CN110928297A - Intelligent bus route planning method based on multi-objective dynamic particle swarm optimization - Google Patents

Intelligent bus route planning method based on multi-objective dynamic particle swarm optimization Download PDF

Info

Publication number
CN110928297A
CN110928297A CN201911031001.1A CN201911031001A CN110928297A CN 110928297 A CN110928297 A CN 110928297A CN 201911031001 A CN201911031001 A CN 201911031001A CN 110928297 A CN110928297 A CN 110928297A
Authority
CN
China
Prior art keywords
dynamic
track
fitness
particle
static
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.)
Granted
Application number
CN201911031001.1A
Other languages
Chinese (zh)
Other versions
CN110928297B (en
Inventor
余伶俐
魏亚东
况宗旭
周开军
霍淑欣
王正久
白宇
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.)
Central South University
Original Assignee
Central South University
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 Central South University filed Critical Central South University
Priority to CN201911031001.1A priority Critical patent/CN110928297B/en
Publication of CN110928297A publication Critical patent/CN110928297A/en
Application granted granted Critical
Publication of CN110928297B publication Critical patent/CN110928297B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

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
    • G05D1/0231—Control of position or course in two dimensions specially adapted to land vehicles using optical position detecting means
    • G05D1/0238—Control of position or course in two dimensions specially adapted to land vehicles using optical position detecting means using obstacle or wall sensors
    • G05D1/024—Control of position or course in two dimensions specially adapted to land vehicles using optical position detecting means using obstacle or wall sensors in combination with a laser
    • 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
    • G05D1/0212—Control of position or course in two dimensions specially adapted to land vehicles with means for defining a desired trajectory
    • G05D1/0223—Control of position or course in two dimensions specially adapted to land vehicles with means for defining a desired trajectory involving speed control of the vehicle
    • 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
    • G05D1/0231—Control of position or course in two dimensions specially adapted to land vehicles using optical position detecting means
    • G05D1/0246—Control of position or course in two dimensions specially adapted to land vehicles using optical position detecting means using a video camera in combination with image processing means
    • 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
    • G05D1/0257—Control of position or course in two dimensions specially adapted to land vehicles using a radar
    • 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
    • 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
    • G05D1/0278—Control of position or course in two dimensions specially adapted to land vehicles using signals provided by a source external to the vehicle using satellite positioning signals, e.g. GPS

Abstract

The invention discloses an intelligent bus route planning method based on multi-objective dynamic particle swarm optimization, which comprises the following steps: acquiring vehicle and road information in real time to generate a global reference path; constructing a two-dimensional environment model based on the road rule lines and the global reference path, and initializing each particle in the particle swarm: each dimension of the particles corresponds to one coordinate point, and a curve segment is set between every two adjacent dimension coordinate points to obtain a track corresponding to the particles; designing a static multi-objective fitness function of the track according to the path length, the smoothness and the static safety index; then, extracting an optimal track candidate set by adopting a particle swarm algorithm and applying a static multi-target fitness function; and designing a dynamic multi-target fitness function and a constraint acceleration relation according to the dynamic barrier, combining with the static safety design fitness function, and selecting a track with optimal comprehensive fitness from the optimal track candidate set. The invention improves the comfort index and greatly improves the dynamic safety performance.

Description

Intelligent bus route planning method based on multi-objective dynamic particle swarm optimization
Technical Field
The invention relates to the technical field of intelligent driving and control thereof, in particular to an intelligent bus route planning method based on multi-objective dynamic particle swarm optimization.
Background
In recent years, intelligent driving is always the focus of social attention, with the continuous promotion of intelligent driving technology research, intelligent vehicles gradually move to the reality, and the structured urban roads and clear road signs also enable the urban buses to have relatively simple road conditions, so that the buses are promoted to become the breakthrough of the popularization of the automatic driving technology in daily life, and the core part of the automatic driving technology is to plan a safe and collision-free optimal driving path for the intelligent vehicles.
The path planning method is mainly researched to enable the intelligent vehicle to avoid obstacles. The existing path planning algorithm can be divided into two phases of global planning and local planning. In the global planning phase, global paths and vehicle states are determined by a digital map and positioning system. In the local planning stage, the local path may be realized according to the global path and peripheral information acquired by sensors such as a camera and a radar. However, a track planning method for intelligent buses and comprehensively considering static and dynamic multi-obstacle targets and riding comfort is lacked in the prior art.
Disclosure of Invention
The invention aims to solve the technical problem of providing an intelligent bus route planning method based on multi-objective dynamic particle swarm optimization, which greatly improves the dynamic safety performance while improving the comfort index.
In order to achieve the technical purpose, the invention adopts the following technical scheme:
an intelligent bus route planning method based on multi-objective dynamic particle swarm optimization comprises the following steps:
step 1, generating a global reference path;
acquiring vehicle and road information in real time according to a vehicle-mounted sensor, extracting a road rule line from the road information, and generating a global reference path according with the road rule line;
step 2, initializing a particle swarm and a corresponding track;
constructing a two-dimensional environment model with a longitudinal X axis in a forward direction and a transverse Y axis in a vertical direction based on a road rule line and a global reference path, and initializing a particle swarm in the two-dimensional environment model;
wherein the position encoding rule of each particle is as follows: each dimension corresponds to one coordinate point in the two-dimensional environment model, a curve segment is set between every two adjacent dimension coordinate points, the sum of M-1 curve segments obtains an optimizing track corresponding to the particles, and the particle swarm corresponds to obtain a track set;
step 3, static multi-target setting and dynamic obstacle avoidance are carried out;
designing a static multi-objective fitness function according to the length and smoothness of each optimizing track and a static safety index associated with a static obstacle;
adopting a particle swarm algorithm and applying a static multi-target fitness function, and screening out the first K optimizing tracks with optimal fitness from the particle swarm track set to form an optimal track candidate set;
and designing a dynamic multi-target fitness function according to the dynamic barrier and the relation between the dynamic barrier and the current position, combining the dynamic multi-target fitness function with the static safety design fitness function to obtain a comprehensive fitness function, and selecting a track with the optimal comprehensive fitness from the optimal track candidate set.
Furthermore, curves set between every two adjacent dimensional coordinate points of each particle are obtained by a cubic spline interpolation method.
Further, when generating the M-1 spline curve, the following constraints are included:
a, starting point constraint, wherein the function value and the first derivative value of the starting point coordinate are both 0;
b, end point constraint, wherein the function value of the end point coordinate is known and the first derivative value is 0;
and c, constraining intermediate points, wherein the function values of each intermediate point on two adjacent curve segments are the same, the first-order derivative values are the same, and the second-order derivative values are the same.
Further, the expression of the static multi-objective fitness function is specifically:
fstastic_fitness＝ωdffit_distence+ωsffit_smoothness+ωuffit_security；
in the above formula, fstastic_fitnessStatic multi-objective fitness, f, representing a trajectoryfit_distenceIndicating the length of the track, ωdRepresenting length weight, ffit_smoothnessIndicates smoothness, ωsRepresenting a smoothness weight, ffit_securityIndicating static degree of safety, ωuRepresents a static security level weight and has:
Figure RE-GDA0002276918520000031
Figure RE-GDA0002276918520000032
Figure RE-GDA0002276918520000033
in the above formula, fcalculate_securityRepresenting the shortest distance, f, of the obstacle to each curved section of the trajectory(j+1)Represents a curve segment of the track between the j-th dimensional coordinate and the j + 1-th dimensional coordinate, f ″(j+1)Representing a curve segment f(j+1)(x) of (c)j,yj) The coordinates of j dimension of the particles are represented, and the horizontal coordinates of M dimensions of the particles are distributed at equal intervals in sequence, wherein the interval is L1；
If the shortest distance from the obstacle to the track is smaller than the geometric radius of the obstacle, setting the static safety degree as follows: f. ofcalculate_security＝9999。
Further, ω isd＝4，ωs＝400，ωu＝2。
Further, the design rule of the dynamic multi-objective fitness function is as follows:
fdynamic_fitness＝ωdynamic(L4-r)|a|，
Figure RE-GDA0002276918520000034
Figure RE-GDA0002276918520000035
Figure RE-GDA0002276918520000036
in the formula (f)dynamic_fitnessDynamic multi-target fitness, omega, representing trajectorydynamicRepresenting the dynamic fitness weight, a representing the acceleration of the vehicle travelling along the trajectory, L4Indicating the distance travelled by the vehicle following the surrounding vehicle along the track, L5Represents the path arc length, v, of the vehicle between the current position of the trajectory and the collision point of the dynamic collision object0Indicating the initial speed, v, of the vehicle travelling along the tracklimitIndicating the speed limit of the vehicle.
