to it; each of the fourth root node or the fourth leaf nodes includes a set: a fourth node time period, a fourth node trajectory set, and a fourth node depth; the topological structure of the second policy tree is consistent with that of the second prediction tree corresponding to the same working cycle: the root nodes correspond to each other, the leaf nodes correspond to each other, and the connection relationships of the parent and child nodes at all levels are consistent; Poll all the first similar node pairs, and denote the currently polled first similar node pair as the current similar node pair; and in the current similar node pair, denote the node corresponding to the first prediction tree as the first prediction tree node, and the node corresponding to the second prediction tree as the second prediction tree node; and based on the topological correspondence between the first prediction tree and the first policy tree, denote the second root node or the second leaf node in the first policy tree corresponding to the first prediction tree node as the first policy tree node; and based on the topological correspondence between the second prediction tree and the second policy tree, denote the second root node or the second leaf node in the second policy tree corresponding to the second prediction tree node as the second policy tree node; and initialize the fourth node time period of the second policy tree node to the second node time period of the first policy tree node, initialize the fourth node trajectory set of the second policy tree node to the second node trajectory set of the first policy tree node, and initialize the fourth node depth of the second policy tree node to the second node depth of the first policy tree node.

6. An electronic device, characterized in that, Comprising: A memory, a processor, and a transceiver; The processor is used to be coupled with the memory, read and execute instructions in the memory to implement the method according to any one of claims 1-5; The transceiver is coupled with the processor, and the processor controls the transceiver to perform message sending and receiving.

7. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions, and when the computer instructions are executed by a computer, the computer is caused to execute the method according to any one of claims 1-5.

Images