CN103077425A - Immune genetic algorithm for AUV (Autonomous Underwater Vehicle) real-time path planning - Google Patents

Abstract

The invention relates to a real-time path planning method of AUV (Autonomous Underwater Vehicle), in particular to a method for carrying out online, real-time local path planning according to an online map in an AUV real-time collision preventation process. The method comprises the steps of: setting the quantity of small populations according to the quantity of path points of the AUV, initializing; carrying out immune selection on each small population to obtain subgroups; carrying out genetic manipulation on one subgroup, carrying out cell cloning on the other subgroup; then clustering through a vaccination and an antibody to form the next generation of small population, judging whether the next generation of small population meets the conditions or not; if yes, selecting optimal individuals of the small populations; and selecting the optimal individuals from the set consisting of all optimal individuals to be used as a planning path. According to the invention, the diversity of the population is maintained by using an antibody clustering principle, the premature convergence of an algorithm is avoided, and the global optimization is facilitated. The established immune genetic algorithm is used for clustering and analyzing generated filial generations by adopting a self-regulating mechanism, and the diversity of the population is ensured.

Description

A kind of immune genetic algorithm for the planning of AUV real-time route

Technical field

The present invention relates to the real-time route planing method of autonomous underwater robot (AUV, Autonomous Underwater Vehicle), more particularly, be in the AUV Realtime collision free process according to Online Map carry out online, the real-time method of local paths planning.

Background technology

Autonomous underwater robot is a kind of self-contained energy, relies on autonomous navigation system, and by the programmed decision-making of intelligence, operating area, the underwater vehicle of the mission that independently fulfils assignment are arrived in autonomous navigation.Independence requires autonomous underwater robot independently to adapt to changeable, complicated external environment under without extraneous controlled condition, particularly will tackle the obstacle of prior the unknown.This just needs autonomous underwater robot when possessing the Realtime collision free function, also should possess the ability of online, real-time local paths planning.

Real-time route planning is defined as: the Online Map that generates according to sensor information in AUV navigation process is sought the process of a preferred path from the starting point to the impact point according to certain evaluation criterion.Real-time route planning algorithm commonly used has Artificial Potential Field Method, A* or D* algorithm, genetic algorithm etc.Artificial Potential Field Method has good real-time, but has trap area and can not find the shortcomings such as path between close barrier.The Optimizing Search such as A* or D* algorithm more is applicable to solve single-object problem.Genetic algorithm is a kind of based on natural selection and the natural global optimization approach of heredity, and employing colony method is carried out the parallel search of multi thread to purpose-function space, more is applicable to the AUV real-time route and plans this class multi-objective optimization question.But traditional genetic algorithm exists precocious and slow two difficult problems of speed of convergence, and is not suitable for AUV real-time route planning requirement.

Summary of the invention

The technical problem to be solved in the present invention provides a kind of have better speed of convergence and constringent AUV real-time route planing method, when AUV Realtime collision free strategy is absorbed in endless loop, the method can guide AUV to avoid obstacle, jump out endless loop, and continues to carry out the mission task.

The technical scheme that the present invention adopts for achieving the above object is: a kind of immune genetic algorithm for the planning of AUV real-time route, may further comprise the steps,

1) counts out according to the AUV path and set the Small Population number, the scale of initialization Small Population, maximum evolutionary generation and generate at random the individuality of Small Population;

2) each Small Population is carried out Immune Selection after, each Small Population obtains two subgroups; Genetic manipulation is carried out in one of them subgroup, and another carries out cell clone; Vaccine inoculation and antibody cluster are carried out in two subgroups that obtain, form Small Population of future generation;

3) judge whether Small Population of future generation satisfies maximum evolutionary generations or Pareto optimal solution conditions; If satisfy, then select the optimum individual of these Small Populations according to the value of affinity; If do not satisfy, then return step 2);

4) from the set that each Small Population optimum individual forms, select of affinity value maximum as optimum individual according to the affinity of each optimum individual, this optimum individual is the path of planning.

Described Immune Selection is specially: select the high individuality of affinity, and individual number satisfies setting value; The value of affinity is passed through formula $A\left(P_{\mathrm{mk}}\right)=\frac{1}{1+|J\left(P_{\mathrm{mk}}\right)-\underset{i=1}{\overset{N_{\mathrm{bm}}}{\min }}J\left(B_{\mathrm{mi}}\right)|}$ Obtain;

Wherein, adaptive value

$J\left(P_{\mathrm{mk}}\right)=\frac{\underset{l=0}{\overset{m}{\mathrm{\Σ}}}d(p_{k,l},p_{k,l+1})-d(p_s,p_e)}{d(p_s,p_e)}+\underset{l=1}{\overset{m}{\mathrm{\Σ}}}\frac{{\mathrm{\ψ}}_{k,l}}{\mathrm{\π}}$

$+\underset{l=o}{\overset{m}{\mathrm{\Σ}}}\underset{g}{\overset{G}{\max }}b\left(p_{k,l}{p}_{k,l+1}{\mathrm{\Ω}}_g\right)+\underset{1=0}{\overset{m}{\mathrm{\Σ}}}g(p_{k,l}{p}_{k,l+1},T_{\mathrm{AUV}})$

p _s, p _eThe starting point and the terminal point that represent respectively the path, p _{K, 0}=p _s, p _{K, m+1}=p _eD (p _x, p _y) be distance function, the space length between representing at 2; Ψ _KlExpression line segment p _{K, l-1}p _{K, l}With line segment p _{K, l}p _{K, l+}The angle of 1 extended line, Ψ _Kl∈ [0, π]; P _MkBe the P of colony _m(t) antibody in, B _MiBe the P of colony _m(t) antigen in;

B (p _{K, l}p _{K, l+1}, Ω _g) expression line segment p _{K, l}p _{K, l+1}With obstacle Ω _gThe coefficient of relative orientation; d _MinThe minimum safe distance of expression AUV and obstacle, O _dThe penalty coefficient that expression line segment and obstacle intersect, d (p _{K, l}p _{K, l+1}, Ω _g) expression line segment p _{K, l}p _{K, l+1}With obstacle Ω _gMinimum distance; B (p _{K, l}p _{K, l+1}, Ω _g) computing formula is:

$b(p_{k,l}{p}_{k,l+1},{\mathrm{\Ω}}_g)=\left\{\begin{array}{cc}0& \mathrm{ifd}(p_{k,l}{p}_{k,l+1},{\mathrm{\Ω}}_g)\≥d_{\min }\\ O_d& \mathrm{otherwise}\end{array}\right.$

G (p _{K, l}p _{K, l+1}, T _AUV) expression path p _{K, l}p _{K, l+1}Whether with AUV ship trajectory T _AUVIntersect, if path and AUV ship trajectory intersect, then corresponding chromosome is punished that penalty value is larger positive number O _g:

Described genetic manipulation may further comprise the steps:

From the current P of colony _m(t) in according to selecting probability

Select the individuality of some, generate the mating pond, T＞0th wherein, annealing temperature, J _Mk=J (P _Mk), N _mBe the P of colony _m(t) scale;

From the mating pond, select two chromosomes to carry out interlace operation, be specially: two chromosome is separating from same position at random all, a chromosomal first half and another chromosomal latter half combination, another chromosomal first half and previous chromosomal latter half combination enter new subgroup thereby form two brand-new individualities;

Again the individuality in the mating pond is carried out mutation operation, at the evolution initial stage, adopt unified modes of reproduction, also claim asexual intersection: select a chromosome, the one or more chromosomal genes of randomly changing; After converging to a certain degree, use non-unified breeding instead, also claim heuristic intersection: the gene location that the chosen distance obstacle is nearer, produce sudden change along vertical ideal path direction by length resolution, the individuality that obtains enters new subgroup.

Described cell clone is specially: according to antibody X _iWith antigen Y _jSuper sudden change formula X _i← X _i+ β (Y _j-X _i) carry out antibody breeding; β ∈ [0, α] wherein,

Described antibody cluster may further comprise the steps: with given colony

Be divided into q subgroup Q _k: $Q_k=\{P_{k1},{\mathrm{<figure-callout\; id="2"\; label="P\; k"\; filenames="HDA00002456710400022.png"\; state="\{\{state\}\}">P</figure-callout>}}_k,\&CenterDot;\&CenterDot;\&CenterDot;P_{k{p}_k}\},$ ${\&ForAll;P}_{\mathrm{ki}},P_{\mathrm{kj}}\&Element;Q_k,$ M(P _ki,P _kj)≤δ；

Similarity $M(P_{\mathrm{mk}1},P_{\mathrm{mk}2})=\underset{i=0}{\overset{m}{\mathrm{\&Sigma;}}}d(p_{k1,i},{\mathrm{<figure-callout\; id="2"\; label="p\; k"\; filenames="HDA00002456710400022.png"\; state="\{\{state\}\}">p</figure-callout>}}_k),$ δ is self-defined constant;

According to | F (P _Ki)-F (P _Kj) |≤δ ₀, P _Ki, P _Kj∈ Q _kThe low individual P of punishment excitation degree _Kj, the individuality of not being punished is deposited among the Q;

Wherein, excitation degree

Wherein β is regulatory factor, β 〉=1; $C\left(P_{\mathrm{mk}}\right)=\frac{\left|\right\{P_{\mathrm{mi}}\&Element;X|M(P_{\mathrm{mk}},P_{\mathrm{mi}})\&le;\mathrm{\&delta;}\}|}{N_{\mathrm{pm}}},$ $X\&Subset;P_m$ Be P _MkAffiliated antibody population, N _PmScale for Small Population.

The present invention has the following advantages:

1. the present invention utilizes antibody population cluster mechanism to keep the diversity of colony, has both avoided algorithm Premature Convergence problem, is conducive to again reach global optimization.The immune genetic algorithm of setting up adopts self-regulatory mechanism that cluster analysis is carried out in the filial generation that generates, and has guaranteed the diversity of colony.

2. the local search ability of genetic algorithm has been strengthened in the cell clone operation of the present invention's foundation.Affine sudden change and even mutation strategy are more emphasized the Local Search to current subspace, have also strengthened local search ability when guaranteeing the genetic algorithm ability of searching optimum.

3. the present invention utilizes immunological memory mechanism, keeps the efficient feasible antibody that occurs in the computation process as vaccine, thereby and be population vaccine inoculation quickening convergence of algorithm speed in good time.

Description of drawings

Fig. 1 is the double-deck AUV Realtime collision free system construction drawing that the present invention sets up;

Fig. 2 is the immune genetic algorithm process flow diagram that the present invention sets up;

Fig. 3 is immune genetic algorithm and the contrast that makes two kinds of method path plannings and speed of convergence among the embodiment one;

Fig. 4 is the AUV ship trajectory that has or not among the embodiment two in the real-time route planning situation;

Fig. 5 is the result of twice real-time route planning among the embodiment two.

Embodiment

The present invention is described in further detail below in conjunction with drawings and Examples.

As shown in Figure 1, the present invention increases the real-time route planning module in original autonomous underwater robot Realtime collision free system framework, set up double-deck Realtime collision free system framework.The real-time route planning module is triggered by the monitoring planning module, plans the expected path that makes new advances according to Online Map and AUV current state that data processing module generates.

The real-time route planning module that the present invention sets up adopts the real-time route planning technology scheme based on immune genetic algorithm, as shown in Figure 2, the core is comprised of Immune Selection, genetic manipulation, cell clone and 4 modules of antibody cluster, and main flow process is: at first from treating Advanced group species P _m(t) determinacy ground selects the strong antibody of responsibility to form the subgroup in

Carry out immune response, residue is individual to be formed

Right

In individuality carry out cell clone operation, right

In individuality carry out genetic manipulation, the result of two operations forms

Carry out the antibody cluster, remove same or analogous antibody.At last from

The middle individuality of optimum of selecting forms the P of colony of future generation _m(t+1).Treat that Advanced group species represents mulitpath, individuality represents the path, counts out according to the AUV path and sets the Small Population number.

(1) Immune Selection

The purpose of Immune Selection is by selecting Probability p _oFrom P _m(t) the higher individuality of selective affinity forms the P of colony to be evolved in _m ¹(t), wherein individual number satisfies setting value (the present embodiment setting value is 100);So-called affinity refers to the matching degree of antibody and antigen, is representing the matching degree of individuality to be selected and current optimum individual.The affinity that the present invention sets up is the size of weighing affinity with adaptive value difference, at first finds the antigen of adaptive value minimum, individual P _MkAdaptive value and the adaptive value of this antigen do difference, individual P _MkAffinity and this difference be inversely proportional to.Individual P _MkExpression antibody, B _MiExpression antigen (can be set as tens), when affinity is maximum, P _MkAnd B _MiMate most.The computing formula of affinity is:

$A\left(P_{\mathrm{mk}}\right)=\frac{1}{1+|J\left(P_{\mathrm{mk}}\right)-\underset{i=1}{\overset{N_{\mathrm{bm}}}{\min }}J\left(B_{\mathrm{mi}}\right)|}---\left(1\right)$

J (P in the formula (1) _Mk) be individual P _MkFitness function, this function representation path, security and smoothness more can be reacted the demand of AUV real-time route planning.

Computing formula is:

$J\left(P_{\mathrm{mk}}\right)=\frac{\underset{l=0}{\overset{m}{\mathrm{\&Sigma;}}}d(p_{k,l},p_{k,l+1})-d(p_s,p_e)}{d(p_s,p_e)}+\underset{l=1}{\overset{m}{\mathrm{\&Sigma;}}}\frac{{\mathrm{\&psi;}}_{k,l}}{\mathrm{\&pi;}}---\left(2\right)$

$+\underset{l=o}{\overset{m}{\mathrm{\&Sigma;}}}\underset{g}{\overset{G}{\max }}b\left(p_{k,l}{p}_{k,l+1}{\mathrm{\&Omega;}}_g\right)+\underset{1=0}{\overset{m}{\mathrm{\&Sigma;}}}g(p_{k,l}{p}_{k,l+1},T_{\mathrm{AUV}})$

In the formula (2) $\frac{\underset{l=0}{\overset{m}{\mathrm{\&Sigma;}}}d(p_{k,l},p_{k,l+1})-d(p_s,p_e)}{d(p_d,p_e)}$ The length adaptability in expression path, $\underset{l=1}{\overset{m}{\mathrm{\&Sigma;}}}\frac{{\mathrm{\&psi;}}_{k,l}}{\mathrm{\&pi;}}$ The adaptability of expression path smooth degree, $\underset{l=o}{\overset{m}{\mathrm{\&Sigma;}}}\underset{g}{\overset{T}{\max }}b(p_{k,l}{p}_{k,l+1},{\mathrm{\&Omega;}}_g)$ Expression path security yardstick, $\underset{1=0}{\overset{m}{\mathrm{\&Sigma;}}}g(p_{k,l}{p}_{k,l+1},T_{\mathrm{AUV}})$ Whether intersect with the AUV ship trajectory in the expression path; Whether above-mentioned length adaptability, path smooth degree adaptability, path security yardstick, path can calculate according to online electronic chart and AUV current state (being sensing data) with data such as the AUV ship trajectory intersect.

p _s, p _eThe starting point and the terminal point that represent respectively the path, and definition p _{K, 0}=p _s, p _{K, m+1}=p _eD (p _x, p _y) be distance function, the space length between representing at 2.Ψ _KlExpression line segment p _{K, l-1}p _{K, l}With line segment p _{K, l}p _{K, l+1}The angle of extended line, Ψ _Kl[0, π].

B (p _{K, l}p _{K, l+1}, Ω _g) expression line segment p _{K, l}p _{K, l+1}With obstacle Ω _gThe coefficient of relative orientation.If d _MinThe minimum safe distance of expression AUV and obstacle, O _dThe penalty coefficient (usually being set as larger positive integer) that expression line segment and obstacle intersect, d (p _{K, l}p _{K, l+1,}Ω _g) expression line segment p _{K, l}p _{K, l+1}With obstacle Ω _gMinimum distance.B (p then _{K, l}p _{K, l+1}, Ω _g) computing formula is:

$b(p_{k,l}{p}_{k,l+1},{\mathrm{\&Omega;}}_g)=\left\{\begin{array}{cc}0& \mathrm{ifd}(p_{k,l}{p}_{k,l+1},{\mathrm{\&Omega;}}_g)\&GreaterEqual;d_{\min }\\ O_d& \mathrm{otherwise}\end{array}\right.---\left(3\right)$

G (p _{K, l}p _{K, l+1}, T _AUV) expression path p _{K, l}p _{K, l+1}Whether with AUV ship trajectory T _AUVIntersect, if path and AUV ship trajectory intersect, then corresponding chromosome is punished that penalty value is larger positive number O _g:

Determinacy ground is selected to reply the strong antibody of antigenic capacity and is carried out immune response, participates in cell clone and affine sudden change.Affine sudden change can be finely tuned the path point position near obstacle, and cell clone operation acting in conjunction, strengthens near the local search ability of barrier zone.

Immune Selection provides more more options chance on the one hand the high antibody of affinity, and provides chance for survival to affinity and all low antibody of concentration, so that the antibody population of survival has diversity.

(2) genetic manipulation

The genetic manipulation module that the present invention sets up is comprised of three genetic operators: select operator, special crossover operator and mutation operator.

Selecting operator is from the current P of colony _m(t) select the individuality of some in, generate the mating pond.If T＞0th, annealing temperature, select probability to be:

$S_s\left(P_{\mathrm{mk}}\right)=\frac{e^{\frac{J_{\mathrm{mk}}}{T}}}{\underset{k=1}{\overset{N_m}{\mathrm{\&Sigma;}}}{e}^{\frac{J_{\mathrm{mk}}}{t}}}---\left(5\right)$

Wherein, J _Mk=J (P _Mk), N _mBe the P of colony _m(t) scale.

Crossover operator is to select two chromosomes from the mating pond, equal separating from same position at random, a chromosomal first half and another chromosomal latter half combination, another chromosomal first half and previous chromosomal latter half combination, thus two brand-new individualities formed.

Mutation operator is divided into two kinds, at the evolution initial stage, adopts unified modes of reproduction, also claims asexual intersection: select a chromosome, the one or more chromosomal genes of randomly changing.After converging to a certain degree, use non-unified breeding instead, also claim heuristic intersection: the gene location that the chosen distance obstacle is nearer produces sudden change along vertical ideal path direction by length resolution.

(3) cell clone

Cell clone refers under given breeding number, all antibody breeding clones' mapping in the antibody population.Set X, Y is respectively given antibody population and antigen group, antibody

Antigen

The antibody X that the present invention sets up _iThe breeding number

Computing formula is:

$N_{x_i}={(N_{\mathrm{bm}}-\frac{1}{\mathrm{\&lambda;A}\left(X_i\right)})}^{\mathrm{\&theta;}},$ $\mathrm{\&lambda;}\&Element;[\frac{1}{2(1+A\left(X_i\right))},\frac{1}{(1+A\left(X_i\right))}]$

Wherein λ is random number, expression antibody X _iBreeding potential; 1.0＜θ＜1.5 are setup parameter.

The cell clone process that the present invention sets up is antibody population P _m ¹(t) each antibody is counted formula breeding clone, the N that then breeds according to above-mentioned breeding in _BmIndividual clone and antigen group B _m(t) antigen in surpasses sudden change.To remaining clone, select at random B _m(t) antigen in carries out even random mutation.

The antibody X that the present invention sets up _iWith antigen Y _jSuper sudden change formula:

X _i←X _i+β(Y _j-X _i)，β∈[0,α]， $\mathrm{\&alpha;}=1-e^{-\left|\right|x_i-Y_j\left|\right|}---\left(6\right)$

Wherein β is the random number on [0, α].

Antibody X _iWith antigen Y _jEven random mutation refer to antibody X _iWith mutation rate α ₀As probability to the random mutation between 0 to 9 integer of the gene on its each gene location, λ bit constant, wherein α ₀Determined by following formula.

${\mathrm{\&alpha;}}_0=1-{\mathrm{\&lambda;e}}^{-\left|\right|x_i-Y_j\left|\right|}---\left(7\right)$

(4) antibody cluster

The present invention introduces clustering algorithm and processes individuality superfluous in the antibody population.Given colony

P is divided into q subgroup Q _k

$Q_k=\{P_{k1},{\mathrm{<figure-callout\; id="2"\; label="P\; k"\; filenames="HDA00002456710400022.png"\; state="\{\{state\}\}">P</figure-callout>}}_k,\&CenterDot;\&CenterDot;\&CenterDot;P_{k{p}_k}\},$ ${\&ForAll;P}_{\mathrm{ki}},P_{\mathrm{kj}}\&Element;Q_k,$ M(P _ki,P _kj)≤δ (8)

According to

|F(P _ki)-F(P _kj)|≤δ ₀,P _ki,P _kj∈Q _k (9)

Punishment P _Ki, P _KjThe low individuality of middle excitation degree deposits the individuality of not being punished among the Q in.

Wherein degree of excitation refers to antibody response antigen in the antibody population and the integration capability that is activated by other antibody, and the present invention is defined as function F with it:

Computing formula is:

$F\left(P_{\mathrm{mk}}\right)=A\left(P_{\mathrm{mk}}\right)e^{-\frac{C\left(P_{\mathrm{mk}}\right)}{\mathrm{\&beta;}}}---\left(10\right)$

β is regulatory factor in the formula (10), β 〉=1,

Be P _MkAffiliated antibody population.Antibody concentration C (P _Mk) computing formula be:

$C\left(P_{\mathrm{mk}}\right)=\frac{\left|\right\{P_{\mathrm{mi}}\&Element;X|M(P_{\mathrm{mk}},P_{\mathrm{mi}})\&le;\mathrm{\&delta;}\}|}{N_{\mathrm{pm}}}$

M (P in the formula _Mk1, P _Mk2) be similarity, its value is larger, and similarity is less; Computing formula is:

$M(P_{\mathrm{mk}1},P_{\mathrm{mk}2})=\underset{i=0}{\overset{m}{\mathrm{\&Sigma;}}}d(p_{k1,i},{\mathrm{<figure-callout\; id="2"\; label="p\; k"\; filenames="HDA00002456710400022.png"\; state="\{\{state\}\}">p</figure-callout>}}_k)$

Clustering algorithm impels in the antibody population same or analogous being determined property of antibody ground to remove, and its effect not only is to keep population diversity, and alleviates selection pressure for the Immune Selection operator antibody of selecting to survive.

As shown in Figure 2, the present invention forms a kind of new immune genetic algorithm that is used for the planning of AUV real-time route take genetic algorithm as immunologic mechanism main, that introduce antibody recognition antigen, and its basic procedure is:

Step 1, if the line of path starting point and impact point satisfies formula (2), then with starting point and impact point as antibody Forwarded for the 9th step to; Otherwise, count out according to the path and require to set Small Population number M, common 2≤M≤100; If the set of vaccine is the memory cell group, set up another the line of initial point and impact point as memory cell group's initial value M ₁(t), make m=2;

Step 2 is set Small Population P _m(t) scale N _Pm, maximum evolutionary generation T _m, make t=1;

Step 3 take the memory cell group as the basis, generates N at random _PmThe individual Small Population P that forms _m(t);

Step 4 is calculated P according to formula (2) _m(t) each individual adaptive value in;

Step 5 is from antibody population P _m(t) select the larger individuality of adaptive value to form alternative antigen group in, with original antigen group B _m(t) carry out together the antigen cluster, thus neoantigen group B more _m(t), antigen group scale is N _Bm

Step 6 selects feasible optimal path as vaccine from up-to-date antigen, adds former memory cell group M _M-1(t) form new memory cell group M _m(t);

Step 7 is calculated P _m(t) affinity of antibody in; And as measuring by selecting Probability p _oFrom P _m(t) select N in _Pm1Individual optimized individual forms the P of colony to be evolved _m ¹(t), all the other group of individuals become the P of colony _m ²(t);

Step 8 is with P _m ¹(t) carry out the cell clone operation, form a filial generation P _m ¹¹(t);

Step 9 is to P _m ²(t) individuality carries out the genetic manipulations such as crossover and mutation in, generates P _m(t) filial generation P _m ²¹(t);

Step 10 makes P _m ³(t)=[P _m ¹¹(t) P _m ²¹(t)], calculate P _m ³(t) each individual adaptive value in if the corresponding path of optimum individual is infeasible, then selects part individuality and gene position to carry out vaccine inoculation at random.

Step 11, antagonist group P _m ³(t) carry out cluster analysis, select the individuality composition filial generation that adaptive faculty is stronger in each cluster to represent P _m ⁴(t);

Step 12, when | P _m ⁴(t) | 〉=N _PmThe time, from P _m ⁴(t) select optimum N in _PmIndividual antibody forms the P of colony of future generation _m(t+1); When | P _m ⁴(t) |＜N _PmThe time, the at random new individuality of generating portion and P _m ⁴(t) form together the P of colony of future generation _m(t+1);

Step 13 is calculated P _m(t+1) each individual adaptive value in;

Step 14, t=t+1; If t＞T _mOr satisfy the Pareto optimal solution conditions, export the optimum individual in the current colony

Forwarded for the 15th step to; If t≤T _mAnd do not satisfy the Pareto optimal solution conditions, forwarded for the 5th step to;

Step 15, m=m+1; If m＞M or the satisfied cycling condition that stops forwarded for the 16th step to; If m≤M and the satisfied cycling condition that stops returned for the 2nd step;

Step 16 is from the set of each Small Population optimum individual composition In, select a peaked optimum individual of affinity that obtains according to formula (1), generate preferred path.Preserve vaccine information Mm (t).

Fig. 3 is embodiment one, establishes the current point of AUV and is (24,10), and impact point is (24,50).If m=2 adopts respectively immune genetic algorithm (IGA), immune algorithm (IA) and genetic algorithm (GA) to carry out path planning, three kinds of algorithms are all used identical fitness function.Get the optimum individual in the 100th generation as program results, then Fig. 3 a is the route result of three kinds of algorithmic rules.Obviously, the path of immune genetic algorithm planning is shorter, more excellent.The adaptive value of getting every generation optimum individual represents speed of convergence, and shown in Fig. 3 b, three kinds of algorithms all are in convergence state when 100 generation, but convergency value is different.The adaptive value of immune genetic algorithm the 100th generation optimum individual is 0.36967, and the adaptive value of genetic algorithm the 100th generation optimum individual is 1.02796, and the adaptive value of immune algorithm the 100th generation optimum individual is 0.41559.

Fig. 4 and Fig. 5 are embodiment two, the expected path of AUV is AB, Fig. 4 a is the ship trajectory of AUV when not adopting the real-time route planning strategy, obviously AUV can't pass through semi-enclosed barrier zone, Fig. 4 b carries out the ship trajectory of the rear AUV of real-time route planning for the immune genetic algorithm that adds the present invention's foundation, this shows that the immune genetic real-time route planing method that the present invention sets up provides the means of escaping the trap area such as semiclosed for AUV.

Carried out altogether 4 real-time route planning at the 643.5th second, 1234.5 seconds, 724 seconds and 1347 seconds among the embodiment two, the result of real-time route planning when Fig. 5 a is the 643.5th second, the result of real-time route planning when Fig. 5 b is the 1347th second.

Claims (5)

1. an immune genetic algorithm that is used for the planning of AUV real-time route is characterized in that may further comprise the steps,

1) counts out according to the AUV path and set the Small Population number, the scale of initialization Small Population, maximum evolutionary generation and generate at random the individuality of Small Population;

2) each Small Population is carried out Immune Selection after, each Small Population obtains two subgroups; Genetic manipulation is carried out in one of them subgroup, and another carries out cell clone; Vaccine inoculation and antibody cluster are carried out in two subgroups that obtain, form Small Population of future generation;

3) judge whether Small Population of future generation satisfies maximum evolutionary generations or Pareto optimal solution conditions; If satisfy, then select the optimum individual of these Small Populations according to the value of affinity; If do not satisfy, then return step 2);

4) from the set that each Small Population optimum individual forms, select of affinity value maximum as optimum individual according to the affinity of each optimum individual, this optimum individual is the path of planning.

2. a kind of immune genetic algorithm for the planning of AUV real-time route according to claim 1 is characterized in that described Immune Selection is specially: select the high individuality of affinity, and individual number satisfies setting value; The value of affinity is passed through formula $A\left(P_{\mathrm{mk}}\right)=\frac{1}{1+|J\left(P_{\mathrm{mk}}\right)-\underset{i=1}{\overset{N_{\mathrm{bm}}}{\min }}J\left(B_{\mathrm{mi}}\right)|}$ Obtain;

Wherein, adaptive value

$J\left(P_{\mathrm{mk}}\right)=\frac{\underset{l=0}{\overset{m}{\mathrm{\&Sigma;}}}d(p_{k,l},p_{k,l+1})-d(p_s,p_e)}{d(p_s,p_e)}+\underset{l=1}{\overset{m}{\mathrm{\&Sigma;}}}\frac{{\mathrm{\&psi;}}_{k,l}}{\mathrm{\&pi;}}$

$+\underset{l=o}{\overset{m}{\mathrm{\&Sigma;}}}\underset{g}{\overset{G}{\max }}b\left(p_{k,l}{p}_{k,l+1}{\mathrm{\&Omega;}}_g\right)+\underset{1=0}{\overset{m}{\mathrm{\&Sigma;}}}g(p_{k,l}{p}_{k,l+1},T_{\mathrm{AUV}})$

p _s, p _eThe starting point and the terminal point that represent respectively the path, p _{K, 0}=p _s, p _{K, m+1}=p _eD (p _x, p _y) be distance function, the space length between representing at 2; Ψ _KlExpression line segment p _{K, l-1}p _{K, l}With line segment p _{K, l}p _{K, l+1}The angle of extended line, Ψ _Kl∈ [0, π]; P _MkBe the P of colony _m(t) antibody in, B _MiBe the P of colony _m(t) antigen in;

B (p _{K, l}p _{K, l+1}, Ω _g) expression line segment p _{K, l}p _{K, l+1}With obstacle Ω _gThe coefficient of relative orientation; d _MinThe minimum safe distance of expression AUV and obstacle, O _dThe penalty coefficient that expression line segment and obstacle intersect, d (p _{K, l}p _{K, l+1}, Ω _g) expression line segment p _{K, l}p _{K, l+1}With obstacle Ω _gMinimum distance; B (p _{K, l}p _{K, l+1}, Ω _g) computing formula is:

$b(p_{k,l}{p}_{k,l+1},{\mathrm{\&Omega;}}_g)=\left\{\begin{array}{cc}0& \mathrm{ifd}(p_{k,l}{p}_{k,l+1},{\mathrm{\&Omega;}}_g)\&GreaterEqual;d_{\min }\\ O_d& \mathrm{otherwise}\end{array}\right.$

G (p _{K, l}p _{K, l+1}, T _AUV) expression path p _{K, l}p _{K, l+1}Whether with AUV ship trajectory T _AUVIntersect, if path and AUV ship trajectory intersect, then corresponding chromosome is punished that penalty value is larger positive number O _g:

3. a kind of immune genetic algorithm for AUV real-time route planning according to claim 1 is characterized in that described genetic manipulation may further comprise the steps:

From the current P of colony _m(t) in according to selecting probability Select the individuality of some, generate the mating pond, T＞0th wherein, annealing temperature, J _Mk=J (P _Mk), N _mBe the P of colony _m(t) scale;

From the mating pond, select two chromosomes to carry out interlace operation, be specially: two chromosome is separating from same position at random all, a chromosomal first half and another chromosomal latter half combination, another chromosomal first half and previous chromosomal latter half combination enter new subgroup thereby form two brand-new individualities;

Again the individuality in the mating pond is carried out mutation operation, at the evolution initial stage, adopt unified modes of reproduction, also claim asexual intersection: select a chromosome, the one or more chromosomal genes of randomly changing; After converging to a certain degree, use non-unified breeding instead, also claim heuristic intersection: the gene location that the chosen distance obstacle is nearer, produce sudden change along vertical ideal path direction by length resolution, the individuality that obtains enters new subgroup.

4. a kind of immune genetic algorithm for the planning of AUV real-time route according to claim 1 is characterized in that described cell clone is specially: according to antibody X _iWith antigen Y _jSuper sudden change formula X _i← X _i+ β (Y _j-X _i) carry out antibody breeding; β ∈ [0, α] wherein,

5. a kind of immune genetic algorithm for the planning of AUV real-time route according to claim 1 is characterized in that described antibody cluster may further comprise the steps: with given colony

Be divided into q subgroup

Q _k： $Q_k=\{P_{k1},P_{k2},\&CenterDot;\&CenterDot;\&CenterDot;P_{k{p}_k}\},$ ${\&ForAll;P}_{\mathrm{ki}},P_{\mathrm{kj}}\&Element;Q_k,$ M(P _ki,P _kj)≤δ；

Similarity $M(P_{\mathrm{mk}1},P_{\mathrm{mk}2})=\underset{i=0}{\overset{m}{\mathrm{\&Sigma;}}}d(p_{k1,i},p_{k2,i}),$ δ is self-defined constant;

According to | F (P _Ki)-F (P _Kj) |≤δ ₀, P _Ki, P _Kj∈ Q _kThe low individual P of punishment excitation degree _Kj, the individuality of not being punished is deposited among the Q;

Wherein, excitation degree

Wherein β is regulatory factor, β 〉=1; $C\left(P_{\mathrm{mk}}\right)=\frac{\left|\right\{P_{\mathrm{mi}}\&Element;X|M(P_{\mathrm{mk}},P_{\mathrm{mi}})\&le;\mathrm{\&delta;}\}|}{N_{\mathrm{pm}}},$ $X\&Subset;P_m$ Be P _MkAffiliated antibody population, N _PmScale for Small Population.

Images