Further, the dynamic collider keeps uniform linear motion, only deceleration avoidance is considered when the track avoids the dynamic barrier, and the track in the optimal track candidate set has the following conditions when the track avoids the dynamic barrier: the fitness value of the collision point of the track and the dynamic obstacle is larger than that of the collision point of the track and the dynamic obstacle, namely: f. of1 nocollison_fitness＝f1 stastic_fitness+0>+f2 collison_fitness＝f2 stastic_fitness+f2 dynamic_fitness。
Further, the specific process of step 1 is:
step 1.1, obtaining vehicle and road information, and extracting road ruled lines from the vehicle and road information;
acquiring current positioning and state information of a vehicle by using a positioning device, acquiring road edge data by using a radar and extracting geometric description of a road edge, and acquiring lane line data by using a camera and extracting geometric description of a lane line;
step 1.2, generating a global reference path;
if the road ruled line extracted in the step 1.1 comprises the geometric description of the lane line, the global reference path is obtained by moving the geometric description which is parallel to the lane line and is opposite to the right side of the central lane line by a half lane to the left;
if the road information extracted in the step 1.1 does not include the geometric description of the lane line but includes the geometric description of the road edge, the global reference path is obtained by moving a half lane left relative to the right road edge and is parallel to the road edge;
if the road information extracted in step 1.1 does not include the geometric description of the lane line and the geometric description of the road edge, detecting whether the sensor has a fault, and re-executing step 1 under the normal condition of the sensor.
Further, the method for initializing the particle swarm in the two-dimensional environment model comprises the following steps:
initially encoding the positions of the particles into a set of M-dimensional coordinate points in a two-dimensional environment model: xi＝([x1,y1],[x2,y2]，…，[xM,yM]) I represents a particle number in the particle group, and M represents a dimension of the particle;
wherein the longitudinal de-fixing of the particle code is [ x ]1,x2,…,xM]＝[0,L1,…,(M-1)L1]，
Separating the lateral solution space of the particle [ -L ]2,L2]At an interval of L3Is divided into
Figure RE-GDA0002276918520000041
A discrete value, i.e. the transverse solution space of the particle is
Figure RE-GDA0002276918520000042
j represents the dimension index of the particle;
and setting a curve segment between every two adjacent dimensional coordinate points of the particles, and obtaining the optimization track corresponding to the particles by the sum of the M-1 curve segments, wherein the particle swarm corresponds to a track set.
Advantageous effects
The invention has the beneficial effects that:
1. the multi-objective fitness function design is designed by comprehensively considering performance indexes such as the length, smoothness, static safety, dynamic safety and the like of the track, so that the safety and the comfort of intelligent bus route planning are greatly improved;
2. due to the introduction of the dynamic barrier, the dynamic safety of the trajectory planning is improved, and meanwhile, the vehicle acceleration planning is output, so that the method is particularly suitable for intelligent buses;
3. by adopting a particle swarm algorithm, when a curve between every two adjacent dimensional coordinates of the multi-dimensional particles is fitted, the constraints of a starting point, a terminal point and a middle point are considered, the smoothness of the track is increased, and the comfort of path planning is further improved;
4. the particle swarm algorithm is adopted, the longitudinal solution of the particle codes is fixed according to the forward direction characteristic of the path planning, the transverse solution is discretized into a limited number of values according to the deviation upper limit of the track, the particle optimization can be completed only by setting the population quantity and the iteration period of the small particle swarm, and the completion period of the path planning is greatly shortened.
5. The completion period based on the path planning is short, the motion of the dynamic barrier can be simplified into uniform linear motion when the dynamic barrier avoidance planning is carried out, and the difficulty of the path planning by the dynamic barrier avoidance is reduced while the safety of the path planning is ensured.
Drawings
FIG. 1 is a flow chart of multi-objective dynamic particle swarm optimization path planning according to an embodiment of the present invention;
FIG. 2 is a schematic diagram of a two-dimensional environment model established according to an embodiment of the present invention;
FIG. 3 is a flowchart illustrating an optimal trajectory candidate set generation and multi-target setting process according to an embodiment of the present invention;
fig. 4 is a flow chart of dynamic obstacle avoidance according to an embodiment of the present invention.
Detailed Description
The following describes embodiments of the present invention in detail, which are developed based on the technical solutions of the present invention, and give detailed implementation manners and specific operation procedures to further explain the technical solutions of the present invention.
In the embodiment, the intelligent vehicle refitted from the bus with the length of 12m, the width of 2.5m and the height of 3.3m is provided with the laser radar, the millimeter wave radar, the GPS positioning system and the machine vision system, and a path planning experiment is carried out on a structured road.
The intelligent bus route planning method based on multi-objective dynamic particle swarm optimization provided by the embodiment is shown in fig. 1, 3 and 4, and comprises the following processes:
step 1, generating a global reference path;
the method comprises the following steps of acquiring vehicle and road information in real time according to a vehicle-mounted sensor, extracting a road rule line from the road information, and generating a global reference path according with the road rule line, wherein the specific process comprises the following steps:
step 1.1, obtaining vehicle and road information, and extracting road ruled lines from the vehicle and road information;
the vehicle information is in particular the current position and status information of the vehicle, i.e.
Figure RE-GDA0002276918520000061
Obtained by using a GPS positioning system.
The road information specifically refers to road edge data and lane line data. The road edge data are acquired through a laser radar and a millimeter wave radar, and then the geometric description of the road edge can be extracted from the road edge data: including geometric description of the left edge
Figure RE-GDA0002276918520000062
And geometric description of the right edge
Figure RE-GDA0002276918520000063
The lane line data is acquired by an industrial camera and then the number of lane lines can be countedExtracting geometric description of lane line
Figure RE-GDA0002276918520000064
Wherein i represents the serial number of the lane lines, i belongs to { 0., K }, and K is the total number of the detected lane lines. The geometric description of the road edge and the geometric description of the lane line are the extracted road ruled line.
Step 1.2, generating a global reference path;
if the road ruled line obtained in step 1.1 includes the geometric description of the lane line, the global reference path
Figure RE-GDA0002276918520000065
Obtained by shifting the geometric description parallel to the lane line and to the right of the central lane line by half a lane to the left, i.e.
Figure RE-GDA0002276918520000066
k∈{1,2,3}，
Figure RE-GDA0002276918520000067
If the road information obtained in step 1.1 does not include the geometric description of the lane line but includes the geometric description of the road edge, the global reference path
Figure RE-GDA0002276918520000068
Parallel to the road edge and shifted to the left by half a lane relative to the lateral road edge, i.e.
Figure RE-GDA0002276918520000069
Or
Figure RE-GDA00022769185200000610
And is provided with
Figure RE-GDA00022769185200000611
Or
Figure RE-GDA00022769185200000612
If the road information obtained in step 1.1 does not include the geometric description of the lane line and the geometric description of the road edge, detecting whether the sensor has a fault, and re-executing step 1 under the normal condition of the sensor.
Step 2, initializing a particle swarm and a corresponding track;
constructing a two-dimensional environment model with a longitudinal X axis in a forward direction and a transverse Y axis in a vertical direction based on the road ruled lines and the global reference path, as shown in FIG. 2, wherein particle swarms are initialized in the two-dimensional environment model; wherein the position encoding rule of each particle is as follows: each dimension corresponds to one coordinate point in the two-dimensional environment model, a curve segment is obtained between every two adjacent dimension coordinate points by adopting a cubic spline interpolation method, the sum of M-1 curve segments obtains an optimizing track corresponding to the particles, and the particle swarm corresponds to obtain a track set;
the method for initializing the particle swarm in the two-dimensional environment model specifically comprises the following steps:
step 2.1, initially encoding the positions of the particles into M coordinate point sets in a two-dimensional environment model: xi＝([x1,y1],[x2,y2]，…，[xM,yM]) I represents a particle number in the particle group, M represents a dimension of the particle, and M is 5 in this embodiment;
step 2.2, the longitudinal de-fixing of the particle code is then [ x ]1,x2,…,xM]＝[0,L1,…,(M-1)L1]And separating the longitudinal solution space [ -L ] of the particles2,L2]At an interval of L3Is divided into
Figure RE-GDA0002276918520000071
Discrete values, i.e. the longitudinal solution space of the particles is defined as
Figure RE-GDA0002276918520000072
j represents the dimension index of the particle; wherein L2The specific value is the offset upper limit of the track offset global reference path. In this embodiment, L is set2＝6m，L3＝0.2m。
And 2.3, setting a curve segment between every two adjacent dimensional coordinate points of the particles, and obtaining the optimization track corresponding to the particles by the sum of the M-1 curve segments, wherein the particle swarm corresponds to a track set.
In this embodiment, each curve segment is obtained by a cubic spline interpolation method, and the kth curve segment is represented as
Figure RE-GDA0002276918520000073
k is 1, 2, …, M-1. When generating the M-1 curve segments, the following constraints are included:
a, starting point constraint, and the function value and the first derivative value of the starting point coordinate are both 0, namely:
Figure RE-GDA0002276918520000074
b, end point constraint, function value of end point coordinate is known and first derivative value is 0, i.e.
Figure RE-GDA0002276918520000075
c, constraint of intermediate points, wherein the intermediate points are positioned at the joint of two adjacent curve segments, and the function value of each intermediate point on the two adjacent curve segments is the same
Figure RE-GDA0002276918520000076
The first derivative values are the same
Figure RE-GDA0002276918520000077
Second derivative value is the same
Figure RE-GDA0002276918520000078
Step 3, static multi-target setting and dynamic obstacle avoidance are carried out;
step 3.1, designing a static multi-objective fitness function:
designing a static multi-objective fitness function according to the length and smoothness of each optimizing track and a static safety index associated with a static obstacle, wherein the expression is as follows:
fstastic_fitness＝ωdffit_distence+ωsffit_smoothness+ωuffit_security；
in the above formula, fstastic_fitnessStatic multi-objective fitness, f, representing a trajectoryfit_distenceDenotes the length, ωdRepresenting length weight, ffit_smoothnessIndicates smoothness, ωsRepresenting a smoothness weight, ffit_securityIndicating static degree of safety, ωuRepresents the static security weight, in this example, ωd＝4，ωs＝400，ωu＝2。
Wherein the length ffit_distenceSince the vertical solutions of all particle codes are fixed to [ x ] in this embodiment1,x2,…,xM]＝[0,L1,…,(M-1)L1]I.e. the length of the trajectory represented by the particle is equal in the longitudinal direction, the length in the transverse direction can be considered when designing the fitness function, i.e. only the transverse distances of the curve segments are summed up, as follows:
Figure RE-GDA0002276918520000081
(xj,yj) A coordinate representing the jth dimension of the particle,
smoothness ffit_smoothnessSince the trajectory represented by the particle with each dimension M is divided into M-1 curve segments and the curvature (second derivative) has a positive or negative division, the embodiment designs the smoothness of the trajectory represented by the particle as the sum of the squares of the curvatures of the M-1 curve segments, which is expressed as:
Figure RE-GDA0002276918520000082
f(j+1)represents a curve segment of the track between the j-th dimensional coordinate and the j + 1-th dimensional coordinate, f ″(j+1)Representing a curve segment f(j+1)The curvature of (a);
degree of static safety ffit_securityConsidering the existence of obstacles in the path planning process, the obstacles are set as central circles with the geometric radius of r and the pose coordinates of r[x,y]Then the static security level is expressed as:
Figure RE-GDA0002276918520000083
fcalculate_securityrepresenting the shortest distance of the obstacle to each curved segment in the trajectory. If the shortest distance from the obstacle to the track is smaller than the geometric radius of the obstacle, setting the static safety degree as follows: f. ofcalculate_security＝9999。
Step 3.2, extracting an optimal track candidate set:
and adopting a particle swarm algorithm and applying a static multi-target fitness function, and screening the first K optimizing tracks with the optimal fitness from the particle swarm track set to form an optimal track candidate set.
After the population size of the particle swarm and the highest iteration number of the particle swarm algorithm are set, and the particle positions of the particle swarm are initialized and encoded according to the step 2, the static multi-target fitness function designed in the step 3.1 can be used as the fitness function of the particle swarm algorithm, and the particle swarm is iteratively updated by adopting the following position and speed updating formula:
Vi＝ωVi+c1r1(Pbest-Xi)+c2r2(Gbest-Xi)，
Xi＝Xi+Vi，
wherein Vi＝(vi1,vi2,...,viM) Representing the velocity of the ith particle, with dimension M being the same as the particle dimension; c. C1＝c10.5 is a learning factor, also called acceleration constant, ω is an inertia factor, r1、r2Is [0,1 ]]Random number between, PbestThe optimal position for the ith particle is called the individual extremum, GbestThe optimal position searched by the current particle swarm is called a global extremum. Due to the speed updating of c in the formula1、c2、r1、r2And omega, the calculated particle velocity can generate decimal number, the decimal number is rounded, and the velocity is integer to be in accordance with the form of particle coding integer.
And when the maximum iteration times is reached, selecting the first K extreme values with the minimum fitness from all the individual extreme values of the particle swarm, wherein the tracks corresponding to the K extreme value particles are the first K optimizing tracks with the optimal fitness. In this embodiment, K is 5.
Step 3.3, dynamic obstacle avoidance:
because the completion period of the path planning is in millisecond level, especially when particle encoding is carried out by adopting a particle swarm algorithm in the embodiment, the longitudinal solution of the particle position encoding is fixed, and the transverse solution is dispersed into a limited number according to the deviation upper limit of the track, the particle optimization can be completed only by setting the smaller particle swarm number and the iteration period, and the completion period of the path planning is greatly shortened. Therefore, the present embodiment may assume that the moving speed and the moving direction of the dynamic obstacle are substantially unchanged within the millisecond time of the path planning, so as to assume that the dynamic obstacle keeps moving linearly within the path planning time.
The present embodiment additionally assumes the following: (1) the fitness value of the collision point of the track and the dynamic obstacle is larger than that of the collision point of the track and the dynamic obstacle, namely: f. of1 nocollison_fitness＝f1 stastic_fitness+0>+f2 collison_fitness＝f2 stastic_fitness+f2 dynamic_fitness(ii) a (2) The dynamic barrier avoidance only considers deceleration avoidance and does not consider the condition of accelerating overtaking.
Therefore, after all tracks in the optimal track candidate set are subjected to static obstacle avoidance, a dynamic multi-target fitness function and an acceleration constraint relation of the vehicle running along the tracks can be designed according to the dynamic obstacles and the relation between the dynamic obstacles and the current position, the dynamic multi-target fitness function is combined with the static safety design fitness function to obtain a comprehensive fitness function, and one track with the optimal comprehensive fitness is selected from the optimal track candidate set on the basis of meeting the acceleration constraint relation.
The specific dynamic obstacle avoidance planning method comprises the following steps:
according to the assumed conditions, the dynamic barrier keeps uniform linear motion to generate a track line, and the intersection point of the track line and the track in the optimal track candidate set is the collision point.
Firstly, intercepting a path between the current position of each track in the optimal track candidate set and a collision point, wherein the arc length of the path is L5(ii) a And the time required for the vehicle to travel from the starting point to the collision point along the track is v0t＝L2-r，v0Representing an initial speed of the vehicle traveling along the track;
then, the vehicle acceleration constraint relation is set: the running distance of the vehicle following the dynamic barrier along the track is L4And satisfy
Figure RE-GDA0002276918520000091
The acceleration takes the value of
Figure RE-GDA0002276918520000092
And acceleration satisfies
Figure RE-GDA0002276918520000093
vlimitIndicating the speed limit of the vehicle. In the present embodiment, v is set0＝30km/h，vlimit＝10km/h。
And finally, aiming at K tracks in the optimal track candidate set, introducing a dynamic multi-target fitness function, and comprehensively considering the static and dynamic multi-target fitness functions to finally generate the optimal track and the acceleration of dynamic obstacle avoidance. The dynamic multi-target fitness function is as follows:
fdynamic_fitness＝ωdynamic(L4-r)|a|；
in the formula (f)dynamic_fitnessDynamic multi-target fitness, omega, representing trajectorydynamicRepresenting the dynamic fitness weight, taking omegadynamic＝0.25。
The above embodiments are preferred embodiments of the present application, and those skilled in the art can make various changes or modifications without departing from the general concept of the present application, and such changes or modifications should fall within the scope of the claims of the present application.

Claims (9)

1. An intelligent bus route planning method based on multi-objective dynamic particle swarm optimization is characterized by comprising the following steps:
step 1, generating a global reference path;
acquiring vehicle and road information in real time according to a vehicle-mounted sensor, extracting a road rule line from the road information, and generating a global reference path according with the road rule line;
step 2, initializing a particle swarm and a corresponding track;
constructing a two-dimensional environment model with a longitudinal X axis in a forward direction and a transverse Y axis in a vertical direction based on a road rule line and a global reference path, and initializing a particle swarm in the two-dimensional environment model;
wherein the position encoding rule of each particle is as follows: each dimension corresponds to one coordinate point in the two-dimensional environment model, a curve segment is set between every two adjacent dimension coordinate points, the sum of M-1 curve segments obtains an optimizing track corresponding to the particles, and the particle swarm corresponds to obtain a track set;
step 3, static multi-target setting and dynamic obstacle avoidance are carried out;
designing a static multi-objective fitness function according to the length and smoothness of each optimizing track and a static safety index associated with a static obstacle;
adopting a particle swarm algorithm and applying a static multi-target fitness function, and screening out the first K optimizing tracks with optimal fitness from the particle swarm track set to form an optimal track candidate set;
and designing a dynamic multi-target fitness function according to the dynamic barrier and the relation between the dynamic barrier and the current position, combining the dynamic multi-target fitness function with the static safety design fitness function to obtain a comprehensive fitness function, and selecting a track with the optimal comprehensive fitness from the optimal track candidate set.
2. The method according to claim 1, wherein the curve set between every two adjacent dimensional coordinate points of each particle is obtained by a cubic spline interpolation method.
3. The method of claim 2, wherein the following constraints are included in generating the M-1 spline curve:
a, starting point constraint, wherein the function value and the first derivative value of the starting point coordinate are both 0;
b, end point constraint, wherein the function value of the end point coordinate is known and the first derivative value is 0;
and c, constraining intermediate points, wherein the function values of each intermediate point on two adjacent curve segments are the same, the first-order derivative values are the same, and the second-order derivative values are the same.
4. The method according to claim 1, wherein the expression of the static multi-objective fitness function is specifically:
fstastic_fitness＝ωdffit_distence+ωsffit_smoothness+ωuffit_security；
in the above formula, fstastic_fitnessStatic multi-objective fitness, f, representing a trajectoryfit_distenceIndicating the length of the track, ωdRepresenting length weight, ffit_smoothnessIndicates smoothness, ωsRepresenting a smoothness weight, ffit_securityIndicating static degree of safety, ωuRepresents a static security level weight and has:
Figure FDA0002250143700000021
Figure FDA0002250143700000022
Figure FDA0002250143700000023
in the above formula, fcalculate_securityRepresenting obstacles to curved sections of the trajectoryThe shortest distance of f(j+1)Represents a curve segment of the track between the j-th dimensional coordinate and the j + 1-th dimensional coordinate, f ″(j+1)Representing a curve segment f(j+1)(x) of (c)j,yj) The coordinates of j dimension of the particles are represented, and the horizontal coordinates of M dimensions of the particles are distributed at equal intervals in sequence, wherein the interval is L1；
If the shortest distance from the obstacle to the track is smaller than the geometric radius of the obstacle, setting the static safety degree as follows: f. ofcalculate_security＝9999。
5. The method of claim 4, wherein ω isd＝4，ωs＝400，ωu＝2。
6. The method of claim 1, wherein the design rule of the dynamic multi-objective fitness function is:
fdynamic_fitness＝ωdynamic(L4-r)|a|，
Figure FDA0002250143700000024
Figure FDA0002250143700000025
Figure FDA0002250143700000026
in the formula (f)dynamic_fitnessDynamic multi-target fitness, omega, representing trajectorydynamicRepresenting the dynamic fitness weight, a representing the acceleration of the vehicle travelling along the trajectory, L4Indicating the distance travelled by the vehicle following the surrounding vehicle along the track, L5Represents the path arc length, v, of the vehicle between the current position of the trajectory and the collision point of the dynamic collision object0Indicating the initial speed, v, of the vehicle travelling along the tracklimitIndicating the speed limit of the vehicle.
7. The method according to claim 6, wherein the dynamic collider keeps a uniform linear motion, the trajectory only considers deceleration avoidance when avoiding the dynamic obstacle, and the trajectory in the optimal trajectory candidate set has the following conditions when avoiding the dynamic obstacle: the fitness value of the collision point of the track and the dynamic obstacle is larger than that of the collision point of the track and the dynamic obstacle, namely: f. of1 nocollison_fitness＝f1 stastic_fitness+0>+f2 collison_fitness＝f2 stastic_fitness+f2 dynamic_fitness。
8. The method according to claim 1, wherein the specific process of step 1 is as follows:
step 1.1, obtaining vehicle and road information, and extracting road ruled lines from the vehicle and road information;
acquiring current positioning and state information of a vehicle by using a positioning device, acquiring road edge data by using a radar and extracting geometric description of a road edge, and acquiring lane line data by using a camera and extracting geometric description of a lane line;
step 1.2, generating a global reference path;
if the road ruled line extracted in the step 1.1 comprises the geometric description of the lane line, the global reference path is obtained by moving the geometric description which is parallel to the lane line and is opposite to the right side of the central lane line by a half lane to the left;
if the road information extracted in the step 1.1 does not include the geometric description of the lane line but includes the geometric description of the road edge, the global reference path is obtained by moving a half lane left relative to the right road edge and is parallel to the road edge;
if the road information extracted in step 1.1 does not include the geometric description of the lane line and the geometric description of the road edge, detecting whether the sensor has a fault, and re-executing step 1 under the normal condition of the sensor.
9. The method of claim 1, wherein the method for initializing the particle swarm in the two-dimensional environment model comprises:
initially encoding the positions of the particles into a set of M-dimensional coordinate points in a two-dimensional environment model: xi＝([x1,y1],[x2,y2]，…，[xM,yM]) I represents a particle number in the particle group, and M represents a dimension of the particle;
wherein the longitudinal de-fixing of the particle code is [ x ]1,x2,…,xM]＝[0,L1,…,(M-1)L1]，
Separating the lateral solution space of the particle [ -L ]2,L2]At an interval of L3Is divided into
Figure FDA0002250143700000031
A discrete value, i.e. the transverse solution space of the particle is
Figure FDA0002250143700000032
j represents the dimension index of the particle;
and setting a curve segment between every two adjacent dimensional coordinate points of the particles, and obtaining the optimization track corresponding to the particles by the sum of the M-1 curve segments, wherein the particle swarm corresponds to a track set.
CN201911031001.1A 2019-10-28 2019-10-28 Intelligent bus route planning method based on multi-objective dynamic particle swarm optimization Active CN110928297B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201911031001.1A CN110928297B (en) 2019-10-28 2019-10-28 Intelligent bus route planning method based on multi-objective dynamic particle swarm optimization

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201911031001.1A CN110928297B (en) 2019-10-28 2019-10-28 Intelligent bus route planning method based on multi-objective dynamic particle swarm optimization

Publications (2)

Publication Number Publication Date
CN110928297A true CN110928297A (en) 2020-03-27
CN110928297B CN110928297B (en) 2020-10-30

Family

ID=69849576

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201911031001.1A Active CN110928297B (en) 2019-10-28 2019-10-28 Intelligent bus route planning method based on multi-objective dynamic particle swarm optimization

Country Status (1)

Country Link
CN (1) CN110928297B (en)

Cited By (11)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111766875A (en) * 2020-06-18 2020-10-13 珠海格力智能装备有限公司 Obstacle avoidance method and device for dust collection equipment and electronic equipment
CN112148002A (en) * 2020-08-31 2020-12-29 西安交通大学 Local trajectory planning method, system and device
CN112562372A (en) * 2020-11-30 2021-03-26 腾讯科技（深圳）有限公司 Track data processing method and related device
CN112859845A (en) * 2021-01-04 2021-05-28 广州小鹏自动驾驶科技有限公司 Trajectory planning method and device and vehicle
CN113011060A (en) * 2021-03-03 2021-06-22 西北工业大学 Temperature process curve optimization method and system for composite material
CN113759887A (en) * 2020-06-04 2021-12-07 北京京东乾石科技有限公司 Path planning method and device, storage medium and electronic equipment
CN114035586A (en) * 2021-11-22 2022-02-11 江苏科技大学 Workshop AGV trolley path planning method for improving ant colony algorithm and dynamic window
WO2022083166A1 (en) * 2020-10-19 2022-04-28 山东交通学院 Method and system for reconstructing vehicle's driving trajectory when checkpoint data is missing
CN114413906A (en) * 2022-01-18 2022-04-29 哈尔滨工业大学 Three-dimensional trajectory planning method based on improved particle swarm optimization algorithm
CN116811884A (en) * 2023-08-25 2023-09-29 格陆博科技有限公司 Intelligent driving environment perception analysis method and system
CN117250868A (en) * 2023-11-14 2023-12-19 泉州装备制造研究所 Shunt control method, system and storage medium based on multi-target particle swarm algorithm

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106843236A (en) * 2017-03-31 2017-06-13 深圳市靖洲科技有限公司 The unmanned bicycle paths planning method of particle cluster algorithm is improved based on weight
CN107992051A (en) * 2017-12-26 2018-05-04 江南大学 Unmanned vehicle paths planning method based on improved multi-objective particle swarm algorithm
CN108645411A (en) * 2018-05-15 2018-10-12 深圳大学 Robot path planning method, device based on particle cluster algorithm and terminal device
CN110081894A (en) * 2019-04-25 2019-08-02 同济大学 A kind of real-time planing method of unmanned wheel paths based on the fusion of road structure weight

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106843236A (en) * 2017-03-31 2017-06-13 深圳市靖洲科技有限公司 The unmanned bicycle paths planning method of particle cluster algorithm is improved based on weight
CN107992051A (en) * 2017-12-26 2018-05-04 江南大学 Unmanned vehicle paths planning method based on improved multi-objective particle swarm algorithm
CN108645411A (en) * 2018-05-15 2018-10-12 深圳大学 Robot path planning method, device based on particle cluster algorithm and terminal device
CN110081894A (en) * 2019-04-25 2019-08-02 同济大学 A kind of real-time planing method of unmanned wheel paths based on the fusion of road structure weight

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
LINGLI YU等: "A Path Planning and Navigation Control System Design for Driverless Electric Bus", 《IEEE ACCESS》 *
余伶俐等: "智能车辆深度强化学习的模型迁移轨迹规划方法", 《控制理论与应用》 *

Cited By (17)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113759887A (en) * 2020-06-04 2021-12-07 北京京东乾石科技有限公司 Path planning method and device, storage medium and electronic equipment
CN111766875A (en) * 2020-06-18 2020-10-13 珠海格力智能装备有限公司 Obstacle avoidance method and device for dust collection equipment and electronic equipment
CN112148002A (en) * 2020-08-31 2020-12-29 西安交通大学 Local trajectory planning method, system and device
CN112148002B (en) * 2020-08-31 2021-12-28 西安交通大学 Local trajectory planning method, system and device
WO2022083166A1 (en) * 2020-10-19 2022-04-28 山东交通学院 Method and system for reconstructing vehicle's driving trajectory when checkpoint data is missing
CN112562372A (en) * 2020-11-30 2021-03-26 腾讯科技（深圳）有限公司 Track data processing method and related device
CN112562372B (en) * 2020-11-30 2021-11-16 腾讯科技（深圳）有限公司 Track data processing method and related device
CN112859845A (en) * 2021-01-04 2021-05-28 广州小鹏自动驾驶科技有限公司 Trajectory planning method and device and vehicle
CN113011060A (en) * 2021-03-03 2021-06-22 西北工业大学 Temperature process curve optimization method and system for composite material
CN114035586A (en) * 2021-11-22 2022-02-11 江苏科技大学 Workshop AGV trolley path planning method for improving ant colony algorithm and dynamic window
CN114035586B (en) * 2021-11-22 2024-03-29 江苏科技大学 Workshop AGV trolley path planning method for improving ant colony algorithm and dynamic window
CN114413906A (en) * 2022-01-18 2022-04-29 哈尔滨工业大学 Three-dimensional trajectory planning method based on improved particle swarm optimization algorithm
CN114413906B (en) * 2022-01-18 2022-12-13 哈尔滨工业大学 Three-dimensional trajectory planning method based on improved particle swarm optimization algorithm
CN116811884A (en) * 2023-08-25 2023-09-29 格陆博科技有限公司 Intelligent driving environment perception analysis method and system
CN116811884B (en) * 2023-08-25 2023-11-28 格陆博科技有限公司 Intelligent driving environment perception analysis method and system
CN117250868A (en) * 2023-11-14 2023-12-19 泉州装备制造研究所 Shunt control method, system and storage medium based on multi-target particle swarm algorithm
CN117250868B (en) * 2023-11-14 2024-03-01 泉州装备制造研究所 Shunt control method, system and storage medium based on multi-target particle swarm algorithm

Also Published As

Publication number Publication date
CN110928297B (en) 2020-10-30

Similar Documents

Publication Publication Date Title
CN110928297B (en) Intelligent bus route planning method based on multi-objective dynamic particle swarm optimization
Dai et al. Modeling vehicle interactions via modified LSTM models for trajectory prediction
CN110749333B (en) Unmanned vehicle motion planning method based on multi-objective optimization
CN112673234B (en) Path planning method and path planning device
WO2018176593A1 (en) Local obstacle avoidance path planning method for unmanned bicycle
Pérez et al. Trajectory generator for autonomous vehicles in urban environments
Khodayari et al. A historical review on lateral and longitudinal control of autonomous vehicle motions
CN108088456A (en) A kind of automatic driving vehicle local paths planning method with time consistency
CN107813820A (en) A kind of unmanned vehicle lane-change paths planning method for imitating outstanding driver
CN114005280B (en) Vehicle track prediction method based on uncertainty estimation
Yi et al. Model predictive trajectory planning for automated driving
CN112577506B (en) Automatic driving local path planning method and system
US20230150549A1 (en) Hybrid log simulated driving
CN116118780A (en) Vehicle obstacle avoidance track planning method, system, vehicle and storage medium
Zhang et al. Structured road-oriented motion planning and tracking framework for active collision avoidance of autonomous vehicles
CN116069037A (en) Unmanned vehicle formation track planning control method, device, equipment and storage medium
Malayjerdi et al. Practical path planning techniques in overtaking for autonomous shuttles
CN112706760B (en) Unmanned parking path planning method for special road scene
CN113515111B (en) Vehicle obstacle avoidance path planning method and device
Park et al. Path generation algorithm based on crash point prediction for lane changing of autonomous vehicles
CN114537435A (en) Real-time whole vehicle track planning method in automatic driving
CN115079699A (en) Model prediction control-based motion planning method for man-machine co-driving automobile
CN115246394A (en) Intelligent vehicle overtaking obstacle avoidance method
CN117916682A (en) Motion planning using a time-space convex corridor
Fu et al. Unmanned Driving Technology in Coal Mine Based on Semantic Information Method

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant