CN115775463A - Navigation method for automatically driving automobile - Google Patents
Navigation method for automatically driving automobile Download PDFInfo
- Publication number
- CN115775463A CN115775463A CN202211329153.1A CN202211329153A CN115775463A CN 115775463 A CN115775463 A CN 115775463A CN 202211329153 A CN202211329153 A CN 202211329153A CN 115775463 A CN115775463 A CN 115775463A
- Authority
- CN
- China
- Prior art keywords
- vehicle
- information
- current
- potential risk
- distance
- 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.)
- Withdrawn
Links
- 238000000034 method Methods 0.000 title claims abstract description 37
- 230000002159 abnormal effect Effects 0.000 claims abstract description 39
- 238000012545 processing Methods 0.000 claims abstract description 22
- 230000009471 action Effects 0.000 claims abstract description 8
- 230000033001 locomotion Effects 0.000 claims description 27
- 238000001514 detection method Methods 0.000 claims description 18
- 230000001133 acceleration Effects 0.000 claims description 17
- 230000002452 interceptive effect Effects 0.000 claims description 16
- 230000003044 adaptive effect Effects 0.000 claims description 11
- 230000007246 mechanism Effects 0.000 claims description 11
- 230000005856 abnormality Effects 0.000 claims description 10
- 238000004364 calculation method Methods 0.000 claims description 10
- 230000008859 change Effects 0.000 claims description 10
- 230000004927 fusion Effects 0.000 claims description 9
- 238000011156 evaluation Methods 0.000 claims description 8
- 230000008439 repair process Effects 0.000 claims description 8
- 238000003745 diagnosis Methods 0.000 claims description 7
- 238000000605 extraction Methods 0.000 claims description 7
- 230000000087 stabilizing effect Effects 0.000 claims description 6
- 230000002068 genetic effect Effects 0.000 claims description 5
- 239000011159 matrix material Substances 0.000 claims description 3
- 230000003287 optical effect Effects 0.000 claims description 3
- 239000011241 protective layer Substances 0.000 claims description 3
- 230000035484 reaction time Effects 0.000 claims description 3
- 238000012502 risk assessment Methods 0.000 claims description 3
- 238000012549 training Methods 0.000 claims description 3
- 230000009466 transformation Effects 0.000 claims description 3
- 230000001502 supplementing effect Effects 0.000 claims description 2
- 230000006399 behavior Effects 0.000 description 10
- 239000010410 layer Substances 0.000 description 8
- 206010039203 Road traffic accident Diseases 0.000 description 4
- 238000011161 development Methods 0.000 description 4
- 230000006870 function Effects 0.000 description 4
- 230000008569 process Effects 0.000 description 4
- 238000005516 engineering process Methods 0.000 description 3
- 230000003993 interaction Effects 0.000 description 3
- 238000004458 analytical method Methods 0.000 description 2
- 238000013473 artificial intelligence Methods 0.000 description 2
- 238000004891 communication Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 230000007613 environmental effect Effects 0.000 description 2
- 230000003068 static effect Effects 0.000 description 2
- 206010063385 Intellectualisation Diseases 0.000 description 1
- 230000004913 activation Effects 0.000 description 1
- 230000001154 acute effect Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 230000033228 biological regulation Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 230000006835 compression Effects 0.000 description 1
- 238000007906 compression Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000003912 environmental pollution Methods 0.000 description 1
- 230000003137 locomotive effect Effects 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000006855 networking Effects 0.000 description 1
- 230000008447 perception Effects 0.000 description 1
- 238000011176 pooling Methods 0.000 description 1
- 230000001737 promoting effect Effects 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000012827 research and development Methods 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 230000006641 stabilisation Effects 0.000 description 1
- 238000011105 stabilization Methods 0.000 description 1
Images
Classifications
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
- Y02T10/00—Road transport of goods or passengers
- Y02T10/10—Internal combustion engine [ICE] based vehicles
- Y02T10/40—Engine management systems
Landscapes
- Traffic Control Systems (AREA)
Abstract
The invention relates to the technical field of unmanned driving, in particular to a navigation method for automatically driving an automobile. The method comprises the following steps: the method comprises the steps of obtaining real-time picture information around a current vehicle, obtaining vehicle and pedestrian information of a lane adjacent to the current vehicle, carrying out local decision-making picture enhancement processing on the obtained vehicle real-time picture information to obtain an enhanced key image, carrying out safety distance marking on the adjacent vehicle, and obtaining abnormal driving vehicle behavior prediction information. The invention carries out the action of interactively predicting the navigation track of the vehicle by planning the interactively predicted navigation track of the vehicle so as to improve the driving safety and the comfort of passengers.
Description
Technical Field
The invention relates to the technical field of automatic driving, in particular to a navigation method for an automatic driving automobile.
Background
The progress of the automobile industry remarkably improves the life of common people, benefits the life, facilitates the transportation and the trip, and also plays a great role in promoting the development of the whole national economy. It should be noted that there are many "side effects" that follow as the automobile industry develops, such as environmental problems, energy problems, and problems of traffic congestion, traffic safety, etc. are becoming more acute. Therefore, more and more national government agencies and automobile manufacturers start to conduct policy guidance and strategic direction adjustment on future development of the automobile industry, many countries develop policy regulations which are good for research and development and production of new energy automobiles, and corresponding manufacturers actively respond to policy calls, and a great deal of leading-edge technical research is developed in the technical field of new energy.
The progress of the automobile industry is remarkably convenient for the life of modern people, but the popularization of automobiles also causes problems such as traffic jam, environmental pollution and the like, so that the automobiles tend to develop towards more intellectualization, networking and electromotion in the future. The unmanned driving of the vehicle is used as the final form of the automobile science and technology development, and the path tracking control is the technical premise for realizing the real unmanned driving of the vehicle. The automatic driving technology of the vehicle means that the vehicle can realize information perception of a driving environment by using sensing technologies such as vision, laser and radar, combines environmental information such as real-time road conditions, traffic conditions and pedestrian states, obtains reasonable driving paths and vehicle motion control information through high-speed calculation processing core planning, and is delivered to a bottom layer motion control system comprising a steering system, a driving/braking system and the like to finally finish automatic driving operation of the vehicle.
Aiming at the increasingly serious road traffic safety at present, the unmanned automobile is developed rapidly by means of the rapid development of artificial intelligence. In real road scenarios, it is a great challenge to consider the prediction of interaction trajectories between surrounding vehicles in the planning of the motion of an autonomous vehicle.
Disclosure of Invention
The invention provides a navigation method for automatically driving an automobile, which aims to solve at least one technical problem.
The invention provides a navigation method for automatically driving an automobile, which comprises the following steps:
acquiring real-time picture information around the current vehicle through a monocular camera;
acquiring binocular real-time picture information through a binocular camera, wherein the binocular real-time picture information comprises information of vehicles and pedestrians adjacent to a current vehicle in a lane;
carrying out local decision-level picture enhancement processing on the obtained real-time picture information around the current vehicle to obtain an enhanced key image;
calculating to obtain the distance between the vehicles and the pedestrians in the adjacent lanes in the binocular real-time picture information, and marking the safety distance of the vehicles in the adjacent lanes according to the highest speed per hour of the adjacent lanes;
analyzing vehicles in the adjacent lanes and vehicles in front of the current vehicle in the same lane by using a track prediction model to obtain abnormal running vehicle information influencing safe driving, and correspondingly predicting the running route of the abnormal running vehicle information to obtain abnormal running vehicle behavior prediction information;
detecting and processing the enhanced key image to obtain key information in the enhanced key image, extracting key place names and traffic indication information in the key information, extracting text information in the key place names and the traffic indication information, and marking the text information as emergency navigation text information; marking the shelters around the current vehicle, and recording as the potential risk estimation information;
performing latent risk judgment on the latent risk estimation information by using a latent risk evaluation detection model to obtain final latent risk information;
detecting the current satellite condition, and judging and repairing the reason of the satellite abnormality by using a satellite repairing model when the satellite is detected to be in an abnormal state to obtain accurate satellite positioning;
when the repair is failed, performing data fusion on the emergency LBS positioning and the drawn route map which is driven in an abnormal state by using an emergency positioning algorithm in the navigation system so as to calculate the current repairability accurate position positioning;
planning an interactive predicted navigation track of the vehicle by utilizing accurate satellite positioning or repairability accurate position positioning and combining abnormal running vehicle behavior prediction information and potential risk prediction information;
and judging the motion control mechanism of the current vehicle and the state of the current vehicle by using the motion control detection mechanism, and performing the action of interactively predicting the navigation track of the vehicle by using an adaptive uncertain intelligent vehicle steering stabilizing system.
Optionally, the step of performing local decision-level picture enhancement processing on the obtained vehicle real-time picture information to obtain an enhanced key image includes the following steps:
pre-extracting key information from the current real-time picture information around the vehicle by using a key information pre-extraction network to obtain preliminary key information;
and carrying out local decision-level picture enhancement processing on the primary key information obtained by the real-time picture information of the vehicle by using the picture enhancement model to obtain an enhanced key image.
Optionally, the step of obtaining the distance between the vehicle and the pedestrian in the binocular real-time image information through calculation, and marking the safe distance of the vehicle at the highest speed per hour of the vehicle road of the adjacent lane comprises the following steps:
the distance from the current vehicle to the vehicle behind the adjacent lane is calculated by using the following formula:
wherein P is the position of the target, Z L Distance of left camera to target, Z R Distance of right camera to target, O L Is the center point of the aperture of the left camera, O R Is the center point of the aperture of the right camera, D is the distance between the optical centers of the two cameras, F is the focal length of the cameras, P L And P R Respectively projecting the object to the points on the camera photosensitive element through the central points of the diaphragms of the left camera and the right camera; p is L ' and P R ' is the distance from the outermost protective layer of the left and right cameras to the center point of the aperture of the left and right cameras, P L ' and P R ' abscissa is X L And X R Y is the linear distance from the current camera to the tail of the vehicle;
the vehicles are safety distance marked at the highest speed per hour of the vehicle road in the adjacent lane, and in conjunction with the current weather conditions.
Optionally, the step of analyzing the vehicle in front of the adjacent lane and the current vehicle in the same lane by using the trajectory prediction model to obtain information of the abnormally-traveling vehicle affecting safe driving, and predicting the traveling route of the abnormally-traveling vehicle accordingly, wherein the step of obtaining the behavior prediction information of the abnormally-traveling vehicle comprises the following steps:
detecting and processing the enhanced key images to obtain states of the road sign and the traffic lights, interpreting semantics of text information in the road sign through a road traffic self-attention model in combination with road traffic conditions, thereby reordering or supplementing the text information contents and marking the text information contents as emergency navigation text information;
and classifying the vehicles, the obstacles on the road surface and the pedestrians appearing in the enhanced key image, marking the sheltered vehicles, the obstacles on the current lane and the pedestrians with the distance less than the safe distance threshold value, and recording as the potential risk prediction information.
Optionally, detecting the enhanced key image to obtain key information in the image, extracting a key place name and traffic indication information, extracting text information in the image, and marking the text information as emergency navigation text information; the method for marking the shielding object row around as the latent risk estimation information comprises the following steps:
extracting the potential risk estimation information, and utilizing a potential risk evaluation detection model to evaluate and calculate a potential risk expected value:
wherein
For the expected value of the potential risk, H is the distance between the vehicle and the vehicle in front of the through lane, W is the environment adaptive value, G is the number of lanes on the road where the vehicle is located, and N is the number of lanes on the road where the vehicle is locatedThe front traffic flow level value, V is the speed of the current vehicle, V0 is the highest hourly speed of the current lane compliance, VT is the speed of the vehicle behind the adjacent lane, and I is the inclination angle of the vehicle head behind the adjacent lane;
the W environment adaptive value is the probability of the emergency situation according to the semantic judgment of the emergency navigation text information and the vehicle driving experience, and is calculated by a naive Bayes classifier;
marking risk events larger than a potential risk expectation value threshold as preliminary potential risk information by using the formula;
calculating the loss value of each potential risk event by using a loss calculation formula of the potential risk assessment detection model:
where β is the loss value for each risk potential event, μ is the number of turns of the lane that need to be crossed to avoid the risk time, γ is the reaction time to avoid the risk,
the method comprises the steps that a preset dynamic change value of a potential risk scene is obtained, omega is a predicted economic loss cost, and alpha is a preset collision weight value of different objects;
and comparing the loss values of the potential risk events to the preliminary potential risk information, so as to confirm that the preliminary potential risk information with the largest loss value is the final potential risk information.
Optionally, the step of determining and repairing the cause of the satellite abnormality includes the following steps:
judging the satellite signal to be an abnormal reason caused by physical failure or malicious network attack through historical training experience, and judging the type of the current satellite abnormality;
through a satellite diagnosis and repair algorithm, multi-channel data fusion, feature extraction and abnormal diagnosis result output are carried out on the sensor data of the GPS and the LIDAR, so that the vehicle information suffering from malicious network attack is repaired.
Alternatively, the step of drawing a route map for traveling in the abnormal state includes the steps of:
extracting the characteristic points obtained by each frame by using a LiDAR odometer, establishing a characteristic point coordinate system, and placing the subsequent characteristic points in the coordinate system;
and matching the current characteristic points extracted from the current coordinate system of the LiDAR odometer with the characteristic point set extracted from the previous frame, solving a relative pose transformation matrix and calculating pose information so as to obtain a route map which is driven in an abnormal state.
Optionally, when the repair fails, the step of performing data fusion on the emergency LBS location and the route map drawn in the abnormal state by using an emergency location algorithm in the navigation system includes the following steps:
calibrating the time of the two positioning systems, respectively placing the LBS positioning position and the calculated pose information in a navigation map, and drawing two corresponding preliminary motion trail maps;
and comparing the positions of the unified time points of the two preliminary motion trail maps, and performing parameter calculation on the map positions with the distance difference larger than a set threshold value to obtain an accurate positioning result.
Optionally, the step of planning the interactive predicted navigation track of the vehicle by combining the abnormal driving vehicle behavior prediction information and the potential risk prediction information includes the following steps:
by utilizing a motion planning algorithm, taking a lane central line as a reference line, taking the road width as boundary constraint, establishing a global Cartesian coordinate system, and drawing a corresponding vehicle predicted navigation track in the coordinate system;
and correcting the predicted navigation track of the vehicle by combining the abnormal running vehicle behavior prediction information and the potential risk prediction information, and planning the interactive predicted navigation track of the vehicle.
Optionally, a motion control detection mechanism is used for judging the motion control mechanism of the current vehicle and the state of the current vehicle, and an adaptive uncertain intelligent vehicle steering stabilizing system is used, so that the action of interactively predicting the navigation track of the vehicle can be better executed;
detecting the state of the tire by utilizing a motion control detection mechanism according to the real-time rotation of a vehicle chassis gear;
optimizing parameters of the controller by using a genetic algorithm, inputting a difference value between a reference speed and a current speed in a vehicle interactive prediction navigation track to obtain a required acceleration, and using the acceleration as a unified acceleration signal of a brake and an accelerator;
by using an intelligent vehicle steering stabilizing system with uncertain adaptability, combining a unified acceleration signal and a vehicle interactive prediction navigation track, and performing coordinated control on each system of a chassis through a controller, the stable control on tires according to the states of different tires is realized.
The embodiment of the invention aims to optimize the parameters of the controller by using a genetic algorithm and perform online approximation on the unknown nonlinearity of the system by considering the influence of the unknown nonlinearity on the control performance, thereby improving the practicability of the controller, reducing the occurrence of traffic accidents caused by the controller and further improving the experience of passengers.
Drawings
FIGS. 1-2 are schematic flow charts illustrating steps of a navigation method for an autonomous vehicle according to an embodiment of the present invention;
FIG. 3 is a schematic illustration of a road according to an embodiment;
FIG. 4 is a schematic road view of another embodiment;
fig. 5 is a schematic road diagram according to still another embodiment.
The specific embodiment is as follows:
it should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.
The application provides a navigation method for automatically driving an automobile. The execution subject of the navigation method of the automatic driving automobile includes, but is not limited to, a server, a terminal, and the like, which can be configured to execute at least one of the devices of the method provided by the embodiments of the present application.
In other words, the method may be performed by software or hardware installed in the terminal device or the server device, and the software may be a block chain platform. The server includes but is not limited to: a single server, a server cluster, a cloud server or a cloud server cluster, and the like. The server may be an independent server, or may be a cloud server that provides basic cloud computing services such as cloud service, a cloud database, cloud computing, cloud functions, cloud storage, web service, cloud communication, middleware service, domain name service, security service, content Delivery Network (CDN), and a big data and artificial intelligence platform.
Referring to fig. 1 and fig. 2, a flow chart of a navigation method for automatically driving a vehicle according to an embodiment of the present invention is shown. In this embodiment, the navigation method of the autonomous vehicle includes the steps of:
step S1: acquiring real-time picture information around the current vehicle through a monocular camera;
in the embodiment of the invention, the real-time picture information around the current vehicle is acquired so as to be used for extracting the enhanced key image from the real-time picture information subsequently.
In detail, the real-time picture information refers to information around the vehicle obtained by monocular cameras distributed around the vehicle sound, is processed to obtain 360-degree video information of the vehicle, and is marked as real-time picture information;
further, the 360 ° video information is shot by a panoramic camera to convert a static panoramic picture into a dynamic video image. The method is not in a single static panoramic picture form, has all-purpose factors such as depth of field, dynamic images and sound, and has the functions of positioning and synchronizing sound and picture.
Step S2: acquiring information of vehicles and pedestrians adjacent to a lane of a current vehicle through a binocular camera, and marking the information as binocular real-time picture information;
in the embodiment of the invention, binocular real-time picture information is obtained through a binocular camera for subsequently calculating the distance of a vehicle close to the rear of a lane;
in detail, the binocular real-time picture information refers to picture information simultaneously obtained by two cameras which are arranged on a vehicle rearview mirror and have a certain distance;
further, two monocular cameras that are movable and have a little distance are installed at the left and right rear-view mirrors of the vehicle, respectively, thereby constituting a binocular camera system. For example, when a complete video recording of an adjacent lane cannot be captured due to an angle, the binocular camera system may complete the capturing by adjusting.
And step S3: carrying out local decision-level picture enhancement processing on the obtained vehicle real-time picture information to obtain an enhanced key image;
in detail, a key information pre-extraction network is utilized to pre-extract key information from the vehicle real-time picture information to obtain preliminary key information;
further, the pre-extracting of the key information of the vehicle real-time picture information comprises: extracting key frames of the real-time picture information of the vehicle, and carrying out image annotation on images of the key frames to obtain preliminary key information; exemplarily, a traffic indicator light lamp post, a road sign, a pedestrian, a vehicle, a billboard and a building appear in the key frame image, and at this time, the system marks the traffic indicator light, the road sign, the pedestrian and the vehicle as preliminary key information;
the embodiment of the invention aims to pre-extract the real-time images of the vehicle to obtain key frames containing key information and pre-process the key information in the information so as to reduce the workload of subsequent image enhancement processing;
in detail, a picture enhancement model is utilized to carry out local decision-level picture enhancement processing on primary key information obtained by real-time picture information of the vehicle to obtain an enhanced key image;
the embodiment of the invention aims to reduce the processing time, improve the data processing efficiency, ensure the accuracy of the subsequent character information, improve the safety of vehicles and reduce the economic cost brought by violating the traffic rules by carrying out local decision-level picture enhancement processing on the primary key information;
and step S4: calculating to obtain the distance between the vehicle and the pedestrian in the binocular real-time picture information, and marking the safety distance of the vehicle according to the highest speed per hour of the vehicle road of the adjacent lane;
in detail, the distance to the target is calculated using the following formula:
wherein P is the position of the target, Z L Distance of left camera to target, Z R Distance of right camera to target, O L Is the center point of the aperture of the left camera, O R Is the center point of the aperture of the right camera, D is the distance between the optical centers of the two cameras, F is the focal length of the cameras, P L And P R Respectively projecting the object to the points on the camera photosensitive element through the central points of the diaphragms of the left camera and the right camera; p is L ' and P R ' is the distance from the outermost protective layer of the left and right cameras to the center point of the aperture of the left and right cameras, P L ' and P R The abscissa is X L And X R Y is the linear distance from the current camera to the tail of the vehicle;
the embodiment of the invention aims to calculate the distance of the vehicle behind the adjacent lane by using the binocular camera, so that the situation that the distance of the vehicle behind the adjacent lane cannot be calculated when a radar system fails, and the lane change operation cannot be carried out or the feedback of the lane change operation cannot be carried out can be effectively prevented;
in detail, the vehicles are safety distance marked at the highest speed per hour of the vehicle road in the adjacent lane, and in conjunction with the current weather conditions.
The embodiment of the invention aims to mark the safe distance of the vehicle, so that the vehicle and the vehicle in the adjacent lane keep the corresponding safe distance, and the corresponding evasive operation can be performed according to the operation of the vehicle in the adjacent lane and the vehicle system can be ensured to have sufficient time to react; for example, when the vehicle runs on the highway in rainy days and is in a middle lane, the safe distance threshold value of the left adjacent lane is set to be 300 meters, and the safe distance threshold value of the right adjacent lane is set to be 200 meters;
step S5: analyzing the vehicles in the adjacent lanes and in front of the current vehicle in the same lane by using a track prediction model to obtain abnormal driving vehicle information influencing safe driving, and correspondingly predicting the driving route of the abnormal driving vehicle information to obtain abnormal driving vehicle behavior prediction information;
in detail, the analysis of the adjacent lane and the vehicle in front of the current vehicle in the same lane means that the states and the speed of the vehicle lamps in the adjacent lane, the inclination angle of the head of the vehicle, the brake lamps in the same lane are predicted;
the embodiment of the invention aims to predict the running track of the surrounding vehicles, so that the path of the vehicle can be better planned in the follow-up process; illustratively, when the vehicle is about to drive to a crossroad and is in a middle straight lane and needs to turn left, through prediction analysis, when the vehicle in the left lane needs to go to the middle lane, the abnormal driving vehicle behavior prediction information can be obtained;
step S6: detecting and processing the enhanced key image to obtain key information in the image, extracting key place names and traffic indication information, extracting text information in the image, and marking the text information as emergency navigation text information; marking the surrounding shielding object rows as potential risk estimation information;
in detail, the states of the road sign and the traffic light in the enhanced key image are obtained by detecting and processing the enhanced key image, the semantics of the text information in the road sign are explained by a road traffic self-attention model in combination with the road traffic condition, so that the content of the text information is reordered or supplemented, and the text information is marked as emergency navigation text information;
the embodiment of the invention aims to hopefully carry out global traffic processing on the enhanced key image through the self-attention model, ensure the driving safety of vehicles and make better selection in the subsequent path planning; illustratively, the traffic sign displays an icon of a school road section, and has no corresponding character mark, when the traffic signal lamp displays green and can pass, and when people still pass in the pedestrian crossing at a slow speed, the content of the emergency navigation text information is no starting, and the current school road section is decelerated and slow-moved;
in detail, vehicles, road surface obstacles and pedestrians appearing in the enhanced key image are classified, and sheltered vehicles, current lane obstacles and pedestrians with a distance smaller than a safe distance threshold are marked and recorded as potential risk estimation information.
The embodiment of the invention aims to predict the potential risk so as to reduce the occurrence of traffic accidents, and when the nearby vehicles shield the pictures at two sides of the road, the accidents caused by the sudden acceleration of bicycles, pedestrians and the like at the two sides of the road are prevented; for example, if the current vehicle speed is 60 kilometers per hour, the distance between the pedestrian and the current lane is defined as 1m, and the distance between the pedestrian and the current lane is the safety distance of the pedestrian, and when the distance between the pedestrian and the current lane is smaller than the safety distance, the safety distance is marked as potential risk prediction information, so that the situation that the pedestrian suddenly flees out of the road and the emergency brake is not timely is prevented.
Step S7: extracting the potential risk estimation information, and judging the final result of the potential risk by using a potential risk evaluation detection model to obtain final potential risk information;
in detail, extracting the potential risk estimation information, and performing potential risk expectation value evaluation calculation by using a potential risk evaluation detection model:
wherein
For the expected value of the potential risk, H is the distance between the vehicle and the vehicle in front of the through lane, W is the environment adaptive value, G is the number of lanes on the road where the vehicle is located, N is the current traffic level value, V is the current speed of the vehicle, and V is the current traffic level value 0 The highest speed per hour, V, of the current lane compliance T For vehicles behind adjacent lanesI is the inclination angle of the vehicle head behind the adjacent lane;
the W environment adaptive value is the probability of the emergency situation according to the semantic judgment of the emergency navigation text information and the vehicle driving experience, and is calculated by a naive Bayes classifier;
according to the embodiment of the invention, the environment adaptive value W is utilized, and the corresponding potential risk expectation value can be derived well according to the environment changing in real time and the current time, so that the potential risk can be avoided more accurately; by using the inclination angle I of the vehicle head behind the adjacent lane and the maximum speed per hour V of the current lane compliance 0 And the speed V of the vehicle behind the adjacent lane T The method has obvious beneficial effects on whether the current vehicle needs to be subjected to lane changing operation; the sin function is used because
The method is characterized in that an increasing function is used in the front, and the side surface proves that the larger the inclination angle of the locomotive is, the larger the potential risk expectation value is;
the embodiment of the invention aims to calculate the probability of some possible events by utilizing various current conditions through a naive Bayes classifier, thereby preliminarily judging the current possible accidents; for example, when a cell passes, the current time is the time close to work and school, the current road is crowded, and the rear vehicles have obvious lane change behaviors, the situations of lane change vehicles, suddenly crossed pedestrians and electric vehicles can be easily calculated;
in detail, the risk events larger than the potential risk expectation value threshold are marked as preliminary potential risk information by using the formula;
calculating the loss value of each potential risk event by using a loss calculation formula of the potential risk assessment detection model:
whereinBeta is the loss value for each risk potential event, mu is the number of turns of the lane that need to be crossed to avoid the risk time, gamma is the reaction time to avoid the risk,
the method comprises the steps that a preset dynamic change value of a potential risk scene is obtained, omega is a predicted economic loss cost, and alpha is a preset collision weight value of different objects;
the embodiment of the invention aims to assign different loss values to different objects, so that pedestrians and nearby vehicles can be better protected in subsequent planning; through collision weight values alpha of different objects, casualties caused by the fact that the current vehicle loses to reduce economic cost can be effectively prevented, damage to pedestrians is avoided at any time, safety of the pedestrians around the current vehicle is improved, and worry of the pedestrians is eliminated;
in detail, the preliminary risk potential information is compared for loss values of the risk potential event to confirm the final risk potential information.
The embodiment of the invention aims to determine the final potential risk through the loss value, facilitate subsequent path planning and guarantee the safety of the current traffic participants. For example, when the vehicle in front suddenly brakes and stops, if the left side is a road guard fence and the right side has a pedestrian, the safety of the vehicle and the pedestrian in front is finally ensured, and the vehicle and the pedestrian in front are determined as final potential risk information; if the left side is the road guard fence, and the right side has no pedestrians and obstacles, the safety of the vehicle and the front vehicle is guaranteed, the road guard fence and the front vehicle are defined as final potential risk information, and the subsequent timely avoidance is carried out;
step S8: detecting the current satellite condition, and judging and repairing the reason of the satellite abnormality by using a satellite repairing model when the satellite is detected to be in an abnormal state to obtain accurate satellite positioning;
in detail, the abnormal reason of the satellite signals caused by physical failure or malicious network attack and the type of the current satellite abnormality are judged through historical training experience;
further, the network attacks suffered by unmanned vehicles may be internal attacks and external attacks. It is understood that external attacks mainly refer to attacks and interferences on communication equipment, signal detection equipment and other basic black equipment of vehicle data; the internal attack refers to attack and interference on a vehicle-mounted control system, a navigation sensor and other vehicle-mounted sensors;
for example, the signal detection device, the result of the determination may be divided into: a normal state without hierarchy; the system comprises medium-grade X-direction offset, Y-direction offset, X-direction delay drift and Y-direction delay drift with different degrees, and high-grade fixed point attack (namely, in the driving process, satellite signals are displayed on a certain point for a long time);
in detail, through a satellite diagnosis and repair algorithm, multi-channel data fusion, feature extraction and abnormal diagnosis result output are carried out on sensor data of a GPS and a LIDAR, so that vehicle information suffering from malicious network attacks is repaired.
Furthermore, original sensor multi-channel measurement data of a GPS and a LIDAR of the unmanned vehicle are directly input into an input layer to automatically complete data fusion and data processing, a plurality of ID convolution layers, relu activation layers and pooling layers in a convolution network layer are used for carrying out feature extraction on the original data, finally, the compression of dimension reduction and parameter quantity is completed, and dimension conversion between a one-dimensional convolution network layer and an output layer and the output of a diagnosis result are adaptively associated;
the embodiment of the invention aims to better select how to repair the abnormal condition of the satellite in the follow-up process by judging the type of the satellite abnormality so as to reduce the time for processing the satellite abnormality and better and more quickly plan a path in the abnormal state of the satellite;
step S9: when the repair is failed, performing data fusion on the emergency LBS positioning and the drawn route map which runs in the abnormal state by using an emergency positioning algorithm in the navigation system, thereby calculating the current repairability accurate position positioning;
in detail, the LiDAR odometer is utilized to extract the characteristic points obtained by each frame, a characteristic point coordinate system is established, and the subsequent characteristic points are placed in the coordinate system;
in detail, matching the current characteristic points extracted from the current coordinate system of the LiDAR odometer with the characteristic point set extracted from the previous frame, solving a relative pose transformation matrix and calculating pose information so as to obtain a route map for driving in an abnormal state;
furthermore, the time of the two positioning systems is calibrated, LBS positioning positions and position and pose information are calculated and respectively placed in a navigation map, and two corresponding preliminary motion trail maps are drawn;
further, preliminarily drawing the obtained route map which is driven under the abnormal state on a map, correcting the route which is obviously out of the road, and synchronizing the route to two corresponding preliminary motion track maps;
and further, comparing the positions of the unified time points of the two preliminary motion track maps, and performing parameter calculation on the map positions with the distance difference larger than a set threshold value to obtain an accurate positioning result.
The embodiment of the invention aims to carry out double point location when the GPS fails through LiDAR odometer and LBS positioning so as to obtain point location information more accurately and effectively;
step S10: planning an interactive predicted navigation track of the vehicle by utilizing accurate satellite positioning or repairability accurate position positioning and combining abnormal running vehicle behavior prediction information and potential risk prediction information;
in detail, a motion planning algorithm is utilized, a lane center line is used as a reference line, the road width is used as boundary constraint, a global Cartesian coordinate system is established, and a corresponding vehicle predicted navigation track is drawn in the coordinate system;
in detail, correcting the predicted navigation track of the vehicle by combining the abnormal running vehicle behavior prediction information and the potential risk prediction information, and planning the interactive predicted navigation track of the vehicle;
as shown in fig. 3 to 5, a route is planned for three types of roads commonly seen in actual road traffic;
FIG. 3 is a schematic diagram of a movement plan without consideration of multi-vehicle interaction, with an untimely response to lane-change actions;
FIG. 4 is a schematic view of a road with excessive acceleration and reduced ride comfort;
FIG. 5 is a schematic illustration of excessive vehicle speed and potential overspeed;
for example, in the deceleration driving scenario shown in fig. 3, the initial vehicle speed of C3 is 5.5m/s, the speed of the surrounding vehicle C4 is 8m/s, and the speed of the surrounding vehicle C2 is 2m/s, as can be seen from the interactive predicted trajectory of C1 and C2, when C2 is driven out of the fork to turn right, it is closer to C1 with a higher vehicle speed, so C1 is suddenly driven to the left into the current lane of the host vehicle C3 to avoid collision with C2 during driving, and the initial vehicle speeds of the following host vehicles C1, C2, and C3 are the same as those in fig. 3;
as shown in fig. 4, the movement plan without considering multi-vehicle interaction does not react to the lane change action of C1 in time, the main vehicle is still at a high driving speed within 0 to 10 seconds, and the C1 that has already driven into the lane where C3 is currently located is correspondingly decelerated only within 10 seconds, so that the speed changes suddenly, the driving efficiency is reduced, and the collision risk is increased. In fig. 4 and 5, the acceleration and the acceleration change rate fluctuate greatly, the turning width becomes excessively large, and the ride comfort is reduced. Compared with the motion planning without considering the track prediction, the speed following curve planned by interactively predicting the navigation track is more stable, the fluctuation of the acceleration and the acceleration change rate is smaller, the turning is smoother, and the driving comfort is higher.
The embodiment of the invention aims to improve the riding experience of passengers and reduce the occurrence of traffic accidents by interactively predicting the navigation track;
step S11: judging the motion control mechanism of the current vehicle and the state of the current vehicle by using the motion control detection mechanism, and enabling the chassis to execute the action of vehicle interactive prediction navigation track by using an adaptive uncertain intelligent vehicle steering stabilization system through a controller;
in detail, a motion control detection mechanism is utilized to detect the state of a tire according to the real-time rotation of a vehicle chassis gear;
in detail, parameters of the controller are optimized by using a genetic algorithm, a difference value between a reference speed and a current speed in a vehicle interactive prediction navigation track is input, and a required acceleration is obtained and is used as a unified acceleration signal of a brake and an accelerator;
the embodiment of the invention aims to optimize the parameters of the controller by considering the influence of unknown nonlinearity on the control performance and utilizing a genetic algorithm to perform online approximation on the unknown nonlinearity of the system, thereby improving the practicability of the controller, reducing the occurrence of traffic accidents caused by the controller and further improving the experience of passengers;
in detail, an intelligent vehicle steering stabilizing system with adaptability uncertainty is utilized, a unified acceleration signal and a vehicle interactive prediction navigation track are combined, and the controller is used for coordinately controlling all systems of a chassis, so that the tires are stably controlled according to the states of different tires;
the embodiment of the invention aims to realize that the steer-by-wire system can still obtain satisfactory control performance under the conditions of control gain, actuator dead zone and faults by considering the influence of unknown control gain, actuator dead zone and faults on the control performance, and realize real-time control on tires, thereby better executing the action of interactively predicting the navigation track of the vehicle;
the foregoing are merely exemplary embodiments of the present invention, which enable those skilled in the art to understand or practice the present invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the generic principles defined herein may be applied to other embodiments without departing from the spirit or scope of the invention. Thus, the present invention is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.
Claims (10)
1. A navigation method for an autonomous vehicle, the method comprising the steps of:
step S1: acquiring real-time picture information around the current vehicle through a monocular camera;
step S2: acquiring binocular real-time picture information through a binocular camera, wherein the binocular real-time picture information comprises information of vehicles and pedestrians adjacent to a current vehicle;
and step S3: carrying out local decision-level picture enhancement processing on the obtained real-time picture information around the current vehicle to obtain an enhanced key image;
and step S4: calculating to obtain the distance between the vehicles and the pedestrians in the adjacent lanes in the binocular real-time picture information, and marking the safety distance of the vehicles in the adjacent lanes according to the highest speed per hour of the roads of the vehicles in the adjacent lanes;
step S5: analyzing vehicles in the adjacent lanes and vehicles in front of the current vehicle in the same lane by using a track prediction model to obtain abnormal running vehicle information influencing safe driving, and correspondingly predicting the running route of the abnormal running vehicle information to obtain abnormal running vehicle behavior prediction information;
step S6: detecting and processing the enhanced key image to obtain key information in the enhanced key image, extracting key place names and traffic indication information in the key information, extracting text information in the key place names and the traffic indication information, and marking the text information as emergency navigation text information; marking the shelters around the current vehicle, and recording as the potential risk estimation information;
step S7: carrying out potential risk judgment on the potential risk estimated information by using a potential risk evaluation detection model to obtain final potential risk information;
step S8: detecting the current satellite condition, and judging and repairing the reason of the satellite abnormality by using a satellite repairing model when the satellite is detected to be in an abnormal state to obtain accurate satellite positioning;
step S9: when the repair is failed, performing data fusion on the emergency LBS positioning and the drawn route map which runs in the abnormal state by using an emergency positioning algorithm in the navigation system, thereby calculating the current repairability accurate position positioning;
step S10: planning an interactive predicted navigation track of the vehicle by utilizing accurate satellite positioning or repairability accurate position positioning and combining abnormal running vehicle behavior prediction information, potential risk prediction information and emergency navigation text information;
step S11: and judging the motion control mechanism of the current vehicle and the state of the current vehicle by using the motion control detection mechanism, and performing the action of interactively predicting the navigation track of the vehicle by using an adaptive uncertain intelligent vehicle steering stabilizing system.
2. The method according to claim 1, wherein step S3 is specifically:
pre-extracting key information from real-time picture information around the current vehicle by using a key information pre-extraction network to obtain preliminary key information;
and carrying out local decision-level picture enhancement processing on the preliminary key information by using a picture enhancement model to obtain an enhanced key image.
3. The method of claim 1, wherein the step of performing the safe distance marking in step S4 is:
the distance from the current vehicle to the adjacent lane target is calculated by the following formula:
wherein P is the position of the target, Z L Distance of left camera to target, Z R Distance of right camera to target, O L Is the center point of the aperture of the left camera, O R Is the center point of the aperture of the right camera, D is the distance between the optical centers of the two cameras, F is the focal length of the cameras, P L And P R Respectively projecting the object to the points on the camera photosensitive element through the central points of the diaphragms of the left camera and the right camera; p L ' and P R ' is the distance from the outermost protective layers of the left and right cameras to the center point of the aperture of the left and right cameras,P L ' and P R ' abscissa is X L And X R Y is the linear distance from the current camera to the tail of the vehicle;
the vehicles are marked with safe distance at the highest speed per hour of the vehicle road in the adjacent lane, and in combination with the current weather conditions.
4. The method according to claim 1, wherein step S6 is specifically:
detecting and processing the enhanced key images to obtain states of the road sign and the traffic lights, interpreting semantics of text information in the road sign by a road traffic self-attention model in combination with road traffic conditions, thereby reordering or supplementing the text information contents and marking the text information contents as emergency navigation text information;
and classifying the vehicles, the road surface obstacles and the pedestrians appearing in the enhanced key image, marking the sheltered vehicles, the current lane obstacles and the pedestrians with the distance less than the safe distance threshold value, and recording as the potential risk estimation information.
5. The method according to claim 1, wherein step S7 is specifically:
extracting the latent risk estimation information, and carrying out the evaluation calculation of the latent risk expected value by using a latent risk evaluation detection model:
wherein
For the expected value of the potential risk, H is the distance between the vehicle and the vehicle in front of the traffic lane, W is an environment adaptive value, G is the number of lanes on the road where the vehicle is located, N is the current traffic level value, V is the speed of the current vehicle, V0 is the highest speed per hour of the current lane compliance, VT is the speed of the vehicle behind the adjacent lane, and I is the speed of the vehicle behind the adjacent laneThe inclination angle of the square vehicle head;
the W environment adaptive value is the probability of the emergency situation according to the semantic judgment of the emergency navigation text information and the vehicle driving experience, and is calculated by a naive Bayes classifier;
using the formula to mark risk events larger than the expected value threshold value of the potential risk as preliminary potential risk information;
calculating the loss value of each potential risk event by using a loss calculation formula of the potential risk assessment detection model:
where β is the loss value for each risk potential event, μ is the number of turns of the lane that need to be crossed to avoid the risk time, γ is the reaction time to avoid the risk,
the method comprises the steps that a preset dynamic change value of a potential risk scene is obtained, omega is a predicted economic loss cost, and alpha is a preset collision weight value of different objects;
and comparing the loss values of the potential risk events to the preliminary potential risk information, so as to confirm that the preliminary potential risk information with the largest loss value is the final potential risk information.
6. The method according to claim 1, wherein the determining and repairing the cause of the satellite abnormality in step S8 comprises the following specific steps:
judging the satellite signal to be an abnormal reason caused by physical failure or malicious network attack through historical training experience, and judging the type of the current satellite abnormality;
through a satellite diagnosis and repair algorithm, multi-channel data fusion, feature extraction and abnormal diagnosis result output are carried out on the sensor data of the GPS and the LIDAR, so that the vehicle information suffering from malicious network attack is repaired.
7. The method according to claim 1, wherein the step S9 of drawing a route pattern traveled in an abnormal state comprises the specific steps of:
extracting the characteristic points obtained by each frame by using a LiDAR odometer, establishing a characteristic point coordinate system, and placing the subsequent characteristic points in the characteristic point coordinate system;
and matching the current characteristic points extracted from the coordinate system of the current characteristic points of the LiDAR odometer with the characteristic point set extracted from the previous frame, solving a relative pose transformation matrix and calculating pose information so as to obtain a route map driven in an abnormal state.
8. The method according to claim 5, wherein the position fusion in step S8 comprises the following specific steps:
calibrating the time of the two positioning systems, respectively placing the LBS positioning positions and the calculated pose information in a navigation map, and drawing two corresponding preliminary motion trail maps;
and comparing the positions of the unified time points of the two preliminary motion trail maps, and carrying out parameter calculation on the map positions with the distance difference larger than a set threshold value so as to obtain an accurate positioning result.
9. The method of claim 1, wherein the step S10 of planning the vehicle interactive predicted navigation track comprises the following specific steps:
by utilizing a motion planning algorithm, taking a lane central line as a reference line, taking the road width as boundary constraint, establishing a global Cartesian coordinate system, and drawing a corresponding vehicle predicted navigation track in the coordinate system;
and correcting the predicted navigation track of the vehicle by combining the abnormal running vehicle behavior prediction information, the potential risk prediction information and the emergency navigation text information, and planning the interactive predicted navigation track of the vehicle.
10. The method according to claim 1, wherein the step S11 comprises the following specific steps:
detecting the state of the tire by utilizing a motion control detection mechanism according to the real-time rotation of a vehicle chassis gear;
optimizing parameters of the controller by using a genetic algorithm, acquiring a difference value between a reference speed and a current speed in a vehicle interactive prediction navigation track, acquiring a required acceleration, and using the acceleration as a unified acceleration signal of a brake and an accelerator;
by using an intelligent vehicle steering stabilizing system with uncertain adaptability, combining a unified acceleration signal and a vehicle interactive prediction navigation track, and performing coordinated control on each system of the chassis through a controller, the tires are stably controlled according to the states of different tires.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211329153.1A CN115775463A (en) | 2022-10-27 | 2022-10-27 | Navigation method for automatically driving automobile |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211329153.1A CN115775463A (en) | 2022-10-27 | 2022-10-27 | Navigation method for automatically driving automobile |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115775463A true CN115775463A (en) | 2023-03-10 |
Family
ID=85388584
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211329153.1A Withdrawn CN115775463A (en) | 2022-10-27 | 2022-10-27 | Navigation method for automatically driving automobile |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115775463A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117405124A (en) * | 2023-12-13 | 2024-01-16 | 融科联创(天津)信息技术有限公司 | Path planning method and system based on big data |
-
2022
- 2022-10-27 CN CN202211329153.1A patent/CN115775463A/en not_active Withdrawn
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117405124A (en) * | 2023-12-13 | 2024-01-16 | 融科联创(天津)信息技术有限公司 | Path planning method and system based on big data |
| CN117405124B (en) * | 2023-12-13 | 2024-02-27 | 融科联创(天津)信息技术有限公司 | Path planning method and system based on big data |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US11990036B2 (en) | Driver behavior monitoring | |
| CN111540237B (en) | Method for automatically generating vehicle safety driving guarantee scheme based on multi-data fusion | |
| US10345822B1 (en) | Cognitive mapping for vehicles | |
| US10726280B2 (en) | Traffic signal analysis system | |
| CN113165652B (en) | Verifying predicted trajectories using a mesh-based approach | |
| US20240290201A1 (en) | Driver behavior monitoring | |
| WO2022206942A1 (en) | Laser radar point cloud dynamic segmentation and fusion method based on driving safety risk field | |
| CN112700470B (en) | Target detection and track extraction method based on traffic video stream | |
| LU101647B1 (en) | Road pedestrian classification method and top-view pedestrian risk quantitative method in two-dimensional world coordinate system | |
| IL256524A (en) | Improved object detection for an autonomous vehicle | |
| DE112017007906T5 (en) | VEHICLE CONTROL DEVICE, VEHICLE CONTROL METHOD AND PROGRAM | |
| US20200098270A1 (en) | Method And Device For Providing An Environmental Image Of An Environment Of A Mobile Apparatus And Motor Vehicle With Such A Device | |
| CN115298073B (en) | Vehicle travel support method and travel support device | |
| WO2019106789A1 (en) | Processing device and processing method | |
| US11521494B2 (en) | Vehicle eccentricity mapping | |
| DE102021124913A1 (en) | METRIC BACKPROPAGATION FOR EVALUATION OF SUBSYSTEMS PERFORMANCE | |
| US11727694B2 (en) | System and method for automatic assessment of comparative negligence for one or more vehicles involved in an accident | |
| US11640173B2 (en) | Control apparatus, control method, and computer-readable storage medium storing program | |
| CN114932902B (en) | Ghost probe early warning avoiding method and system based on Internet of vehicles technology | |
| CN116703966A (en) | Multi-object tracking | |
| CN115775463A (en) | Navigation method for automatically driving automobile | |
| US20230140079A1 (en) | Method and Systems for Detection Accuracy Ranking and Vehicle Instructions | |
| CN115649198A (en) | Automatic vehicle driving method and system | |
| CN114373297A (en) | Data processing device and method and electronic equipment | |
| CN114756019A (en) | External indication handling for autonomous systems |
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 | ||
| WW01 | Invention patent application withdrawn after publication | ||
| WW01 | Invention patent application withdrawn after publication |
Application publication date: 20230310 |