CN115454051A - Non-structural scene continuous automatic driving perception and path planning method under limited computing power - Google Patents

Non-structural scene continuous automatic driving perception and path planning method under limited computing power Download PDF

Info

Publication number
CN115454051A
CN115454051A CN202210989378.3A CN202210989378A CN115454051A CN 115454051 A CN115454051 A CN 115454051A CN 202210989378 A CN202210989378 A CN 202210989378A CN 115454051 A CN115454051 A CN 115454051A
Authority
CN
China
Prior art keywords
task
suitability
planning
sensing
map
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.)
Pending
Application number
CN202210989378.3A
Other languages
Chinese (zh)
Inventor
邢琰
刘祥
滕宝毅
毛晓艳
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Beijing Institute of Control Engineering
Original Assignee
Beijing Institute of Control Engineering
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Beijing Institute of Control Engineering filed Critical Beijing Institute of Control Engineering
Priority to CN202210989378.3A priority Critical patent/CN115454051A/en
Publication of CN115454051A publication Critical patent/CN115454051A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G05—CONTROLLING; REGULATING
    • G05D—SYSTEMS FOR CONTROLLING OR REGULATING NON-ELECTRIC VARIABLES
    • G05D1/00—Control of position, course or altitude of land, water, air, or space vehicles, e.g. automatic pilot
    • G05D1/02—Control of position or course in two dimensions
    • G05D1/021—Control of position or course in two dimensions specially adapted to land vehicles
    • G05D1/0212—Control of position or course in two dimensions specially adapted to land vehicles with means for defining a desired trajectory
    • G05D1/0214—Control of position or course in two dimensions specially adapted to land vehicles with means for defining a desired trajectory in accordance with safety or protection criteria, e.g. avoiding hazardous areas
    • G—PHYSICS
    • G05—CONTROLLING; REGULATING
    • G05D—SYSTEMS FOR CONTROLLING OR REGULATING NON-ELECTRIC VARIABLES
    • G05D1/00—Control of position, course or altitude of land, water, air, or space vehicles, e.g. automatic pilot
    • G05D1/02—Control of position or course in two dimensions
    • G05D1/021—Control of position or course in two dimensions specially adapted to land vehicles
    • G05D1/0212—Control of position or course in two dimensions specially adapted to land vehicles with means for defining a desired trajectory
    • G05D1/0221—Control of position or course in two dimensions specially adapted to land vehicles with means for defining a desired trajectory involving a learning process
    • G—PHYSICS
    • G05—CONTROLLING; REGULATING
    • G05D—SYSTEMS FOR CONTROLLING OR REGULATING NON-ELECTRIC VARIABLES
    • G05D1/00—Control of position, course or altitude of land, water, air, or space vehicles, e.g. automatic pilot
    • G05D1/02—Control of position or course in two dimensions
    • G05D1/021—Control of position or course in two dimensions specially adapted to land vehicles
    • G05D1/0276—Control of position or course in two dimensions specially adapted to land vehicles using signals provided by a source external to the vehicle

Abstract

The invention relates to a method for continuously and automatically driving, sensing and path planning in a non-structural scene under limited computing power, which comprises a navigation control task and a sensing and obstacle avoidance planning task which are sequentially executed; the navigation control task is a fixed-period task, and each period of the navigation control task executes navigation and control operation and judges whether to initiate a perception and obstacle avoidance task; the sensing and obstacle avoidance planning task is an indefinite-period task and is initiated by a navigation control task; the task executes sensing and planning operation, and after the sensing and planning operation is completed, a feedback state is fed back to the navigation control task. The invention effectively overcomes the constraints of smaller sensing range of the sensor and limited computational power of the vehicle-mounted computer, and improves the sensing and moving efficiency of the autonomous moving vehicle.

Description

Method for continuously and automatically driving, sensing and planning path in non-structural scene under finite computing power
Technical Field
The invention relates to a method for continuously and automatically driving, sensing and path planning in an unstructured scene under limited computational power, which is applied to a moving vehicle working on unstructured terrain, having a small sensing range of a sensor and limited computer computational power, and realizes autonomous local obstacle avoidance in a continuous advancing process.
Background
For an autonomous moving vehicle, in order to ensure the safety of the moving process, obstacle avoidance planning needs to be completed based on local perception.
The local perception is to perform fine perception and modeling on a close-distance local environment in the moving process on the basis of global planning, supplement/update perception information through fusion, and finish obstacle extraction and trafficability analysis based on a modeling result.
The local obstacle avoidance planning depends on local sensing information and finds a safe and collision-free path according to a certain strategy. The local obstacle avoidance planning takes safety and no collision as a main target, and the planning result does not need to be an optimal path.
The local obstacle avoidance planning of the moving vehicle has obvious dependence on the sensing range of the sensing sensor and on-vehicle computing resources. When the sensing range of the sensing sensor is difficult to cover the surrounding area of the vehicle, safety risk is brought to local obstacle avoidance; when the on-vehicle computing resources are limited and the computing power cannot meet the real-time requirement of the data processing of the sensing sensor, the moving speed of the vehicle generally needs to be correspondingly reduced. Especially for a detection vehicle for inspecting and moving the surface of an extraterrestrial celestial body, because the working environment is an unknown and complex unstructured terrain, the terrain modeling and traffic analysis calculation amount is large, and the vehicle often needs to be stopped to complete environment perception and obstacle avoidance planning, thereby seriously affecting the vehicle moving efficiency.
Disclosure of Invention
The technical problem solved by the invention is as follows: aiming at the defects of the prior art, the invention provides a method for continuously and automatically driving, sensing and path planning in a non-structural scene under limited computing power, effectively overcomes the constraints of smaller sensing range of a sensor and limited computing power of a vehicle-mounted computer, and improves the sensing and moving efficiency of a class of autonomous moving vehicles.
The technical scheme adopted by the invention is as follows: a method for continuously and automatically driving, sensing and path planning in a non-structural scene under limited computing power comprises a navigation control task and a sensing and obstacle avoidance planning task which are sequentially executed;
the navigation control task is a fixed-period task, and each period of the navigation control task executes navigation and control operation and judges whether to initiate a perception and obstacle avoidance task;
the sensing and obstacle avoidance planning task is an indefinite-period task and is initiated by a navigation control task; the task executes sensing and planning operation, and after the sensing and planning operation is completed, a feedback state is fed back to the navigation control task.
The navigation control task comprises:
from an initial time t 0 At the beginning, every T 1 Time period, i.e. t k ＝t 0 +i*T 1 I =0 to k; k represents the total number of cycles within the process; executing pose determination and motion control once; the position and posture determination is used for determining the position and posture of the moving vehicle under a certain specified navigation reference coordinate system, and the motion control is used for controlling the moving vehicle to move according to the output result of the obstacle avoidance plan;
after the pose determination and the motion control are executed in each period, the starting condition judgment of the sensing and obstacle avoidance task is executed, and whether the sensing and obstacle avoidance planning task is started or not is determined according to the result.
And executing judgment of starting conditions of the perception and obstacle avoidance task, specifically comprising the following steps:
if: l is u <V*T 2 ，L u ＝L-L e Then: starting a sensing and obstacle avoidance planning task; wherein L is the path length given by the one-time perception planning task, L u For the remaining undriven length of the planned path, L e Is the length of the path, t, that has been traveled since the last perceptual planning task 0 Time L = L e ＝L u =0; v is the moving speed, T 2 Task execution time is planned for each perception.
The perception and obstacle avoidance planning task comprises the following steps:
after receiving a sensing planning task starting instruction sent by a navigation control task, controlling a configured sensing sensor to acquire data;
after data acquisition is completed, environment modeling is carried out by using the current moment position and attitude information obtained by the navigation control task to obtain a rasterized digital elevation map;
after the environment modeling is completed, traffic analysis is completed on the digital elevation grid topographic map, and the traffic suitability of each grid is calculated to obtain a current traffic suitability map, namely a local suitability map; when the passing suitability degree is evaluated, if effective measurement data exist in the grid, the numerical range is 0-1 after the suitability degree evaluation, the grid is impassable, and the grid is safe to pass, wherein 1 is the valid measurement data; if no valid measurement data exists in a certain grid, the value is assigned to be 2, namely invalid;
converting the historical traffic analysis result obtained by the last perception planning task to the current position, and fusing the suitability map to obtain a fused traffic suitability map, namely a global suitability map;
and planning a path on the fused global traffic suitability map, and giving the curvature, the length and the like of the planned path.
The definition of the suitability map is as follows:
and respectively establishing a local suitability map and a global suitability map, wherein the local suitability map is established by each perception planning task based on current perception data, and the global suitability map is obtained by fusing the local suitability maps of the previous times.
The range size of the global suitability map and the local suitability map is determined in advance according to the effective sensing range of the sensing sensor, and the position of the automatic driving unmanned system is taken as an origin.
The local suitability map and the global suitability map both adopt a grid map mode, wherein the grid division line number and the grid column number of the local suitability map are r M ×r M The grid of the global suitability map is divided into the number of rows and the number of columns as r GM ×r GM And r is GM ≥n·r M Coefficient n>1; the traffic suitability evaluation values of each grid of the local suitability map and the global suitability map are respectively shown in an array G cl |r M ×r M And G gl |r GM ×r GM Performing the following steps; each data representation of the array has a size r GM ×r GM The traffic suitability evaluation value of each grid in the grid map of (1).
The local suitability map data G cl And global suitability map data G gl The initial values are all unviable, and G is updated after the perception planning task is executed cl And G gl To data in an area overlapping the local suitability map range.
The fusion method for the suitability map fusion specifically comprises the following steps:
obtaining current local suitability map data G for each perception planning task cl Thereafter, a new global suitability map data G is defined gl_new From the current suitability map origin (x) 0 ,y 0 ) And the origin (x) of the suitability map during the last planning step L0 ,y L0 ) Calculating the relative position relationship of the automatic unmanned system during planning twice before and after, and then utilizing G gl For G gl_new Assigning the grids of the corresponding areas, and finally ordering G gl ＝G gl_new Thereby completing the translation of the global suitability map;
for local suitability map data G cl Each grid (i) of l ,j l ) Find its correspondence in the global suitability map data G gl Grid (i) of (1) g ,j g ) Using G gl (i g ,j g ) For G cl (i l ,j l ) Fusing to obtain fused local suitability map data G cl ；
Using fused local suitability map data G cl Mesh of (i) l ,j l ) Find its correspondence in global suitability map data G gl Cell (i) in g ,j g ) Let G gl (i g ,j g )＝G cl (i l ,j l ) And finishing the updating of the global suitability map.
Obtaining fused local suitability map data G cl The method comprises the following steps:
if G is gl (i g ,j g ) Effective, i.e., the array is in the range of 0 to 1, and G cl (i l ,j l ) Invalid, i.e. array 2, then: g cl (i l ,j l )＝m g0 G gl (i g ,j g )；
If G is gl (i g ,j g ) Effective and G cl (i l ,j l ) And if effective, then:
G cl (i l ,j l )＝m g1 G gl (i g ,j g )+(1-m g1 )G cl (i l ,j l )；
wherein m is g0 The value range is a forgetting factor of the fitness degree [0,1]；m g1 Is a fitness weighting coefficient with a value range of [0,1%]。
Compared with the prior art, the invention has the advantages that:
(1) The invention optimizes the local obstacle avoidance process of the autonomous moving vehicle, divides the environment perception and the trafficability analysis and the obstacle avoidance path selection into two independent links, respectively records the corresponding vehicle position and attitude data, and separates the suitability map origin from the planning starting position, thereby realizing the continuous moving autonomous obstacle avoidance under the condition of limited calculation power and effectively improving the moving efficiency of the vehicle.
(2) The invention improves the suitability map fusion algorithm, enables the map coordinate to be independent of the vehicle course, avoids attitude conversion in the fusion process, and can conveniently realize fusion of suitability maps generated at different sensing positions, thereby expanding the effective coverage range of the suitability map, reducing the safety risk of the vehicle caused by short sensing distance of the sensor, and improving the reliability of local obstacle avoidance planning.
(3) The invention is also suitable for the moving vehicles working on the unstructured terrain, with smaller sensor sensing range and limited computer computing power, realizes autonomous local obstacle avoidance in the continuous advancing process, and has wider application range and strong practicability.
Drawings
FIG. 1 is a flow chart of the method of the present invention.
Detailed Description
The invention relates to a method for continuously and automatically driving, sensing and path planning in an unstructured scene under limited computational power, which is suitable for moving vehicles working in unstructured terrain and constrained in sensing sensors and computer computational power. Wherein the perception sensor is generally a large field of view short focal length stereo camera or a depth camera. The effective sensing distance of the sensor is S meters, the moving speed of the vehicle is V meters per second, and the time T second for finishing the environmental sensing and traffic analysis by the computer meets the relation S >6VT, wherein the effective sensing distance means that the sensing data precision of the sensor within the distance can meet the precision requirements of environmental modeling and traffic analysis. The specific implementation method comprises the following steps:
the method comprises the steps of (I) decomposing a local obstacle avoidance scheme into two parallel tasks which are respectively a navigation control task and a perception and obstacle avoidance planning task.
And secondly, the navigation control task works as follows:
(1) And the navigation control task periodically executes the functions of pose determination and motion control, wherein the pose determination is used for determining the position and the posture of the moving vehicle under a certain specified navigation reference coordinate system, and the motion control is used for controlling the moving vehicle to move according to the output result of the obstacle avoidance plan.
(2) And the navigation control task informs the sensing and obstacle avoidance planning task to execute data acquisition before starting moving, or T seconds before the end of the currently executed obstacle avoidance planning movement time, or when the currently executed obstacle avoidance planning movement mileage is left by V x T meters, and simultaneously transmits the current vehicle position and attitude information to the sensing and obstacle avoidance planning task.
(3) And the navigation control task needs to inform the perception and obstacle avoidance planning task after receiving a 'trafficability analysis' completion signal of the perception and obstacle avoidance planning task, execute 'planning path selection', and simultaneously transmit the current vehicle position and attitude information to the perception and obstacle avoidance planning task.
(4) And after receiving a planning path selection result of the sensing and obstacle avoidance planning task, the navigation control task updates the motion control strategy and controls the vehicle to move according to a new planning output result.
And (III) the perception and obstacle avoidance planning task works as follows:
(1) And after receiving a data acquisition command of the navigation control task, controlling the configured sensing sensor to acquire data.
(2) And after the data acquisition is finished, processing the data, converting the processed three-dimensional point cloud data into a coordinate system required by planning according to the acquisition time position and posture information sent by the navigation control task, carrying out environment modeling, and establishing a rasterized digital elevation topographic map.
(3) And after the 'environment modeling' is finished, traffic analysis is finished on the digital elevation grid topographic map, the suitability evaluation value of each grid is obtained through calculation, meanwhile, the historical traffic analysis result is converted into the current planning coordinate system, the suitability map fusion is carried out, and after the suitability map fusion is finished, a 'traffic analysis' completion signal is sent to a navigation control task.
(4) And after receiving a planning path selection instruction of the navigation control task, generating a current planning path result on a current suitability map by taking the current planning time position sent by the navigation control task as a starting point and selecting a proper local obstacle avoidance planning algorithm according to the attitude information, and feeding back the current planning path result to the navigation control task, wherein the planned longest moving distance is generally not more than S/3.
And (IV) the concrete strategy of the suitability map fusion is as follows:
(1) And respectively establishing a local suitability map and a global suitability map, wherein the local suitability map is used for traffic analysis and local obstacle avoidance planning based on current sensing data, and the global suitability map is used for realizing fusion of historical suitability maps.
(2) And generating a local suitability map, wherein the specific contents are as follows:
local suitability map data G cl Is one size of r M ×r M The physical meaning of the array is an array centered on the origin of the fitness map coordinate system and containing r M ×r M The local suitability map of each grid, each grid stores an evaluation value (namely suitability) for representing whether the corresponding terrain is suitable for the vehicle to pass.
The direction of the local suitability map is consistent with the navigation reference coordinate system and does not change along with the change of the vehicle course. Setting the resolution of each grid of the local suitability map array as l c Then r is M ×l c Generally not less than twice the effective sensing distance S of the sensing sensors. The original point position of the local suitability map is the position of the vehicle at the environmental perception moment, so that the vehicle can sense the environment no matter how muchAnd in which direction the vehicle course faces, the local suitability map can cover the effective perception range of the perception sensor.
For each grid in the local suitability map, a terrain block which takes the terrain block as the center and the size of the terrain block as the vehicle size envelope is taken, the terrain feature data in the terrain block is calculated by utilizing the terrain feature data output by environment modeling, suitability evaluation values of the terrain block such as gradient, roughness and height drop are calculated respectively, and the suitability of the grid in the local suitability map is obtained comprehensively.
(3) And the suitability map is fused, and the specific contents are as follows:
this section utilizes global suitability map data G gl For the local suitability map data G generated at the time of current planning cl Performing fusion while using the fused G cl For G gl And (4) updating.
Global suitability map data G gl Is one size of r GM ×r GM The physical meaning and local suitability map data G of the array cl Same except for the map size ratio G cl Large, so that it can be used to store multiple planning generated G cl General r GM Should not be less than 2r M 。
First, a new global suitability map data G is defined gl_new From the current suitability map origin (x) 0 ,y 0 ) And the origin (x) of the suitability map in the last planning step L0 ,y L0 ) Calculating the relative position relationship of the vehicle during the planning of the two times before and after, and then utilizing G gl For G gl_new Assigning the grids of the corresponding areas, and finally ordering G gl ＝G gl_new And thus, the translation of the global suitability map is completed.
Second, for the local suitability map data G cl Mesh in (i) l ,j l ) Find its correspondence in global suitability map data G gl Cell (i) in (c) g ,j g ) Using G gl For G cl Carrying out fusion, specifically:
if G is gl (i g ,j g ) Effective and G cl (i l ,j l ) And if the data is invalid, then: g cl (i l ,j l )＝m g0 G gl (i g ,j g )；
If G is gl (i g ,j g ) Effective and G cl (i l ,j l ) And if effective, then:
G cl (i l ,j l )＝m g1 G gl (i g ,j g )+(1-m g1 )G cl (i l ,j l )；
wherein m is g0 The value range is a forgetting factor of the fitness degree [0,1]And may be generally 0.95; m is a unit of g1 Is a fitness weighting coefficient with a value range of [0,1%]And may be generally 0.9.
Finally, the fused local suitability map data G is utilized cl Mesh in (i) l ,j l ) Find its correspondence in the global suitability map data G gl Cell (i) in g ,j g ) Let G gl (i g ,j g )＝G cl (i l ,j l ) And finishing the updating of the global suitability map.
The invention is described in detail below with reference to fig. 1, a planetary surface inspection probe is equipped with a pair of wide-angle fisheye obstacle avoidance cameras for local sensing, the effective sensing distance is 3m, the maximum moving speed of the vehicle is designed to be 200m/h, namely about 0.056m/s, and the total time required for completing environmental sensing and traffic analysis is not more than 5s.
The continuous moving local obstacle avoidance method of the inspection detector comprises the following specific implementation method:
the method comprises the following steps of (I) decomposing a local obstacle avoidance scheme into two parallel tasks, namely a navigation control task and a perception and obstacle avoidance planning task.
And (II) the navigation control task works as follows:
(1) And the navigation control task periodically executes the functions of pose determination and motion control, wherein the pose determination is used for determining the position and the posture of the inspection detector in an appointed north-east-ground coordinate system, and the motion control is used for controlling the inspection detector to move according to the output result of the obstacle avoidance plan.
(2) Before the navigation control task starts to move or when the currently executed obstacle avoidance planning movement mileage remains 0.3 m, the navigation control task informs the sensing and obstacle avoidance planning task, executes data acquisition and simultaneously transmits the position and posture information of the current patrol detector to the sensing and obstacle avoidance planning task.
(3) And the navigation control task needs to inform the sensing and obstacle avoidance planning task after receiving a 'trafficability analysis' completion signal of the sensing and obstacle avoidance planning task, execute 'planning path selection', and simultaneously transmit the position and posture information of the inspection detector to the sensing and obstacle avoidance planning task.
(4) And after receiving a planning path selection result of the sensing and obstacle avoidance planning task, the navigation control task updates a motion control strategy and controls the inspection probe to move according to a new planning output result.
And (II) the perception and obstacle avoidance planning task works as follows:
(1) And after receiving a data acquisition command of the navigation control task, controlling the obstacle avoidance camera to acquire images.
(2) And after the data acquisition is finished, processing the image, converting the processed three-dimensional point cloud data into a north-east-ground coordinate system required by planning according to the acquisition time position and attitude information sent by the navigation control task, carrying out environment modeling, and establishing a rasterized digital elevation topographic map.
(3) And after the 'environment modeling' is finished, traffic analysis is finished on the digital elevation grid topographic map, the suitability evaluation value of each grid is obtained through calculation, meanwhile, the historical traffic analysis result is converted into the current planning coordinate system, the suitability map fusion is carried out, and after the suitability map fusion is finished, a 'traffic analysis' completion signal is sent to a navigation control task.
(4) And after receiving a planning path selection instruction of the navigation control task, selecting the curvature and the arc length of the optimal arc path from a group of alternative arc paths as a current planning path result by taking the current planning time position sent by the navigation control task as a starting point on a current suitability map and combining attitude information and selecting a GESTALT algorithm, and feeding back the curvature and the arc length to the navigation control task, wherein the planning arc length is not more than 1 meter.
And (III) the concrete strategy of the suitability map fusion is as follows:
(1) And respectively establishing a local suitability map and a global suitability map, wherein the local suitability map is used for trafficability analysis and local obstacle avoidance planning based on current sensing data, and the global suitability map is used for realizing fusion of historical suitability maps.
(2) And generating a local suitability map, wherein the specific contents are as follows:
local suitability map data G cl Is an array of 50 x 50 in size, with a resolution of 0.2 meters per grid.
The direction of the local suitability map is consistent with the navigation reference coordinate system and does not change along with the change of the vehicle course.
For each grid in the local suitability map, a terrain block which takes the terrain block as the center and the size of the terrain block as the vehicle size envelope is taken, the terrain feature data in the terrain block is calculated by utilizing the terrain feature data output by environment modeling, suitability evaluation values of the terrain block such as gradient, roughness and height drop are calculated respectively, and the minimum value of the three is taken as the suitability of the grid.
(3) And the suitability map is fused, and the specific contents are as follows:
this section utilizes global suitability map data G gl For the local suitability map data G generated during current planning cl Performing fusion while using the fused G cl For G gl And (6) updating.
Global suitability map data G gl Is an array of 100 x 100 in size.
First, a new global suitability map data G is defined gl_new From the current suitability map origin (x) 0 ,y 0 ) And the origin (x) of the suitability map during the last planning step L0 ,y L0 ) Calculating the relative position relationship of the vehicle during the planning of the two times before and after, and then utilizing G gl For G gl_new Network of corresponding areasAssigning the grids and finally ordering G gl ＝G gl_new And thus, the translation of the global suitability map is completed.
Second, for the local suitability map data G cl Mesh in (i) l ,j l ) Find its correspondence in the global suitability map data G gl Cell (i) in g ,j g ) Using G gl For G cl Carrying out fusion, specifically:
if G is gl (i g ,j g ) Effective and G cl (i l ,j l ) And if the data is invalid, then: g cl (i l ,j l )＝m g0 G gl (i g ,j g )；
If G is gl (i g ,j g ) Effective and G cl (i l ,j l ) And if effective, then:
G cl (i l ,j l )＝m g1 G gl (i g ,j g )+(1-m g1 )G cl (i l ,j l )；
wherein m is g0 The fitness forgetting factor is 0.95; m is g1 The fitness weighting coefficient is 0.9.
Finally, the fused local suitability map data G is utilized cl Mesh in (i) l ,j l ) Find its correspondence in the global suitability map data G gl Cell (i) in g ,j g ) Let G gl (i g, j g )＝G cl (i l, j l ) And finishing the updating of the global suitability map.
The method is also suitable for moving vehicles working on unstructured terrain, with small sensor sensing range and limited computer computing power, and realizes autonomous local obstacle avoidance in the continuous traveling process.
The above description is only for the best mode of the present invention, but the scope of the present invention is not limited thereto, and any changes or substitutions that can be easily conceived by those skilled in the art within the technical scope of the present invention are included in the scope of the present invention.

Claims (10)

1. A method for continuously and automatically driving, sensing and path planning in a non-structural scene under limited computing power is characterized by comprising a navigation control task and a sensing and obstacle avoidance planning task which are sequentially executed;
the navigation control task is a fixed-period task, and each period of the navigation control task executes navigation and control operation and judges whether to initiate a perception and obstacle avoidance task;
the sensing and obstacle avoidance planning task is an indefinite-period task and is initiated by a navigation control task; the task executes sensing and planning operation, and after the sensing and planning operation is completed, the state is fed back to the navigation control task.
2. The method for continuously and automatically sensing and planning the path of the unstructured scene under the finite computing power of claim 1, wherein the navigation control task comprises:
from an initial time t 0 At first, every T 1 Time period, i.e. t k ＝t 0 +i*T 1 I =0 to k; k represents the total number of cycles within the process; executing pose determination and motion control once; the position and posture determination is used for determining the position and posture of the moving vehicle under a certain specified navigation reference coordinate system, and the motion control is used for controlling the moving vehicle to move according to the output result of the obstacle avoidance plan;
after the pose determination and the motion control are executed in each period, the starting condition judgment of the sensing and obstacle avoidance task is executed, and whether the sensing and obstacle avoidance planning task is started or not is determined according to the result.
3. The method for continuously and automatically driving, sensing and path planning in the unstructured scene under the finite computing power according to claim 2, wherein the sensing and obstacle avoidance task starting condition judgment is carried out as follows:
if: l is u <V*T 2 ，L u ＝L-L e And then: starting a sensing and obstacle avoidance planning task; where L is the path length given by a perceptual planning task,L u for the remaining undriven length of the planned path, L e Is the length of the path, t, that has been traveled since the last perceptual planning task 0 Time L = L e ＝L u =0; v is the moving speed, T 2 Task execution time is planned for each perception.
4. The method for continuously and automatically sensing and planning the path of the unstructured scene under the limited computing power of claim 2, wherein the sensing and obstacle avoidance planning task comprises:
after receiving a sensing planning task starting instruction sent by a navigation control task, controlling a configured sensing sensor to acquire data;
after data acquisition is completed, environment modeling is carried out by using the current moment position and attitude information obtained by the navigation control task to obtain a rasterized digital elevation map;
after the environment modeling is completed, traffic analysis is completed on the rasterized digital elevation map, the traffic suitability of each grid is calculated, and a current traffic suitability map, namely a local suitability map, is obtained; the passing suitability is evaluated, if effective measurement data exist in the grid, the numerical value range is 0-1 after the suitability evaluation, the passing is impossible, and the passing is safe if 1 is available; if no valid measurement data exists in a certain grid, the value is assigned to be 2, namely invalid;
converting the historical traffic analysis result obtained by the last perception planning task to the current position, and fusing the suitability map to obtain a fused traffic suitability map, namely a global suitability map;
and planning a path on the fused global suitability map, and feeding back the state to the navigation control task after the path is planned.
5. The method of claim 4, wherein the suitability map is defined as follows:
the local suitability map is established by each perception planning task based on current perception data, and the global suitability map is obtained by fusing the local suitability maps of the previous times.
6. The method for continuously and automatically driving, sensing and path planning in an unstructured scene with limited computational power as claimed in claim 5, wherein the range of the global suitability map and the local suitability map are determined in advance according to the effective sensing range of the sensing sensor, and both the global suitability map and the local suitability map use the position of the unmanned automatic driving system as the origin.
7. The method as claimed in claim 6, wherein the local suitability map and the global suitability map both use a grid map approach, and wherein the grid division of the local suitability map into rows and columns is r M ×r M The global fitness map is partitioned into a number of rows and a number of columns r GM ×r GM And r is GM ≥n·r M Coefficient n>1; the traffic suitability evaluation values of each grid of the local suitability map and the global suitability map are respectively shown in an array G cl |r M ×r M And G gl |r GM ×r GM Performing the following steps; each data representation of the array has a size r GM ×r GM The traffic suitability evaluation value of each grid in the grid map of (1).
8. The method of claim 7, wherein the local suitability map data G is a global suitability map data cl And global suitability map data G gl The initial values are all unviable, and G is updated after the perception planning task is executed cl And G gl To data in an area overlapping the local suitability map range.
9. The method for continuously and automatically driving perception and path planning in an unstructured scene under limited computing power according to claim 8, wherein the fusion method of the suitability map fusion is as follows:
obtaining current local suitability map data G for each perception planning task cl Thereafter, a new global suitability map data G is defined gl_new From the current suitability map origin (x) 0 ,y 0 ) And the origin (x) of the suitability map during the last planning step L0 ,y L0 ) Calculating the relative position relation of the automatic unmanned system during two times of planning, and then utilizing G gl For G gl_new Assigning the grids of the corresponding areas, and finally ordering G gl ＝G gl_new Thereby completing the translation of the global suitability map;
for local suitability map data G cl Each grid (i) of l ,j l ) Find its correspondence in global suitability map data G gl Grid (i) of g ,j g ) Using G gl (i g ,j g ) For G cl (i l ,j l ) Fusing to obtain fused local suitability map data G cl ；
Using fused local suitability map data G cl Mesh in (i) l ,j l ) Find its correspondence in the global suitability map data G gl Cell (i) in (c) g ,j g ) Let G gl (i g ,j g )＝G cl (i l ,j l ) And finishing the updating of the global suitability map.
10. The method according to claim 9, wherein the fused local suitability map data G is obtained cl The method comprises the following steps:
if G is gl (i g ,j g ) Effective, i.e., the array is in the range of 0 to 1, and G cl (i l ,j l ) Invalid, i.e. array 2, then: g cl (i l ,j l )＝m g0 G gl (i g ,j g )；
If G is gl (i g ,j g ) Is provided withEffective and G cl (i l ,j l ) And if effective, then:
G cl (i l ,j l )＝m g1 G gl (i g ,j g )+(1-m g1 )G cl (i l ,j l )；
wherein m is g0 The value range is a forgetting factor of the fitness degree [0,1]；m g1 Is a fitness weighting coefficient with a value range of [0,1%]。
CN202210989378.3A 2022-08-17 2022-08-17 Non-structural scene continuous automatic driving perception and path planning method under limited computing power Pending CN115454051A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202210989378.3A CN115454051A (en) 2022-08-17 2022-08-17 Non-structural scene continuous automatic driving perception and path planning method under limited computing power

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202210989378.3A CN115454051A (en) 2022-08-17 2022-08-17 Non-structural scene continuous automatic driving perception and path planning method under limited computing power

Publications (1)

Publication Number Publication Date
CN115454051A true CN115454051A (en) 2022-12-09

Family

ID=84298139

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202210989378.3A Pending CN115454051A (en) 2022-08-17 2022-08-17 Non-structural scene continuous automatic driving perception and path planning method under limited computing power

Country Status (1)

Country Link
CN (1) CN115454051A (en)

Similar Documents

Publication Publication Date Title
CN110775052B (en) Automatic parking method based on fusion of vision and ultrasonic perception
CN109341706B (en) Method for manufacturing multi-feature fusion map for unmanned vehicle
CN102192745B (en) Position estimation device and position estimation method
US20170343374A1 (en) Vehicle navigation method and apparatus
CN106257242A (en) For regulating unit and the method for road boundary
CN110096053A (en) Driving locus generation method, system and machine readable media for automatic driving vehicle
JP2018154200A (en) Route determination device, vehicle control device, and route determination method and program
CN111812698A (en) Positioning method, device, medium and equipment
US11754415B2 (en) Sensor localization from external source data
CN110096054A (en) For using multiple threads to generate the method and system of the reference line for automatic driving vehicle
US20220315037A1 (en) Lane changing based only on local information
JPWO2018142566A1 (en) Passing gate determination device, vehicle control system, passing gate determination method, and program
US11436498B2 (en) Neural architecture search system for generating a neural network architecture
CN115525047A (en) Vehicle local track planning method and system with multi-type obstacle avoidance mode
CN111257853B (en) Automatic driving system laser radar online calibration method based on IMU pre-integration
CN115639823A (en) Terrain sensing and movement control method and system for robot under rugged and undulating terrain
WO2023097874A1 (en) Method and device for planning driving track
CN113405555B (en) Automatic driving positioning sensing method, system and device
CN111174765B (en) Planet vehicle target detection control method and device based on visual guidance
CN115454051A (en) Non-structural scene continuous automatic driving perception and path planning method under limited computing power
CN116625394A (en) Robot environment sensing and path optimizing system under unknown road condition
EP4134623A1 (en) Drive device, vehicle, and method for automated driving and/or assisted driving
US11708071B2 (en) Target-orientated navigation system for a vehicle using a generic navigation system and related method
CN116057578A (en) Modeling vehicle environment using cameras
US20230415781A1 (en) Systems and methods for controlling longitudinal acceleration based on lateral objects

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