CN116804560B - Unmanned automobile safety navigation method and device under controlled road section - Google Patents
Unmanned automobile safety navigation method and device under controlled road section Download PDFInfo
- Publication number
- CN116804560B CN116804560B CN202311067417.5A CN202311067417A CN116804560B CN 116804560 B CN116804560 B CN 116804560B CN 202311067417 A CN202311067417 A CN 202311067417A CN 116804560 B CN116804560 B CN 116804560B
- Authority
- CN
- China
- Prior art keywords
- information
- processing
- road
- traffic
- time
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 41
- 238000012545 processing Methods 0.000 claims abstract description 139
- 230000033001 locomotion Effects 0.000 claims abstract description 55
- 238000013439 planning Methods 0.000 claims abstract description 47
- 238000013178 mathematical model Methods 0.000 claims abstract description 39
- 238000011217 control strategy Methods 0.000 claims abstract description 38
- 230000004927 fusion Effects 0.000 claims abstract description 35
- 230000008447 perception Effects 0.000 claims abstract description 28
- 238000013527 convolutional neural network Methods 0.000 claims abstract description 11
- 230000002787 reinforcement Effects 0.000 claims abstract description 10
- 238000010219 correlation analysis Methods 0.000 claims abstract description 8
- 238000004891 communication Methods 0.000 claims description 15
- 238000007499 fusion processing Methods 0.000 claims description 12
- 238000004458 analytical method Methods 0.000 claims description 11
- 230000004888 barrier function Effects 0.000 claims description 10
- 238000012163 sequencing technique Methods 0.000 claims description 8
- 238000007500 overflow downdraw method Methods 0.000 claims description 7
- 230000008859 change Effects 0.000 claims description 6
- 238000000605 extraction Methods 0.000 claims description 6
- 230000003595 spectral effect Effects 0.000 claims description 6
- 238000012300 Sequence Analysis Methods 0.000 claims description 5
- 238000013528 artificial neural network Methods 0.000 claims description 5
- 238000013135 deep learning Methods 0.000 claims description 5
- 238000010845 search algorithm Methods 0.000 claims description 5
- 230000008569 process Effects 0.000 description 10
- 238000010276 construction Methods 0.000 description 8
- 206010039203 Road traffic accident Diseases 0.000 description 6
- 230000007613 environmental effect Effects 0.000 description 5
- 238000005457 optimization Methods 0.000 description 5
- 230000008901 benefit Effects 0.000 description 4
- 239000000284 extract Substances 0.000 description 4
- 230000004044 response Effects 0.000 description 4
- 230000009471 action Effects 0.000 description 3
- 230000009286 beneficial effect Effects 0.000 description 3
- 239000013598 vector Substances 0.000 description 3
- 230000003044 adaptive effect Effects 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 2
- 230000007547 defect Effects 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 239000004973 liquid crystal related substance Substances 0.000 description 2
- 238000007726 management method Methods 0.000 description 2
- 230000007246 mechanism Effects 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000007704 transition Effects 0.000 description 2
- OKTJSMMVPCPJKN-UHFFFAOYSA-N Carbon Chemical compound [C] OKTJSMMVPCPJKN-UHFFFAOYSA-N 0.000 description 1
- 230000002457 bidirectional effect Effects 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 229910052799 carbon Inorganic materials 0.000 description 1
- 230000010485 coping Effects 0.000 description 1
- 125000004122 cyclic group Chemical group 0.000 description 1
- 238000000354 decomposition reaction Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 239000003550 marker Substances 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 238000012913 prioritisation Methods 0.000 description 1
- 238000012216 screening Methods 0.000 description 1
- 238000001228 spectrum Methods 0.000 description 1
- 238000000547 structure data Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 230000002123 temporal effect Effects 0.000 description 1
Landscapes
- Traffic Control Systems (AREA)
Abstract
The invention provides a safe navigation method and device for an unmanned automobile under a controlled road section, which relate to the technical field of unmanned, and comprise the steps of acquiring first information and second information; processing real-time road condition information, traffic state information and real-time perception information by using a space information fusion mathematical model based on a convolutional neural network to generate a dynamic semantic map; carrying out weight distribution processing on the dynamic semantic map according to rescue task information and a preset reinforcement learning mathematical model to obtain a priority weight map; carrying out path planning processing on the priority weight graph according to a preset dynamic path planning mathematical model to obtain a driving route; and carrying out space-time correlation analysis processing on the dynamic semantic map, the driving route and the real-time environment perception information by using a space-time information fusion mathematical model based on a long-short time memory network to obtain a motion control strategy of the unmanned automobile. The invention improves the safety and the passing efficiency of the unmanned automobile in the environment of the complex controlled road section.
Description
Technical Field
The invention relates to the technical field of unmanned aerial vehicles, in particular to a method and a device for safely navigating an unmanned automobile under a controlled road section.
Background
With the rapid development of unmanned technology, unmanned automobiles have been widely used in various scenes. However, in complex controlled road sections, such as crowded city streets, emergency rescue scenes, and the like, the existing unmanned car navigation method often cannot well meet the requirements in terms of safety, efficiency, priority, and the like. The control road section has the characteristics of complex traffic environment, temporary traffic control, special road conditions and the like, the prior art has limitation in processing real-time road condition information, traffic state information and rescue task information, and cannot be well adapted to the complexity of the control road section, for example, traffic accidents occur in one of a certain bidirectional four-lane road, the traffic accidents occur locally in the lane, after the front accident occurs and the traffic is caused, the control method of the unmanned automobile in the prior art directly transfers to a manual driving mode after receiving the front accident, the other control method reduces the speed of the automobile and changes the road to the other lane, but the operation causes the condition that the other normal road is jammed and even the blocked automobile cannot pass, so that the traffic efficiency is reduced, and the on-site rescue work is directly influenced, so that the unmanned automobile safety navigation method under the control road section is needed.
Aiming at the defects of the prior art, the traffic operation management teacher team in the Sichuan traffic occupation technical college comprehensively analyzes the road characteristics, the environment characteristics, the vehicle characteristics, the obstacle characteristics and other on-site information of various control road sections, compares and extracts the information with the related characteristic information under the condition of a common road (no accident or no congestion road section), and provides a navigation method capable of better solving the problem of unmanned automobiles under the control road sections. The method can better relieve local traffic jam, reduce vehicle jam time and achieve the effects of energy saving during transportation and reducing vehicle carbon emission.
Disclosure of Invention
The application aims to provide a safe navigation method and device for an unmanned automobile under a controlled road section, so as to solve the problems. In order to achieve the above purpose, the technical scheme adopted by the application is as follows:
in a first aspect, the present application provides a method for safe navigation of an unmanned car under a controlled road section, including: acquiring first information and second information, wherein the first information comprises real-time road condition information, traffic state information and rescue task information of a controlled road section collected by adopting vehicle-mounted communication equipment and infrastructure communication equipment, and the second information comprises real-time environment perception information collected by an unmanned vehicle sensor;
Processing the real-time road condition information, the traffic state information and the real-time perception information by using a space information fusion mathematical model based on a convolutional neural network to generate a dynamic semantic map containing road attributes, traffic signs, signals and barrier information;
carrying out weight distribution processing on the dynamic semantic map according to the rescue task information and a preset reinforcement learning mathematical model to obtain a priority weight map, wherein the priority weight map comprises comprehensive weights of road traffic conditions, traffic jam degree and rescue priority;
carrying out path planning processing on the priority weight graph according to a preset dynamic path planning mathematical model to obtain at least one driving route adapting to the complex environment of the control road section;
and carrying out space-time correlation analysis processing on the dynamic semantic map, the driving route and the real-time environment perception information by using a space-time information fusion mathematical model based on a long-short time memory network to obtain a motion control strategy of the unmanned automobile, wherein the motion control strategy comprises driving parameters of the unmanned automobile.
In a second aspect, the present application also provides a safety navigation device for an unmanned automobile under a controlled road section, including: the first acquisition unit is used for acquiring first information and second information, wherein the first information comprises real-time road condition information, traffic state information and rescue task information of a controlled road section collected by adopting vehicle-mounted communication equipment and infrastructure communication equipment, and the second information comprises real-time environment perception information collected by an unmanned vehicle sensor;
The first processing unit is used for processing the real-time road condition information, the traffic state information and the real-time perception information by using a space information fusion mathematical model based on a convolutional neural network, and generating a dynamic semantic map containing road attributes, traffic signs, signals and barrier information;
the second processing unit is used for carrying out weight distribution processing on the dynamic semantic map according to the rescue task information and a preset reinforcement learning mathematical model to obtain a priority weight map, wherein the priority weight map comprises comprehensive weights of road traffic conditions, traffic jam degree and rescue priority;
the third processing unit is used for carrying out path planning processing on the priority weight graph according to a preset dynamic path planning mathematical model to obtain at least one driving route which is suitable for the complex environment of the controlled road section;
and the fourth processing unit is used for carrying out space-time correlation analysis processing on the dynamic semantic map, the driving route and the real-time environment perception information by using a space-time information fusion mathematical model based on a long-short time memory network to obtain a motion control strategy of the unmanned automobile, wherein the motion control strategy comprises driving parameters of the unmanned automobile.
The beneficial effects of the invention are as follows:
the invention obtains road, traffic and rescue task information in real time, and utilizes the techniques of convolutional neural network, long-short-time memory network and the like to perform spatial information fusion and space-time correlation analysis on the data, so as to generate a dynamic semantic map, a priority weight map and a motion control strategy. The method overcomes the defects of the prior art in unmanned automobile navigation under the controlled road section, and improves the safety and the passing efficiency of the unmanned automobile in the environment of the complex controlled road section. Meanwhile, the invention can ensure that the unmanned automobile follows the driving route and preferentially passes through the control road section in the process of executing the rescue task, thereby improving the completion efficiency of the rescue task.
Additional features and advantages of the invention will be set forth in the description which follows, and in part will be apparent from the description, or may be learned by practice of the embodiments of the invention. The objectives and other advantages of the invention will be realized and attained by the structure particularly pointed out in the written description and claims thereof as well as the appended drawings.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings that are needed in the embodiments will be briefly described below, it being understood that the following drawings only illustrate some embodiments of the present invention and therefore should not be considered as limiting the scope, and other related drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.
Fig. 1 is a schematic flow chart of a method for safely navigating an unmanned car under a controlled road section according to an embodiment of the invention;
fig. 2 is a schematic structural diagram of an unmanned car safety navigation device under a controlled road section according to an embodiment of the invention.
The marks in the figure: 701. a first acquisition unit; 702. a first processing unit; 703. a second processing unit; 704. a third processing unit; 705. a fourth processing unit; 7021. a first extraction subunit; 7022. a first identification subunit; 7023. a second identification subunit; 7024. a first fusion subunit; 7031. a first sorting subunit; 7032. a first dividing subunit; 7033. a sixth processing subunit; 7034. a first computing subunit; 7035. a first allocation subunit; 7041. a first analysis subunit; 7042. a second sorting subunit; 7043. a seventh processing subunit; 7044. a first adjustment subunit; 7045. a second dividing subunit; 7051. an eighth processing subunit; 7052. a ninth processing subunit; 7053. a tenth processing subunit; 7054. a first predictor unit; 7055. a second fusion subunit; 70241. a first processing subunit; 70242. a second processing subunit; 70243. a third processing subunit; 70244. a fourth processing subunit; 70245. and a fifth processing subunit.
Detailed Description
For the purpose of making the objects, technical solutions and advantages of the embodiments of the present invention more apparent, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention, and it is apparent that the described embodiments are some embodiments of the present invention, but not all embodiments of the present invention. The components of the embodiments of the present invention generally described and illustrated in the figures herein can be arranged and designed in a wide variety of configurations each. Thus, the following detailed description of the embodiments of the invention, as presented in the figures, is not intended to limit the scope of the invention, as claimed, but is merely representative of selected embodiments of the invention. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
It should be noted that: like reference numerals and letters denote like items in the following figures, and thus once an item is defined in one figure, no further definition or explanation thereof is necessary in the following figures. Meanwhile, in the description of the present invention, the terms "first", "second", and the like are used only to distinguish the description, and are not to be construed as indicating or implying relative importance.
Example 1
As shown in fig. 1, the present embodiment provides a method for safe navigation of an unmanned car under a controlled road section, which includes steps S1, S2, S3, S4 and S5.
Step S1, acquiring first information and second information, wherein the first information comprises real-time road condition information, traffic state information and rescue task information of a controlled road section collected by adopting vehicle-mounted communication equipment and infrastructure communication equipment, and the second information comprises real-time environment perception information collected by an unmanned vehicle sensor;
it can be understood that it is important to acquire road information and environmental perception information in real time in the case of road construction, traffic accidents, natural disasters and the like. The information is not only helpful for unmanned vehicles to adjust the driving route in time, avoid potential danger, but also ensure the smooth progress of rescue tasks. Further, the interconnection and intercommunication of the vehicle-mounted communication equipment and the infrastructure communication equipment can realize real-time transmission and sharing of information between the unmanned vehicles, and the perception and response capability of the unmanned vehicles to road environments are improved.
S2, processing real-time road condition information, traffic state information and real-time perception information by using a space information fusion mathematical model based on a convolutional neural network, and generating a dynamic semantic map containing road attributes, traffic signs, signals and barrier information;
It can be understood that by using the convolutional neural network technology in this step, various elements on the road, such as traffic signs, signals, obstacles, etc., can be more accurately identified, and the dynamic semantic map is updated in real time, so as to help the unmanned automobile make more accurate driving decisions in complex road conditions. The real-time updating of the dynamic semantic map can enable the unmanned automobile to be better adapted to the change of road conditions, and misjudgment caused by outdated information is avoided. Further, this step can also enable the fusion of multiple data sources, such as vehicle-self awareness data, data shared by other vehicles, and data provided by infrastructure communication devices. The data fusion can enhance the perception capability of the unmanned automobile to the environment, improve the coping speed of the emergency and reduce the risk of accidents.
Step S3, carrying out weight distribution processing on the dynamic semantic map according to rescue task information and a preset reinforcement learning mathematical model to obtain a priority weight map, wherein the priority weight map comprises comprehensive weights of road traffic conditions, traffic jam degree and rescue priority;
it can be understood that in the step, comprehensive balance is carried out on various information through the reinforcement learning model, so that the unmanned automobile can better meet the requirements of rescue tasks, and meanwhile, the situation that the unmanned automobile falls into a congestion or dangerous road section is avoided. The generation of the priority weight map helps to realize cooperative driving among unmanned automobiles. When a plurality of unmanned vehicles run on the controlled road section at the same time, the unmanned vehicles can cooperatively plan the running route according to the priority weight graph, so that mutual interference is avoided, and the overall road passing efficiency is improved. In addition, the reinforcement learning model can be continuously optimized along with the driving experience of the unmanned automobile, so that the unmanned automobile can make more reasonable and efficient decisions when facing similar scenes.
S4, carrying out path planning processing on the priority weight graph according to a preset dynamic path planning mathematical model to obtain at least one driving route adapting to the complex environment of the control road section;
it can be understood that various obstacles such as construction equipment, staff, roadblocks and the like may exist on the control section of the road construction, and the real-time road condition and the traffic state are changed at any time. Under the condition, the dynamic path planning model comprehensively considers the factors such as road traffic conditions, traffic jam degree, rescue priority and the like of the construction area, and plans a driving route which can safely pass without influencing the construction progress for the unmanned automobile. Similarly, in traffic accidents or natural disasters, real-time road conditions and traffic conditions are affected by accident vehicles, rescue vehicles, damaged roads and other factors. At this time, the dynamic path planning model fully considers the factors, and plans a driving route for the unmanned automobile, wherein the driving route avoids potential danger and ensures the smooth progress of rescue tasks.
And S5, performing space-time correlation analysis processing on the dynamic semantic map, the driving route and the real-time environment perception information by using a space-time information fusion mathematical model based on a long-short time memory network to obtain a motion control strategy of the unmanned automobile, wherein the motion control strategy comprises driving parameters of the unmanned automobile.
It can be understood that the long-short-term memory network is a special cyclic neural network and has strong capability of processing time sequence data and space data. In the step, the long-short-term memory network predicts and decides the motion control of the unmanned automobile under different time and space conditions by analyzing the dynamic semantic map, the driving route and the real-time environment perception information. The time-space information fusion method can provide an accurate and real-time motion control strategy for the unmanned automobile in a complex control road section environment.
In one embodiment of the present disclosure, step S2 includes steps S21, S22, S23 and step S24.
S21, carrying out road attribute extraction processing according to the real-time road condition information to obtain road attribute information, wherein the road attribute information comprises a road type, a width and a speed limit;
it can be understood that the analysis of the real-time road condition information is helpful for the unmanned vehicle to know the basic condition of the current driving road, and the unmanned vehicle can make corresponding driving strategy adjustment according to the current road environment by extracting the road attribute information. For example, on a highway, an unmanned vehicle may need to travel at high speeds according to speed limit requirements while maintaining a stable lane distance according to road width in a multi-lane environment. On urban arterial roads, unmanned vehicles may need to adjust speed according to traffic congestion conditions and pay attention to compliance with relevant regulations such as traffic signals. On rural roads, unmanned vehicles may need to face challenges such as narrow roads, complex terrains, and temporary traffic signs, and thus real-time adjustment of driving strategies is required to ensure safe driving.
S22, carrying out traffic sign and signal lamp identification processing according to real-time traffic state information to obtain traffic sign information and signal lamp information, wherein the traffic sign information comprises a stop sign, a speed limit sign and a traffic prohibition sign, and the signal lamp information comprises a traffic light state and a countdown time;
it will be appreciated that the analysis of the real-time traffic state information in this step facilitates the unmanned vehicles to accurately recognize and follow the indications of traffic signs and lights.
Step S23, performing obstacle recognition processing according to the real-time environment sensing information to obtain obstacle information, wherein the obstacle information comprises pedestrians, vehicles and objects;
it can be understood that the invention provides the unmanned automobile with timely and accurate obstacle information through the obstacle identification processing of the real-time environment perception information, is beneficial to realizing safe and efficient running, and ensures that various challenges can be independently dealt with in a complex road environment.
S24, carrying out fusion processing on road attribute information, traffic sign information, signal lamp information and barrier information according to a spatial information fusion mathematical model to obtain a dynamic semantic map;
it can be understood that the invention provides richer and more accurate environmental perception data for the unmanned automobile by fusing various information sources, and is beneficial to improving the driving safety and efficiency of the unmanned automobile under the complex road condition. In addition, the real-time updating of the dynamic semantic map can enable the unmanned automobile to timely cope with emergencies such as traffic accidents, road construction and the like, so that smoothness and safety of a driving process are ensured.
In one embodiment of the present disclosure, step S24 includes step S241, step S242, step S243, step S244, and step S245.
Step S241, road topological structure processing based on a graph convolution neural network is carried out according to road attribute information, so as to obtain road attribute data;
it can be appreciated that the graph convolutional neural network is a method for deep learning by using graph structure data, and can capture local and global features in the graph structure. In the present invention, by processing road attribute information using a graph roll-up neural network, topology information of a road can be extracted therefrom. The topological structure information comprises the connection relation and the mutual position relation among roads and other information related to road attributes, such as road types, widths, speed limits and the like, and can help the unmanned automobile to better understand the structure and characteristics of the road network, so that more accurate driving decisions can be made in a complex road environment.
Step S242, different weights are given to different types of traffic signs in the traffic sign information, and weighted traffic sign data matched with real-time road conditions are obtained;
it is understood that traffic signs play an important role in road traffic, providing driving rules and restriction information for unmanned vehicles. In this step, different weights are given to different types of traffic signs (such as stop signs, speed limit signs, traffic prohibition signs, etc.) according to real-time road conditions, so as to highlight traffic signs that are more critical to the current road conditions and help the unmanned vehicle make a more appropriate driving decision. For example, in congested road segments, the weight of the speed limit markers may be relatively high to ensure that the unmanned vehicle is traveling within the speed limit. In emergency rescue tasks, however, the weight of the no-pass sign may be reduced to ensure that the unmanned vehicle can reach the destination quickly when necessary. By giving different weights to the traffic sign, the dynamic semantic map can be more in line with real-time road conditions, so that the driving safety and efficiency of the unmanned automobile in a controlled road section and a complex environment are improved.
Step S243, carrying out time sequence analysis according to signal lamp information, and predicting the change trend of the signal lamp state to obtain signal lamp data;
it can be understood that the invention can predict the change trend of the signal lamp state (such as red light and green light) by carrying out time sequence analysis on the signal lamp information. This helps the unmanned vehicle to make ready for travel ahead of time for smoother passage through the traffic light. In addition, through predicting the change trend of signal lamp state, can also improve the efficiency of passing of unmanned vehicles under the complex environment. For example, in an emergency rescue task, the unmanned vehicle can reasonably adjust the driving speed according to the prediction result of the signal lamp state so as to quickly pass through in green light, thereby reaching the destination as soon as possible.
Step S244, performing multi-mode fusion method processing of self-adaptive weights according to the obstacle information, and performing weighted fusion on the obstacle information from different sources to obtain obstacle data;
it can be appreciated that the multi-modal fusion method of adaptive weights can adjust the weight distribution according to different situations, which helps the unmanned vehicle to process obstacle information more flexibly in a complex environment. In a road construction area, the unmanned automobile pays more attention to barrier information such as construction equipment and the like so as to ensure running safety; in busy commercial areas, however, unmanned vehicles may be more concerned with pedestrian and vehicle information to better accommodate complex traffic environments.
And step 245, performing space-time sequence fusion processing on the road attribute data, the weighted traffic sign data, the signal lamp data and the obstacle data by using a deep learning method based on a gating circulating unit to obtain a dynamic semantic map.
It can be understood that the method and the system align the time sequence of the information of different data sources through the time sequence fusion module, and then spatially fuse the information of the different data sources by utilizing the space fusion module, so as to obtain a globally consistent dynamic semantic map. Specifically, first, the road attribute data, the weighted traffic sign data, the signal lamp data, and the obstacle data are converted into a space-time sequence form, and are respectively input into the gate control loop unit. The gating circulation unit is mainly used for processing the sequence data, wherein the gating mechanism can adaptively control the inflow and outflow of information, so that the screening and weighting of the information are realized. The gating loop unit then performs a layer-by-layer processing of the input spatio-temporal sequence to gradually extract the useful feature representation. Preferably, the gating loop unit obtains the corresponding spatio-temporal feature representation by learning the long-range dependency relationship and the time correlation in the spatio-temporal sequence. Meanwhile, by adding a gating mechanism, the gating circulation unit can adaptively carry out weighted fusion on data from different sources, so that a more accurate and reliable dynamic semantic map is obtained. And finally, summarizing and integrating the output of the gating circulation unit to obtain a final dynamic semantic map.
In one embodiment of the present disclosure, step S3 includes step S31, step S32, step S33, step S34, and step S35.
Step S31, performing topology sequencing processing according to each rescue task in the rescue task information to obtain an ordered rescue task list;
it can be understood that in the rescue task, the problem of interdependence and execution sequence of different tasks often exists. For example, when an emergency such as a traffic accident or a natural disaster occurs, rescue needs to be performed as soon as possible, but priority and dependency relationship exist between different rescue tasks, such as first transporting wounded persons and then dragging vehicles. According to the invention, the rescue task list is processed by using the topological ordering, so that the rescue tasks are ensured to be executed according to the correct sequence, and the rescue efficiency and success rate are improved.
S32, performing spectral clustering according to the dynamic semantic map, dividing the map into a plurality of subareas, and obtaining a subarea map list;
it will be appreciated that spectral clustering is an unsupervised learning algorithm that constructs an undirected graph by treating data points as nodes in the graph and calculates the similarity between them. Then, the spectrum clustering algorithm converts the graph into a matrix, and further performs feature vector decomposition to obtain a low-dimensional space composed of feature vectors. Finally, the data points are divided into different categories by carrying out a clustering algorithm such as K-means on the feature vectors. According to the invention, the spectral clustering algorithm can divide the information of similar roads, traffic signs, signal lamps, obstacles and the like in the dynamic semantic map into the same subarea, so that the subsequent processing is convenient.
Step S33, processing based on a heuristic search algorithm is carried out according to the rescue task list, and an optimal subarea of each rescue task in the subarea map list is determined, so that a task-subarea matching relationship is obtained;
it can be understood that the invention searches each rescue task by adopting the heuristic search algorithm to find an optimal sub-region matching scheme, so that the rescue vehicle can reach the task place in the shortest time. In the searching process, the optimal task-subarea matching relationship is obtained through repeated iterative searching by considering the factors such as traffic jam conditions, road traffic conditions, arrival time of rescue vehicles, priority of rescue tasks and the like of each subarea, so that efficient response and rapid rescue of unmanned vehicles are realized.
Step S34, calculating rescue task priorities of all subareas according to the task-subarea matching relationship to obtain a subarea priority list;
it can be understood that the invention distributes each rescue task to the corresponding subarea according to the task-subarea matching relationship, and calculates the priority of the subarea according to the number of rescue tasks in the subarea, the emergency degree and other factors. By generating the sub-region priority list, more accurate task scheduling and path planning can be provided for the unmanned automobile, so that rescue tasks can be completed more efficiently. Meanwhile, according to the subarea priority list, the traffic jam conditions of different subareas can be analyzed and evaluated, and valuable references are provided for urban traffic management. The calculation formula is as follows:
;
Wherein, the liquid crystal display device comprises a liquid crystal display device,rescue task priority indicating the ith sub-area; ->Indicating the number of rescue tasks in the subarea, N indicating the total number of rescue tasks in all subareas; ->Representing the sonEmergency of rescue task in area, +.>Representing the average response time of rescue tasks in this sub-area;>、/>and->The weights of the rescue task number, the emergency degree and the average response time are respectively expressed.
And step S35, carrying out weight distribution on the dynamic semantic map according to the sub-region priority list and a preset depth deterministic strategy gradient mathematical model to obtain a priority weight map.
It can be understood that the invention adopts a preset depth deterministic strategy gradient mathematical model, and the model obtains a better weight distribution strategy by learning historical sub-region selection decisions and corresponding priority weights. Preferably, the model firstly performs preliminary weight distribution on the dynamic semantic map according to the subarea priority list, and then further optimizes the weight by using a depth deterministic strategy gradient algorithm so as to maximize the priority of the rescue task and ensure the balance of each subarea. The finally obtained priority weight map can be used for guiding the unmanned vehicle to run in path planning, so that rescue efficiency and success rate are improved.
In one embodiment of the present disclosure, step S4 includes step S41, step S42, step S43, step S44, and step S45.
S41, performing control road section characteristic analysis processing according to a priority weight graph, and extracting special attributes of a control area to obtain control road section characteristic information, wherein the special attributes comprise an emergency lane and an forbidden traffic area;
it can be understood that the invention can better provide running guide for the unmanned automobile through the characteristic analysis of the controlled road section, help the unmanned automobile to run along the specified route in the controlled road section, avoid the traffic-forbidden area and improve the running safety and efficiency. In special cases, such as emergency or rescue tasks, the unmanned vehicle can flexibly adjust the driving route according to the special attribute of the control road section through the control road section, so that the emergency situation can be better dealt with. In addition, the extraction and analysis of the characteristic information of the control road section also provides necessary data support for subsequent path planning and motion control, so that the unmanned vehicle can more intelligently avoid various obstacles in the control road section, and the driving safety is ensured.
Step S42, carrying out priority ranking processing based on the obstacles and the rescue vehicles according to the characteristic information of the control road section, determining the priorities of the obstacles and the rescue vehicles of different types in path planning, and obtaining the priority information of the obstacles and the rescue vehicles;
It will be appreciated that the prioritization process may be based on the nature of different types of obstacles, such as size, shape, weight, location, etc., to better avoid these obstacles during path planning. For the priority sorting of the rescue vehicles, the priority and the emergency degree of the rescue tasks can be considered, so that the rescue vehicles can be ensured to arrive at the destination safely with priority. Through the step, the method and the device can improve the accuracy and the practicability of path planning and provide better support for the execution of rescue tasks.
Step S43, carrying out mixed integer programming algorithm processing according to the priority weight graph and the priority information, and solving the possible passing conflict problem to obtain a conflict solution;
it can be appreciated that the invention models conflict situations in the managed road section by using a mixed integer programming algorithm, including situations of intersection conflict, vehicle conflict and the like, and converts the problem into a linear programming problem by mathematically describing conflict constraint conditions. And then solving by a linear programming solver to obtain an optimal solution, including a vehicle motion track, a vehicle speed and the like. The algorithm has the advantages that the algorithm can be globally optimized, and the priority of each vehicle in the control road section is ensured, so that the priority of the rescue task is ensured to be met, and the algorithm can be quickly solved. The specific calculation formula is as follows:
Wherein N represents the number of vehicles; k represents the number of selectable running speeds of the vehicle, W represents the number of topological neighbors of the vehicle, T represents the number of time steps V represents the range of values of the speedBoolean variable representing the transition of the ith vehicle from speed k to speed k+1 and to the jth node at time step t, if vehicle i transitions from speed k to speed k+1 and to the jth node at time step t>Has a value of 1, otherwise 0; ->Indicating whether there is an obstacle w on the path from node i to node j at time t>Representing the speed of obstacle w at the jth node in time step t>Indicating the possibility of selecting the speed k of the ith vehicle at time step t;>represents the maximum incoming traffic of the jth node at time step t;>representing the maximum outgoing traffic of the ith node at time step t>Representing the maximum traffic capacity between node i and node j;>indicating whether an edge between node i and node j is present at time step t, if so,/-, is present>And is 1 otherwise 0, deltat representing the time step.
Step S44, dynamically adjusting road traffic weight according to the real-time road condition information according to the conflict solution to obtain real-time road traffic weight information;
It can be understood that the invention dynamically adjusts the road passing weight according to the existing conflict solution by collecting and analyzing the real-time road condition information, so that vehicles with higher priority can obtain more priority passing weight, and meanwhile, the conflict and congestion among vehicles are avoided. Therefore, the unmanned vehicle can efficiently run in real-time traffic under the premise of ensuring safety and efficiency.
And S45, dividing the control road section into a plurality of subareas according to the real-time road traffic weight information and a preset hierarchical path planning strategy, planning paths in each subarea respectively, and connecting the paths of the subareas to obtain at least one driving route adapting to the complex environment of the control road section.
It can be appreciated that in the present invention, the hierarchical path planning strategy can be divided into two levels: high-level path planning and low-level path planning. In high-level path planning, a control road section is divided into a plurality of sub-areas according to specific conditions of rescue tasks and map information, and a target position of each sub-area is preset. And then, determining the priority of each sub-area and the traffic capacity between each sub-area and other sub-areas according to the current real-time road traffic weight information, and obtaining a relation diagram between the sub-areas. Next, by using a graph theory algorithm, preferably using a shortest path algorithm, a minimum spanning tree algorithm, etc., to calculate the optimal path between the sub-regions, a high-level path planning result is obtained. In the low-level path planning, for each sub-area, path planning is further carried out according to real-time road traffic weight information in the area, a path planning algorithm based on a graph, such as Dijkstra algorithm, A-type algorithm and the like, is used for calculating an optimal path from the starting point to the end point of the sub-area, and factors such as obstacle avoidance, priority weight and the like are considered to obtain a low-level path planning result. And finally, combining the high-level path planning result and the low-level path planning result to obtain the driving route adapting to the complex environment of the control road section. The whole hierarchical path planning strategy realizes the global optimization of the managed road section, can ensure that the unmanned vehicle runs safely and efficiently in a complex environment, and meets the requirements of rescue tasks.
In one embodiment of the present disclosure, step S5 includes step S51, step S52, step S53, step S54, and step S55.
Step S51, performing multi-objective optimization processing according to the dynamic semantic map, the driving route and the real-time perception information, balancing the traffic efficiency and the safety of the controlled road section, and obtaining a first motion control strategy of the unmanned automobile;
it can be understood that the invention determines the optimal driving speed and lane by comprehensively considering various factors including the vehicle speed, the lane width, the traffic flow, the obstacle position, etc. on the basis of the real-time perception information. Meanwhile, considering safety, taking obstacle avoidance as an important constraint condition, the unmanned automobile is ensured not to collide with other vehicles or obstacles in the running process. Through the multi-objective optimization processing, the passing efficiency and the safety are balanced, so that a first motion control strategy applicable to the current control road section is obtained, and guidance is provided for subsequent driving.
Step S52, carrying out priority-based path constraint processing according to the driving route and the rescue task information, ensuring that the unmanned automobile follows the driving route and preferentially passes through the control road section in the process of executing the rescue task, and obtaining a constrained second motion control strategy;
It can be understood that the priority weight assignment is carried out on each road section on the path, the priority constraint processing is carried out on the path by marking the position and the priority of the rescue task, so that the unmanned vehicle can pass through the control road sections according to the priority sequence when executing the rescue task, and the priority of the rescue task is ensured to be satisfied. The finally obtained second motion control strategy is to add priority constraint on the basis of the first strategy so as to meet the special requirements of rescue tasks.
Step S53, performing cooperative anti-collision processing according to the real-time environment sensing information, and analyzing the motion states of surrounding vehicles and obstacles to obtain a cooperative anti-collision motion control strategy;
it can be understood that the invention can predict the possible collision situation in advance by sensing the motion states of surrounding vehicles and obstacles in real time and considering the existence, the motion direction and other information in the motion control strategy, and take corresponding measures to avoid the collision, thereby ensuring the safety of rescue tasks.
Step S54, according to the dynamic semantic map and the driving route, predicting the movement track of the unmanned vehicle in the controlled road section through track prediction processing based on a long-short-time memory network to obtain a predicted track;
It can be understood that the motion trail of the unmanned automobile is predicted in the step, so that the accuracy and safety of navigation are improved. Preferably, the invention divides the key nodes and roads on the driving route into track segments, extracts the characteristics of the track segments as the input of a long-short-time memory network, and learns the historical track data through the long-short-time memory network, thereby predicting the position and the speed of the unmanned vehicle reaching the next key node at the current moment. In the prediction process, specific control road section information such as road passing weight, traffic flow and the like is introduced into the long-short-time memory network in consideration of the special properties of the control road section so as to better reflect the real-time road condition in the control road section. And finally, obtaining the movement track of the unmanned vehicle in the controlled road section according to the prediction result. The track prediction method based on the long-short-time memory network has higher accuracy and reliability in the controlled road section, can effectively cope with complex traffic environment and road condition changes in the controlled road section, and provides important guarantee for safe navigation of unmanned vehicles.
And step 55, generating a final motion control strategy through time-space information fusion processing based on a long-short-time memory network according to the first motion control strategy, the second motion control strategy, the cooperative anti-collision motion control strategy and the predicted track.
It will be appreciated that through the processing of the preceding steps, four different motion control strategies are obtained, namely a first motion control strategy, a second motion control strategy, a coordinated anti-collision motion control strategy and a predicted trajectory. In order to generate the final motion control strategy, the four control strategies need to be subjected to temporal-spatial information fusion processing. The time-space information fusion processing refers to combining different motion control strategies together, and comprehensively considering a plurality of factors such as road conditions, safety, comfort, efficiency and the like according to real-time environment perception information and target task information to obtain a final motion control strategy. Preferably, the present invention is implemented using a deep Q network based spatio-temporal information fusion method. The method comprises the steps of firstly taking four different motion control strategies as four actions, and inputting the four different motion control strategies into a deep Q network to obtain a Q value corresponding to each action. And then, obtaining the optimal motion control strategy by continuously updating parameters of the depth Q network by using a strategy iteration algorithm. And finally, selecting a corresponding optimal action from the depth Q network according to the target task information and the real-time environment perception information, and generating a final motion control strategy. The space-time information fusion processing method can comprehensively consider the influence of different factors on motion control, and can realize dynamic adjustment and optimization according to real-time environment perception information and target task information, so that the unmanned vehicle can finish tasks in an optimal strategy in a controlled road section, and the safety and the efficiency are improved.
Example 2
As shown in fig. 2, the present embodiment provides a safe navigation device for an unmanned automobile under a controlled road section, where the device includes a first acquiring unit, and a marker in the first acquiring unit: 701. a first processing unit 702, a second processing unit 703, a third processing unit 704, and a fourth processing unit 705.
The first acquisition unit is marked in the graph: 701, configured to obtain first information and second information, where the first information includes real-time road condition information, traffic state information, and rescue task information of a controlled road section collected by using a vehicle-mounted communication device and an infrastructure communication device, and the second information includes real-time environmental perception information collected by an unmanned vehicle sensor;
the first processing unit 702 processes the real-time road condition information, the traffic state information and the real-time perception information by using a spatial information fusion mathematical model based on a convolutional neural network, and generates a dynamic semantic map containing road attributes, traffic signs, signals and barrier information;
the second processing unit 703 is configured to perform weight distribution processing on the dynamic semantic map according to the rescue task information and a preset reinforcement learning mathematical model to obtain a priority weight map, where the priority weight map includes a comprehensive weight of a road traffic condition, a traffic congestion degree and a rescue priority;
The third processing unit 704 is configured to perform path planning processing on the priority weight map according to a preset dynamic path planning mathematical model, so as to obtain at least one driving route adapted to the complex environment of the controlled road section;
and a fourth processing unit 705, which performs space-time correlation analysis processing on the dynamic semantic map, the driving route and the real-time environment perception information by using a space-time information fusion mathematical model based on a long-short time memory network to obtain a motion control strategy of the unmanned automobile, wherein the motion control strategy comprises driving parameters of the unmanned automobile.
In one embodiment of the present disclosure, the first processing unit 702 includes:
the first extraction subunit 7021 is configured to perform a road attribute extraction process according to real-time road condition information, so as to obtain road attribute information, where the road attribute information includes a road type, a width and a speed limit;
the first recognition subunit 7022 is configured to perform recognition processing on traffic signs and signal lights according to real-time traffic state information, so as to obtain traffic sign information and signal light information, where the traffic sign information includes a stop sign, a speed limit sign, and a traffic prohibition sign, and the signal light information includes a traffic light state and a countdown time;
A second recognition subunit 7023, configured to perform obstacle recognition processing according to the real-time environmental awareness information, so as to obtain obstacle information, where the obstacle information includes pedestrians, vehicles, and objects;
the first fusion subunit 7024 is configured to fuse the road attribute information, the traffic sign information, the signal lamp information, and the obstacle information according to the spatial information fusion mathematical model, so as to obtain a dynamic semantic map.
In one embodiment of the present disclosure, the first fusion subunit 7024 comprises:
the first processing subunit 70241 is configured to perform road topology processing based on the graph convolutional neural network according to the road attribute information to obtain road attribute data;
the second processing subunit 70242 is configured to assign different weights to different types of traffic signs in the traffic sign information, so as to obtain weighted traffic sign data that matches the real-time road condition;
the third processing subunit 70243 is configured to perform time sequence analysis according to the signal lamp information, and predict a variation trend of the signal lamp state to obtain signal lamp data;
a fourth processing subunit 70244, configured to perform multi-mode fusion method processing of adaptive weights according to the obstacle information, and perform weighted fusion on the obstacle information from different sources to obtain obstacle data;
And a fifth processing subunit 70245, configured to perform space-time sequence fusion processing on the road attribute data, the weighted traffic sign data, the signal lamp data and the obstacle data by using a deep learning method based on the gating cycle unit, so as to obtain a dynamic semantic map.
In one embodiment of the present disclosure, the second processing unit 703 includes:
a first sorting subunit 7031, configured to perform topology sorting processing according to each rescue task in the rescue task information, so as to obtain an ordered rescue task list;
the first dividing subunit 7032 is configured to perform spectral clustering according to the dynamic semantic map, divide the map into a plurality of sub-regions, and obtain a sub-region map list;
a sixth processing subunit 7033, configured to perform heuristic search algorithm-based processing according to the rescue task list, determine an optimal sub-area of each rescue task in the sub-area map list, and obtain a task-sub-area matching relationship;
a first calculating subunit 7034, configured to calculate a rescue task priority of each sub-region according to the task-sub-region matching relationship, so as to obtain a sub-region priority list;
the first allocation subunit 7035 is configured to perform weight allocation on the dynamic semantic map according to the sub-region priority list and a preset depth deterministic strategy gradient mathematical model, so as to obtain a priority weight map.
In one embodiment of the present disclosure, the fourth processing unit 705 includes:
a first analysis subunit 7041, configured to perform analysis processing on the characteristics of the controlled road segments according to the priority weight graph, and extract special attributes of the controlled road segments to obtain the characteristics information of the controlled road segments, where the special attributes include an emergency lane and an area where traffic is prohibited;
a second sequencing subunit 7042, configured to perform sequencing processing based on the priorities of the obstacles and the rescue vehicles according to the characteristic information of the controlled road segments, determine the priorities of the different types of obstacles and the rescue vehicles in the path planning, and obtain the priority information of the obstacles and the rescue vehicles;
a seventh processing subunit 7043, configured to perform mixed integer planning algorithm processing according to the priority weight map and the priority information, solve a possible traffic collision problem, and obtain a collision solution;
a first adjustment subunit 7044, configured to dynamically adjust road traffic weight for the real-time road condition information according to the conflict resolution scheme, so as to obtain real-time road traffic weight information;
the second dividing subunit 7045 is configured to divide the controlled road segment into a plurality of sub-areas according to the real-time road traffic weight information and the preset hierarchical path planning strategy, plan paths in each sub-area, and connect the paths of the sub-areas to obtain at least one driving route adapted to the complex environment of the controlled road segment.
In one embodiment of the present disclosure, the fourth processing unit 705 includes:
an eighth processing subunit 7051, configured to perform multi-objective optimization processing according to the dynamic semantic map, the driving route, and the real-time perception information, balance the traffic efficiency and the safety of the controlled road section, and obtain a first motion control policy of the unmanned automobile;
a ninth processing subunit 7052, configured to perform priority-based path constraint processing according to the driving route and the rescue task information, ensure that the unmanned vehicle follows the driving route and preferentially passes through the control road section in the process of executing the rescue task, and obtain a constrained second motion control policy;
a tenth processing subunit 7053, configured to perform cooperative anti-collision processing according to the real-time environmental awareness information, and analyze the motion states of surrounding vehicles and obstacles to obtain a cooperative anti-collision motion control strategy;
the first prediction subunit 7054 is configured to predict, according to the dynamic semantic map and the driving route, a motion track of the unmanned vehicle in the controlled road section through track prediction processing based on the long-short-term memory network, so as to obtain a predicted track;
the second fusion subunit 7055 is configured to generate a final motion control policy according to the first motion control policy, the second motion control policy, the cooperative anti-collision motion control policy, and the predicted trajectory through temporal and spatial information fusion processing based on the long and short time memory network. The above is only a preferred embodiment of the present invention, and is not intended to limit the present invention, but various modifications and variations can be made to the present invention by those skilled in the art. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
The foregoing is merely illustrative embodiments of the present invention, but the scope of the present invention is not limited thereto, and any person skilled in the art can easily think about variations or substitutions within the technical scope of the present invention, and the invention should be covered. Therefore, the protection scope of the invention is subject to the protection scope of the claims.
Claims (4)
1. The unmanned automobile safety navigation method under the control road section is characterized by comprising the following steps of:
acquiring first information and second information, wherein the first information comprises real-time road condition information, traffic state information and rescue task information of a controlled road section collected by adopting vehicle-mounted communication equipment and infrastructure communication equipment, and the second information comprises real-time environment perception information collected by an unmanned vehicle sensor;
processing the real-time road condition information, the traffic state information and the second information by using a space information fusion mathematical model based on a convolutional neural network to generate a dynamic semantic map containing road attributes, traffic signs, signals and barrier information;
carrying out weight distribution processing on the dynamic semantic map according to the rescue task information and a preset reinforcement learning mathematical model to obtain a priority weight map, wherein the priority weight map comprises comprehensive weights of road traffic conditions, traffic jam degree and rescue priority;
Carrying out path planning processing on the priority weight graph according to a preset dynamic path planning mathematical model to obtain at least one driving route adapting to the complex environment of the control road section;
performing space-time correlation analysis processing on the dynamic semantic map, the driving route and the second information by using a space-time information fusion mathematical model based on a long-short time memory network to obtain a motion control strategy of the unmanned automobile, wherein the motion control strategy comprises driving parameters of the unmanned automobile;
the method for generating the dynamic semantic map comprising road attributes, traffic signs, signals and barrier information comprises the following steps of:
extracting road attribute according to the real-time road condition information to obtain road attribute information, wherein the road attribute information comprises road type, width and speed limit;
carrying out traffic sign and signal lamp identification processing according to the traffic state information to obtain traffic sign information and signal lamp information, wherein the traffic sign information comprises a stop sign, a speed limit sign and a traffic prohibition sign, and the signal lamp information comprises a traffic light state and a countdown time;
Performing obstacle recognition processing according to the second information to obtain obstacle information, wherein the obstacle information comprises pedestrians, vehicles and objects;
carrying out fusion processing on the road attribute information, the traffic sign information, the signal lamp information and the obstacle information according to the space information fusion mathematical model to obtain a dynamic semantic map;
the method for obtaining the dynamic semantic map comprises the following steps of:
carrying out road topological structure processing based on a graph convolution neural network according to the road attribute information to obtain road attribute data;
different weights are given to different types of traffic signs in the traffic sign information, and weighted traffic sign data matched with real-time road conditions are obtained;
performing time sequence analysis according to the signal lamp information, and predicting the change trend of the signal lamp state to obtain signal lamp data;
performing multi-mode fusion method processing of self-adaptive weights according to the obstacle information, and performing weighted fusion on the obstacle information from different sources to obtain obstacle data;
Performing space-time sequence fusion processing on the road attribute data, the weighted traffic sign data, the signal lamp data and the obstacle data by using a deep learning method based on a gating circulating unit to obtain a dynamic semantic map;
the method comprises the steps of carrying out path planning processing on the priority weight graph according to a preset dynamic path planning mathematical model to obtain at least one driving route adapting to the complex environment of the control road section, and comprises the following steps:
performing control road section characteristic analysis processing according to the priority weight graph, and extracting special attributes of a control area to obtain control road section characteristic information, wherein the special attributes comprise an emergency lane and an area where traffic is forbidden;
performing priority ranking processing based on the obstacles and the rescue vehicles according to the characteristic information of the control road section, determining the priorities of the different types of obstacles and the rescue vehicles in path planning, and obtaining priority information of the obstacles and the rescue vehicles;
according to the priority weight graph and the priority information, mixed integer programming algorithm processing is carried out, the possible passing conflict problem is solved, and a conflict solution is obtained;
dynamically adjusting the road traffic weight according to the real-time road condition information according to the conflict solution to obtain real-time road traffic weight information;
Dividing the control road section into a plurality of subareas according to the real-time road traffic weight information and a preset hierarchical path planning strategy, respectively planning paths in each subarea, and connecting the paths of the subareas to obtain at least one driving route adapting to the complex environment of the control road section.
2. The method for safely navigating an unmanned car under a controlled road section according to claim 1, wherein the step of performing weight distribution processing on the dynamic semantic map according to the rescue task information and a preset reinforcement learning mathematical model to obtain a priority weight map comprises the following steps:
performing topology sequencing processing according to each rescue task in the rescue task information to obtain an ordered rescue task list;
performing spectral clustering processing according to the dynamic semantic map, dividing the map into a plurality of subareas, and obtaining a subarea map list;
processing based on heuristic search algorithm is carried out according to the rescue task list, and an optimal subarea of each rescue task in the subarea map list is determined, so that a task-subarea matching relationship is obtained;
calculating rescue task priorities of all subareas according to the task-subarea matching relationship to obtain a subarea priority list;
And carrying out weight distribution on the dynamic semantic map according to the sub-region priority list and a preset depth certainty strategy gradient mathematical model to obtain a priority weight map.
3. An unmanned car safety navigation device under a controlled road section, comprising:
the first acquisition unit is used for acquiring first information and second information, wherein the first information comprises real-time road condition information, traffic state information and rescue task information of a controlled road section collected by adopting vehicle-mounted communication equipment and infrastructure communication equipment, and the second information comprises real-time environment perception information collected by an unmanned vehicle sensor;
the first processing unit is used for processing the real-time road condition information, the traffic state information and the second information by using a space information fusion mathematical model based on a convolutional neural network, and generating a dynamic semantic map containing road attributes, traffic signs, signals and barrier information;
the second processing unit is used for carrying out weight distribution processing on the dynamic semantic map according to the rescue task information and a preset reinforcement learning mathematical model to obtain a priority weight map, wherein the priority weight map comprises comprehensive weights of road traffic conditions, traffic jam degree and rescue priority;
The third processing unit is used for carrying out path planning processing on the priority weight graph according to a preset dynamic path planning mathematical model to obtain at least one driving route which is suitable for the complex environment of the controlled road section;
the fourth processing unit is used for carrying out space-time correlation analysis processing on the dynamic semantic map, the driving route and the second information by using a space-time information fusion mathematical model based on a long-short time memory network to obtain a motion control strategy of the unmanned automobile, wherein the motion control strategy comprises driving parameters of the unmanned automobile;
wherein the first processing unit includes:
the first extraction subunit is used for extracting road attribute according to the real-time road condition information to obtain road attribute information, wherein the road attribute information comprises a road type, a width and a speed limit;
the first identification subunit is used for carrying out identification processing on traffic signs and signal lamps according to the traffic state information to obtain traffic sign information and signal lamp information, wherein the traffic sign information comprises a stop sign, a speed limit sign and a traffic prohibition sign, and the signal lamp information comprises a traffic light state and a countdown time;
The second recognition subunit is used for carrying out obstacle recognition processing according to the second information to obtain obstacle information, wherein the obstacle information comprises pedestrians, vehicles and objects;
the first fusion subunit is used for carrying out fusion processing on the road attribute information, the traffic sign information, the signal lamp information and the barrier information according to the spatial information fusion mathematical model to obtain a dynamic semantic map;
wherein the first fusion subunit comprises:
the first processing subunit is used for processing the road topological structure based on the graph convolution neural network according to the road attribute information to obtain road attribute data;
the second processing subunit is used for giving different weights to different types of traffic signs in the traffic sign information to obtain weighted traffic sign data matched with real-time road conditions;
the third processing subunit is used for carrying out time sequence analysis according to the signal lamp information and predicting the change trend of the signal lamp state to obtain signal lamp data;
the fourth processing subunit is used for carrying out multi-mode fusion method processing of self-adaptive weights according to the obstacle information, and carrying out weighted fusion on the obstacle information from different sources to obtain obstacle data;
A fifth processing subunit, configured to perform space-time sequence fusion processing on the road attribute data, the weighted traffic sign data, the signal lamp data, and the obstacle data by using a deep learning method based on a gating cycle unit, so as to obtain a dynamic semantic map;
wherein the third processing unit includes:
the first analysis subunit is used for carrying out control road section characteristic analysis processing according to the priority weight graph, extracting special attributes of a control area to obtain control road section characteristic information, wherein the special attributes comprise an emergency lane and an area where traffic is forbidden;
the second sequencing subunit is used for sequencing the priority of the obstacle-based rescue vehicles according to the characteristic information of the control road section, determining the priority of the obstacle-based rescue vehicles in the path planning, and obtaining the priority information of the obstacle-based rescue vehicles;
a seventh processing subunit, configured to perform mixed integer programming algorithm processing according to the priority weight map and the priority information, solve a possible traffic collision problem, and obtain a collision solution;
the first adjusting subunit is used for dynamically adjusting the road traffic weight according to the real-time road condition information according to the conflict solution to obtain real-time road traffic weight information;
The second dividing subunit is used for dividing the control road section into a plurality of subareas according to the real-time road traffic weight information and a preset hierarchical path planning strategy, planning paths in each subarea respectively, and connecting the paths of the subareas to obtain at least one driving route adapting to the complex environment of the control road section.
4. A controlled section unmanned car safety navigation device according to claim 3, wherein the second processing unit comprises:
the first sequencing subunit is used for performing topology sequencing processing according to each rescue task in the rescue task information to obtain an ordered rescue task list;
the first dividing subunit is used for carrying out spectral clustering processing according to the dynamic semantic map, dividing the map into a plurality of subareas and obtaining a subarea map list;
the sixth processing subunit is used for carrying out heuristic search algorithm-based processing according to the rescue task list, determining the optimal subarea of each rescue task in the subarea map list and obtaining a task-subarea matching relationship;
the first calculating subunit is used for calculating rescue task priorities of all the subareas according to the task-subarea matching relation to obtain a subarea priority list;
The first allocation subunit is used for carrying out weight allocation on the dynamic semantic map according to the sub-region priority list and a preset depth deterministic strategy gradient mathematical model to obtain a priority weight map.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311067417.5A CN116804560B (en) | 2023-08-23 | 2023-08-23 | Unmanned automobile safety navigation method and device under controlled road section |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311067417.5A CN116804560B (en) | 2023-08-23 | 2023-08-23 | Unmanned automobile safety navigation method and device under controlled road section |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN116804560A CN116804560A (en) | 2023-09-26 |
| CN116804560B true CN116804560B (en) | 2023-11-03 |
Family
ID=88079718
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311067417.5A Active CN116804560B (en) | 2023-08-23 | 2023-08-23 | Unmanned automobile safety navigation method and device under controlled road section |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116804560B (en) |
Families Citing this family (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117371750B (en) * | 2023-11-06 | 2024-05-17 | 南京工程学院 | Multi-target power grid fault emergency rescue method considering man-vehicle feature matching |
| CN117196266B (en) * | 2023-11-07 | 2024-01-23 | 成都工业职业技术学院 | Unmanned shared automobile area scheduling method and device based on neural network |
| CN117705090A (en) * | 2023-12-20 | 2024-03-15 | 南京智绘星图信息科技有限公司 | Comprehensive control method of high-precision map |
| CN117804490B (en) * | 2024-02-28 | 2024-05-17 | 四川交通职业技术学院 | Comprehensive planning method and device for vehicle running route |
Citations (18)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106627421A (en) * | 2016-11-21 | 2017-05-10 | 国网山东省电力公司电力科学研究院 | Expressway rescue system with forewarning analysis function for electric vehicles and method of the same |
| CN109492065A (en) * | 2018-10-26 | 2019-03-19 | 桂林电子科技大学 | A kind of abstracting method of indoor semantic map time-space relationship |
| CN111104849A (en) * | 2018-10-29 | 2020-05-05 | 安波福技术有限公司 | Automatic annotation of environmental features in a map during navigation of a vehicle |
| EP3647733A1 (en) * | 2018-10-29 | 2020-05-06 | Aptiv Technologies Limited | Automatic annotation of environmental features in a map during navigation of a vehicle |
| WO2020119684A1 (en) * | 2018-12-14 | 2020-06-18 | 中国科学院深圳先进技术研究院 | 3d navigation semantic map update method, apparatus and device |
| DE102019001899A1 (en) * | 2019-03-18 | 2020-09-24 | Rainer Truöl | Device for the formation of a so-called rescue lane on a vehicle or similar. Congested freeway or street with structurally separated traffic lanes for each direction of travel and facility for variable speed control on these streets |
| US11145076B1 (en) * | 2020-10-27 | 2021-10-12 | R-Go Robotics Ltd | Incorporation of semantic information in simultaneous localization and mapping |
| CN113674416A (en) * | 2021-08-26 | 2021-11-19 | 中国电子科技集团公司信息科学研究院 | Three-dimensional map construction method and device, electronic equipment and storage medium |
| CA3138062A1 (en) * | 2020-11-20 | 2022-05-20 | Rapyuta Robotics Co., Ltd. | Systems and methods for optimizing route plans in an operating environment |
| CN114882182A (en) * | 2022-04-22 | 2022-08-09 | 东南大学 | Semantic map construction method based on vehicle-road cooperative sensing system |
| CN217629570U (en) * | 2022-06-10 | 2022-10-21 | 四川交通职业技术学院 | Road surface warning device for traffic rescue |
| EP4095486A1 (en) * | 2021-05-11 | 2022-11-30 | X Development LLC | Systems and methods for navigating a robot using semantic mapping |
| CN115503694A (en) * | 2022-10-20 | 2022-12-23 | 北京易航远智科技有限公司 | Autonomous learning-based memory parking path generation method and device and electronic equipment |
| CN115578876A (en) * | 2022-10-14 | 2023-01-06 | 浪潮(北京)电子信息产业有限公司 | Automatic driving method, system, equipment and storage medium of vehicle |
| CN115640628A (en) * | 2022-09-29 | 2023-01-24 | 福建师范大学地理研究所 | Design method of indoor fire-fighting map based on space cognition |
| EP4134623A1 (en) * | 2021-08-11 | 2023-02-15 | Hitachi Astemo, Ltd. | Drive device, vehicle, and method for automated driving and/or assisted driving |
| WO2023087181A1 (en) * | 2021-11-17 | 2023-05-25 | 浙江吉利控股集团有限公司 | Vehicle-road collaborative multi-vehicle path planning and right-of-way decision method and system, and roadbed unit |
| CN116266380A (en) * | 2021-12-13 | 2023-06-20 | 广州汽车集团股份有限公司 | Environment data reconstruction method, device, system and storage medium |
Family Cites Families (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US11041730B2 (en) * | 2018-03-02 | 2021-06-22 | DeepMap Inc. | Reverse rendering of an image based on high definition map data |
| CN109703568B (en) * | 2019-02-19 | 2020-08-18 | 百度在线网络技术（北京）有限公司 | Method, device and server for learning driving strategy of automatic driving vehicle in real time |
| US20210191397A1 (en) * | 2019-12-24 | 2021-06-24 | Industrial Technology Research Institute | Autonomous vehicle semantic map establishment system and establishment method |
| US11238650B2 (en) * | 2020-03-13 | 2022-02-01 | Nvidia Corporation | Self-supervised single-view 3D reconstruction via semantic consistency |
| US11981347B2 (en) * | 2021-03-31 | 2024-05-14 | Gm Cruise Holdings Llc | Rules based semantic map sharding |
| US11526535B2 (en) * | 2021-03-31 | 2022-12-13 | Gm Cruise Holdings Llc | System and method for generating shard data based on semantic map sharding |
-
2023
- 2023-08-23 CN CN202311067417.5A patent/CN116804560B/en active Active
Patent Citations (18)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106627421A (en) * | 2016-11-21 | 2017-05-10 | 国网山东省电力公司电力科学研究院 | Expressway rescue system with forewarning analysis function for electric vehicles and method of the same |
| CN109492065A (en) * | 2018-10-26 | 2019-03-19 | 桂林电子科技大学 | A kind of abstracting method of indoor semantic map time-space relationship |
| CN111104849A (en) * | 2018-10-29 | 2020-05-05 | 安波福技术有限公司 | Automatic annotation of environmental features in a map during navigation of a vehicle |
| EP3647733A1 (en) * | 2018-10-29 | 2020-05-06 | Aptiv Technologies Limited | Automatic annotation of environmental features in a map during navigation of a vehicle |
| WO2020119684A1 (en) * | 2018-12-14 | 2020-06-18 | 中国科学院深圳先进技术研究院 | 3d navigation semantic map update method, apparatus and device |
| DE102019001899A1 (en) * | 2019-03-18 | 2020-09-24 | Rainer Truöl | Device for the formation of a so-called rescue lane on a vehicle or similar. Congested freeway or street with structurally separated traffic lanes for each direction of travel and facility for variable speed control on these streets |
| US11145076B1 (en) * | 2020-10-27 | 2021-10-12 | R-Go Robotics Ltd | Incorporation of semantic information in simultaneous localization and mapping |
| CA3138062A1 (en) * | 2020-11-20 | 2022-05-20 | Rapyuta Robotics Co., Ltd. | Systems and methods for optimizing route plans in an operating environment |
| EP4095486A1 (en) * | 2021-05-11 | 2022-11-30 | X Development LLC | Systems and methods for navigating a robot using semantic mapping |
| EP4134623A1 (en) * | 2021-08-11 | 2023-02-15 | Hitachi Astemo, Ltd. | Drive device, vehicle, and method for automated driving and/or assisted driving |
| CN113674416A (en) * | 2021-08-26 | 2021-11-19 | 中国电子科技集团公司信息科学研究院 | Three-dimensional map construction method and device, electronic equipment and storage medium |
| WO2023087181A1 (en) * | 2021-11-17 | 2023-05-25 | 浙江吉利控股集团有限公司 | Vehicle-road collaborative multi-vehicle path planning and right-of-way decision method and system, and roadbed unit |
| CN116266380A (en) * | 2021-12-13 | 2023-06-20 | 广州汽车集团股份有限公司 | Environment data reconstruction method, device, system and storage medium |
| CN114882182A (en) * | 2022-04-22 | 2022-08-09 | 东南大学 | Semantic map construction method based on vehicle-road cooperative sensing system |
| CN217629570U (en) * | 2022-06-10 | 2022-10-21 | 四川交通职业技术学院 | Road surface warning device for traffic rescue |
| CN115640628A (en) * | 2022-09-29 | 2023-01-24 | 福建师范大学地理研究所 | Design method of indoor fire-fighting map based on space cognition |
| CN115578876A (en) * | 2022-10-14 | 2023-01-06 | 浪潮(北京)电子信息产业有限公司 | Automatic driving method, system, equipment and storage medium of vehicle |
| CN115503694A (en) * | 2022-10-20 | 2022-12-23 | 北京易航远智科技有限公司 | Autonomous learning-based memory parking path generation method and device and electronic equipment |
Non-Patent Citations (5)
| Title |
|---|
| Building a grid-semantic map for the navigation of service robots through human–robot interaction;Zhao, C.等;Digital Communications and Networks;第1卷(第4期);253–266 * |
| 城市路网交通管制下的最优路径诱导仿真研究;杨泳;户佐安;;计算机仿真;30(7);147-150 * |
| 基于QR code技术的家庭半未知环境语义地图构建;吴皓, 田国会, 薛英花等.;模式识别与人工智能;第23卷(第4期);464-470 * |
| 泥石流灾害应急救援辅助决策研究——以小金县城隍庙沟泥石流为例;唐尧, 王立娟, 邓琮等;国土资源信息化(第3期);54-60 * |
| 道路交通中攻击性驾驶行为分析与研究;骆勇;中国博士学位论文全文数据库工程科技Ⅱ辑(第3期);C034-30 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN116804560A (en) | 2023-09-26 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN116804560B (en) | Unmanned automobile safety navigation method and device under controlled road section | |
| CN109466543B (en) | Planning autonomous movement | |
| WO2021204092A1 (en) | Track prediction method and apparatus, and device and storage medium | |
| Wei et al. | A survey on traffic signal control methods | |
| US11565716B2 (en) | Method and system for dynamically curating autonomous vehicle policies | |
| JP2023510136A (en) | Geolocation models for perception, prediction or planning | |
| US20220105959A1 (en) | Methods and systems for predicting actions of an object by an autonomous vehicle to determine feasible paths through a conflicted area | |
| KR20150128712A (en) | Lane-level vehicle navigation for vehicle routing and traffic management | |
| Shimosaka et al. | Modeling risk anticipation and defensive driving on residential roads with inverse reinforcement learning | |
| Hao et al. | Reducing CACC platoon disturbances caused by state jitters by combining two stages driving state recognition with multiple platoons’ strategies and risk prediction | |
| Nishi et al. | Fine-grained driving behavior prediction via context-aware multi-task inverse reinforcement learning | |
| KR20230173724A (en) | Improving ride comfort in various traffic scenarios for autonomous vehicles | |
| Emzivat et al. | A formal approach for the design of a dependable perception system for autonomous vehicles | |
| EP3967978B1 (en) | Detecting a construction zone by a lead autonomous vehicle (av) and updating routing plans for following autonomous vehicles (avs) | |
| Lakshna et al. | Smart Traffic: Traffic Congestion Reduction by Shortest Route* Search Algorithm | |
| CN117529429A (en) | Method and system for predicting behavior of an actor in an autonomous vehicle environment | |
| Habibi et al. | Improvement of Multi-agent Routing Guidance with an Intelligent Traffic Light Scheduling and the Ability to Select Intermediate Destinations | |
| Jose et al. | A New approach for accurate path Prediction using multiple Prediction system | |
| CN115503752A (en) | Method and system for operating a vehicle | |
| US20230154317A1 (en) | Method for controlling a mobility system, data processing device, computer-readable medium, and system for controlling a mobility system | |
| US20240028035A1 (en) | Planning autonomous motion | |
| US20230415767A1 (en) | Protolanes for testing autonomous vehicle intent | |
| US20230415773A1 (en) | Validating protolanes with actor behavior flowcharts | |
| US20230418291A1 (en) | Testing reuse across protolane families | |
| Ahmed | Role of smart vehicles concept in reducing traffic congestion on the road |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |