CN114021690B - A natural computing method based on swarm intelligence - Google Patents
A natural computing method based on swarm intelligence Download PDFInfo
- Publication number
- CN114021690B CN114021690B CN202111158377.6A CN202111158377A CN114021690B CN 114021690 B CN114021690 B CN 114021690B CN 202111158377 A CN202111158377 A CN 202111158377A CN 114021690 B CN114021690 B CN 114021690B
- Authority
- CN
- China
- Prior art keywords
- agent
- fitness
- intelligent
- gradient field
- natural
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/004—Artificial life, i.e. computing arrangements simulating life
- G06N3/006—Artificial life, i.e. computing arrangements simulating life based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/004—Artificial life, i.e. computing arrangements simulating life
- G06N3/008—Artificial life, i.e. computing arrangements simulating life based on physical entities controlled by simulated intelligence so as to replicate intelligent life forms, e.g. based on robots replicating pets or humans in their appearance or behaviour
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Molecular Biology (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Robotics (AREA)
- Magnetic Resonance Imaging Apparatus (AREA)
Abstract
The invention discloses a natural computing method based on group intelligence, which is based on a weak person priority evolution strategy in static living body computing, and provides an optimal state retention computing mechanism, wherein the mechanism respectively calculates the fitness of an agent at each sampling point by sampling for a plurality of times in the tracking period of a micro-nano robot, then selects the maximum fitness value of the same agent in the same tracking period as the final fitness of the agent in the tracking period, and can evaluate the fitness of the micro-nano robot in one tracking period for a plurality of times by using the method, thereby extracting the optimal evaluation state, overcoming the time-varying characteristic of the computation of the adaptability of the agent caused by a dynamic biological gradient field and improving the relative accuracy of the fitness evaluation. Therefore, the operation mechanism ensures the accuracy of shared information in the multi-agent collaborative search process under the in-vivo dynamic optimization environment, thereby improving the solving efficiency of the global optimal solution in the natural calculation.
Description
Technical Field
The invention belongs to the field of interdisciplinary, relates to research directions of computational intelligence, micro-nano robots, information medicine and the like, and particularly relates to a natural computing method based on group intelligence.
Background
Traditional medical imaging techniques are limited by the resolution of the imaging system and do not allow for in vivo tissue imaging at the cellular level. In recent years, with the continuous penetration of nano-technology in the medical field, the nanoparticle-based medical contrast imaging technology has also been rapidly developed. For example, existing nanoparticle systems rely on the blood circulation of the human body to power the injection of nanoparticles into the human body, and the transmission of contrast agents is achieved by continuously aggregating nanoparticles around lesions through the high permeability of blood vessels in the lesion tissue area, a process which can be viewed from a computational perspective as brute-force search (brute force search). Since it relies on blood circulation as a kinetic source of the nanoparticles and does not exert external forces to control them, it can result in the nanoparticles being spread throughout the body and not reaching the target site along the most efficient path.
Inspired by the idea that multiple agents cooperate with each other to find an optimal solution in the group intelligent calculation, the invention provides a natural calculation method based on a group magnetic micro-nano robot for improving the conveying efficiency of a contrast agent. The method relies on focus-induced biological gradient field (biological GRADIENT FIELD, BGF for short) information (such as blood flow velocity, pH value, enzyme activity, oxygen concentration, etc.) to provide guidance for the motion of the micro-nano robot, and real-time monitoring is performed through an imaging system. The robot group shares BGF information of the position of the robot group, and then the motion state of the robot group is continuously adjusted through external control equipment, so that efficient transportation of contrast agent molecules is finally realized. The BGF of living tissue is subject to fluid mobility and is subject to dynamic changes over time, requiring approximate replacement with a dynamically changing optimizable function. Based on the method, the invention provides a natural computing method for dynamic BGF, and the method can be used for controlling the group micro-nano robot in vitro to overcome physical constraints brought by in vivo tissue microenvironment, so as to realize efficient delivery of contrast agent molecules at a target position.
Disclosure of Invention
The technical problems to be solved by the invention are as follows: under the condition of being influenced by dynamic changes of biological gradient fields, the group micro-nano robots are ensured to utilize effective biological gradient field information, and the real-time planning of the motion path of the micro-nano robots is carried out by a natural calculation method, so that the improvement of the contrast agent conveying efficiency is realized.
The invention provides a natural computing method based on group intelligence, which is realized based on an external controllable micro-nano robot (hereinafter referred to as micro-nano robot) and is used for overcoming the influence of dynamic change of an in-vivo optimization environment on multi-agent fitness computing and realizing accurate solution under the in-vivo dynamic optimization environment. The research focus of the invention includes a model of natural computation and a solving strategy.
The natural computing method based on group intelligence provided by the invention comprises the following three steps:
1) Establishment of natural computing model
Inspired by the intelligent computing idea of the group, the invention provides a basic theoretical framework of natural computing, and builds a natural computing model on the basis. In the natural computing model, K agents A 1,A2,…,AK are assumed, which are used in the search spaceSearching for an optimal solution
The present invention assumes that the bio-gradient field has periodic and non-periodic variations, respectively, over time. The in vivo objective function to be optimized can be expressed as follows:
Wherein o ms(x,y,z;Ak; t) represents the externally measurable objective function of kth agent a k (k=1, 2, …, K) at position (x, y, z) at time t; o in (x, y, z; t) represents the in vivo objective function at (x, y, z) at time t, i.e., the bio-gradient field function; o fit(x,y,z;Ak; t) is to overcome A k and Is a complex of the interaction o ex(x,y,z;Ak; t) a correction term for the fitness introduced; representing the position coordinates (x, y, z) of Ak in the search space Is a kind of medium. In the natural computing model, a globally optimal solutionThe position of the tumor focus will not change during the whole calculation process, i.e. the position of the tumor focus will remain unchanged during the detection process.
2) Establishment of fitness calculation model
Because the calculation of the fitness of the intelligent agent plays a decisive role in each iterative operation effect of the algorithm in the natural calculation process, the establishment of the fitness calculation model of the intelligent agent is an important link of a natural calculation method, and the establishment of the fitness calculation model comprises the following two substeps:
a) Dynamic biological gradient field characterization
Limited by medical technology, the existing literature records only carry out qualitative experimental observation on the biological gradient field related to the invention, and no widely accepted quantification model exists. Therefore, some representative standard test functions are mainly used at the present stage to represent the overall distribution of the biological gradient field. For example, body fluids in the tumor tissue microenvironment are in a flowing state, and there is dynamic variation in the distribution of biochemical information (e.g., blood flow velocity, pH, enzymatic activity, cytokines, etc.) in the fluid over time. Thus, a static bio-gradient field cannot accurately represent a natural calculated real application scenario. In order to make more intensive research on natural computing theory, the dynamic change situation of the biological gradient field with time needs to be considered.
The invention designs a regularized Biological Gradient Field (BGF) function, wherein the function value gradually decreases along with the approach of the coordinate position to the origin in a parameter space. A spherical region with a radius of 0.5mm at the central position of the parameter space represents the position to be solved. The specific mathematical expression of the function is shown in a formula (2) and represents the static state of the BGF function. Based on the above, in order to perform simulation on the dynamic change condition of the biological gradient field, the invention obtains two corresponding time-varying models by adding a dynamic change factor into the basic function (the biological gradient field function described by the formula (2) is used as the basic function).
The invention relates to o in (x, y, z; t) the function may satisfy periodic sinusoidal variation and possibly non-periodic gaussian variation, and the two corresponding time-varying models are o in (x, y, z; t), wherein H represents the amplitude of the sinusoidal variation of the function, ω represents the frequency of the sinusoidal variation of the function; and o in (x, y, z; t) in step 1) satisfying the aperiodic gaussian variation.
B) Optimal state retention mechanism
Each motion cycle of the natural computation comprises two phases of position update and position estimation of the agent. The control process of the micro-nano robot corresponds to the directional updating stage of the position of the intelligent agent, and the control process directly influences the optimization result of each step of the biological gradient field function. The tracking process of the micro-nano robot corresponds to the position estimation stage of the intelligent agent, and the process realizes the estimation of the wide-area biological gradient field information through the study of the collected local biological gradient field information, so as to guide the control process of the next step. Therefore, the tracking precision of the micro-nano robot directly influences the control effect of the next step.
The method is influenced by the dynamic change of the biological gradient field, and the collected local biological gradient field information is in the dynamic change in the micro-nano robot tracking period, so that the evaluation of the adaptability of the intelligent body is challenged. To solve this problem, the present invention proposes an "optimal state retention" mechanism.
The specific operation steps are as follows:
i) The fitness of each agent at different times S 1,S2,…,Sm of the tracking period is calculated by sampling agent a 1,A2,…,AK of the multi-agent system multiple times during the tracking period, where m represents the total number of samples of one tracking period.
Ii) comparing the fitness calculated by the single agent at time S 1,S2,…,Sm, and selecting the maximum value as the final fitness of the agent.
Iii) And respectively selecting the final fitness A 1/S3,A2/S1,A3/Sm,…,AK/S2 of all the intelligent agents to form a final fitness sequence of the multi-intelligent-agent system in the current tracking period.
3) Multi-agent control strategy
In a group intelligent algorithm (such as PSO (particle swarm optimization algorithm), GSA (gravitation search algorithm), and the like), multiple agents realize iterative update of self positions and movement speeds according to a certain operation rule, and an agent update mode has instantaneity, namely, each agent can complete ideal position update according to own requirements at the same time. In living body calculation, under the action of an external unified control magnetic field, the position update of the intelligent body has consistency. Therefore, in each step of iterative updating, an optimal position updating mode needs to be selected, so that the overall optimal evolution of the multi-agent system is realized.
The invention uses the evolution strategy of 'weak person priority' to control the micro-nano robot group to realize the optimal update of the position of the intelligent agent in living body calculation under the action of an external unified control magnetic field. And selecting an individual corresponding to the minimum fitness in A 1/S3,A2/S1,A3/Sm,…,AK/S2 as a selected individual w 1 in the group, and using a position updating mode of w 1 as an updating mode of the whole micro-nano robot group to realize the state updating of the micro-nano robot group under the control action of an external magnetic field. According to the control strategy, selecting an individual w 2 with the smallest fitness in the next tracking period of iterative operation, and realizing further updating of the group motion condition of the micro-nano robot.
The invention provides an optimal state retention operation mechanism based on a weak priority evolution strategy in static living body calculation. The mechanism calculates the fitness of the intelligent agent of each sampling point respectively by sampling for a plurality of times in the tracking period of the micro-nano robot, then sequences the fitness of a plurality of sampling points of the same intelligent agent in the same tracking period according to the sizes, and selects the maximum value as the final fitness of the intelligent agent in the tracking period. The method can evaluate the fitness of the micro-nano robot in one motion period for multiple times, further extract the optimal evaluation state, overcome the time-varying characteristic of the intelligent agent fitness calculation caused by the dynamic biological gradient field, and improve the relative accuracy of the fitness evaluation. Therefore, the operation mechanism ensures the accuracy of shared information in the multi-agent collaborative search process under the in-vivo dynamic optimization environment, thereby improving the solving efficiency of the global optimal solution in the natural calculation.
Drawings
Fig. 1 is a BGF characterization function: (a) a static base BGF function three-dimensional cross-sectional view; (b) Sectional views of periodically varying BGFs at two different times; (c) Cross-sectional views of non-periodically varying BGFs at two different times;
FIG. 2 is a schematic diagram of an "optimal state reservation" mechanism according to the present invention;
FIG. 3 is a schematic diagram of a multi-agent control strategy of the present invention;
FIG. 4 is a diagram of an example of the natural computing application in contrast agent delivery;
FIG. 5 is a comparison of contrast agent delivery simulation results;
fig. 6 is a statistical diagram of simulation results: (a) applying statistics of a traditional brute force search method; (b) applying the statistical result of the natural computation.
Detailed Description
The invention is explained in detail below with reference to the drawings and examples, and the technical solutions of the invention are clearly described. The examples selected herein are merely illustrative of the invention and are not intended to limit the invention.
The invention provides a natural computing method based on group intelligence, which is realized based on an external controllable micro-nano robot and is used for overcoming the influence of dynamic change of an in-vivo optimization environment on multi-agent fitness computing and realizing accurate solution under the in-vivo dynamic optimization environment, wherein an agent refers to the external controllable micro-nano robot, and is hereinafter referred to as the micro-nano robot; the natural computing method specifically comprises the following steps:
s1) building a natural computing model
Constructing a natural computing model in which K agents A 1,A2,…,AK are assumed to form a multi-agent system for use in searching spaceSearching for a globally optimal solution
Assuming periodic and aperiodic changes in the bio-gradient field over time, respectively, the in vivo objective function to be optimized can be expressed as follows:
Wherein o ms(x,y,z;Ak; t) represents an externally measurable objective function of kth agent a k at position (x, y, z) at time t; o in (x, y, z; t) represents the in vivo objective function at (x, y, z) at time t, i.e., the dynamic bio-gradient field function; o fit(x,y,z;Ak; t) is to overcome A k and Is a complex of the interaction o ex(x,y,z;Ak; t) a correction term for the fitness introduced; Representing the position coordinates (x, y, z) where A k is located, always in the search space Wherein k=1, 2, …, K;
In the natural computing model, a globally optimal solution The position of (2) will not change during the whole calculation process;
s2) establishment of fitness calculation model
Because the calculation of the fitness of the intelligent agent plays a decisive role in each iteration operation effect in the natural calculation process, the establishment of the fitness calculation model of the intelligent agent is an important link of a natural calculation method, and the establishment of the fitness calculation model comprises the following two steps:
s21) dynamic biological gradient field characterization
As shown in fig. 1 (a), designing a regularized biological gradient field function, wherein the function value of the function gradually decreases as the coordinate position approaches to the origin in a parameter space, and a spherical region with the radius of 0.5mm is used as the center of a sphere and represents the position to be solved; the mathematical expression of the regularized biological gradient field function f (x, y, z) is shown in a formula (2), and the three-dimensional image of the function is shown in fig. 1 (a) and represents a static biological gradient field function; adding dynamic change factors into the regularized biological gradient field function to obtain two corresponding time-varying models, wherein the two corresponding time-varying models are respectively states of a dynamic biological gradient field function o in (x, y, z; t) meeting periodic sinusoidal change at two different moments and states of a dynamic biological gradient field function o in (x, y, z; t) meeting aperiodic Gaussian change at two different moments as shown in fig. 1 (b) and (c), and the formula (2) is expressed as follows:
S22) "optimal state reservation" mechanism
Under the influence of the dynamic change of the biological gradient field, in the tracking period of the intelligent agent, the collected local biological gradient field information is in the dynamic change, so that the evaluation of the adaptability of the intelligent agent is challenged, and in order to solve the problem, an optimal state retaining mechanism is provided, as shown in fig. 2, the optimal state retaining mechanism specifically comprises the following operation steps:
i) Calculating the fitness of each intelligent agent at different moments S 1,S2,…,Sm of the tracking period respectively by sampling the intelligent agent A 1,A2,…,AK in the multi-intelligent agent system for a plurality of times in the current tracking period, wherein m represents the total sampling times of one tracking period;
ii) comparing the fitness calculated by the single agent at time S 1,S2,…,Sm, respectively, and selecting the maximum value as the final fitness of the agent;
iii) As shown in fig. 2, the final fitness of all agents is selected separately to form the final fitness sequence of the multi-agent system during the current tracking cycle.
S3) constructing a multi-agent control strategy
In living body calculation, under the action of an external unified control magnetic field, the position update of the intelligent body has consistency, and in each step of iterative update, an optimal position update mode is required to be selected to realize the overall optimal evolution of the multi-intelligent body system, so that a weak person priority evolution strategy is adopted to control the multi-intelligent body to realize the optimal update of the intelligent body position in the living body calculation under the action of the external unified control magnetic field: as shown in fig. 3, selecting an agent corresponding to the minimum fitness in the final fitness sequence of the current tracking period as a selected individual w 1 in the multi-agent system, and using a position update mode of w 1 as an update mode of the whole multi-agent system to realize state update of the multi-agent system under the action of an external unified control magnetic field; according to the control strategy, the individual w 2 with the minimum fitness in the final fitness sequence is selected in the next tracking period of iterative operation, so that the motion condition of the multi-intelligent system is further updated.
The invention is mainly applied to multi-agent-based contrast agent or drug targeted delivery. The main objective is to realize the accurate targeting of the multi-agent system to the target position in the dynamic biological gradient field environment. The method comprises the steps of firstly sampling the intelligent agent in the tracking period for a plurality of times, respectively calculating the adaptability of different sampling points, then utilizing an optimal state retention mechanism to realize the optimal evaluation of the adaptability of the intelligent agent in the tracking period, overcoming the influence of the dynamic change characteristic of the biological gradient field, and finally utilizing a weak person priority evolution strategy to finish the optimal realization of the motion control of multiple intelligent agents. The following is the implementation step of the method of the present invention applied to a specific scene, where the swarm intelligence algorithm selects the PSO algorithm.
The invention provides a natural computing method based on population intelligence, which is applied to detecting the position of a tumor focus in multi-agent-based contrast agent or drug targeting delivery and specifically comprises the following steps:
a. Setting the specific number of the intelligent agents in the multi-intelligent-agent system and the initial position of each intelligent agent in the search space;
b. calculating the initial fitness of the intelligent agent according to the position distribution condition of the intelligent agent in the search space;
c. taking PSO algorithm as basic algorithm as an example, calculating the position of the intelligent agent with' history optimal And "globally optimal" agent locations
D. And (3) sampling the randomly moving agents for a plurality of times in the tracking period, and calculating the final fitness of each agent in each iterative calculation by using an optimal state retention mechanism.
E. the 'weak priority' evolution strategy is used for realizing the directional motion control of the multi-agent system in a control period (the control period and the tracking period form the motion period of the agent).
F. Evaluating fitness and life of each agent, estimating relative distance between each agent and tumor focus, and removing an agent from the multi-agent system when the life of an agent is terminated; when a certain agent reaches the tumor focus position, the agent stops moving;
g. When the maximum running time of the algorithm is reached, calculating the final position distribution condition of the multi-agent system in the search space, and obtaining the space position of the tumor focus and the detection efficiency of the agent on the position of the tumor focus.
Directional motion formula of the agent:
wherein c 1,c2 is two different learning factors, typically any number within the values 0, 4; r 1,r2 represents a random number between (0, 1), respectively; A blood flow rate at time (t+1); representing the velocity of agent a j (j=1, 2, …, K) at time t, as selected by the "weak first" evolution strategy; And The position coordinates of any agent a i (i=1, 2, …, K) at time t and time t+1 are shown, respectively; And The movement speeds of agent a i (i=1, 2, …, K) at time t and time t+1, respectively; d 0 represents the movement speed of the agent under the individual control of the external driving field; The motion speed of A i at the time t+1, which is obtained by applying an optimal state retention mechanism and a weak priority evolution strategy, is represented; and 2 denotes a 2-norm.
The algorithm flow of the above method (simply nat. Comp.) is as follows:
As shown in fig. 4, the natural computing method provided by the invention is applied to the contrast agent delivery problem of tumor medical imaging, so as to improve the delivery efficiency. The number of the intelligent agents (i.e. micro-nano robots) adopted in the example is 20, and the iterative operation process of natural calculation can be completed through the micro-nano robot control and tracking equipment by adopting the algorithm flow. Fig. 5 shows simulation results of targeted delivery of a target region (tumor) using a conventional brute force search (B-F) method and using a natural computing (nat. Comp.) method as proposed by the present invention. It can be seen that the motion trail of the agent adopting the natural computing method can pass through the target area, and most of the agent finally stays at the target area. However, the motion trail of the agent using the brute force search method fails to pass through the target area, and finally most of the agents fail to stay at the target area. It can be seen that the natural computing method proposed by the present invention performs better in this example than the conventional method. FIG. 6 shows a statistical plot of simulation results after 1000 runs of the simulation example. Wherein the horizontal axis represents the number of agents, the vertical axis represents the number of simulation realizations, and η represents the contrast agent delivery efficiency (the ratio of the number of agents staying at the target location to the total amount of agents). It can be seen that the delivery efficiency of the natural computing method is 31.1% (as shown in fig. 6 (b)), which is far higher than that of the conventional brute force searching method with the efficiency of 11.9% (as shown in fig. 6 (a)). Thereby proving the effectiveness of the method proposed by the present invention.
The embodiments described above are only some, but not all, embodiments of the invention. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
Claims (3)
1. The natural computing method based on the group intelligence is characterized by overcoming the influence of dynamic change of an in-vivo optimization environment on multi-agent fitness computing and realizing accurate solution under the in-vivo dynamic optimization environment, wherein the agent refers to an external controllable micro-nano robot; the natural computing method specifically comprises the following steps:
s1) building a natural computing model
Constructing a natural computing model in which K agents A 1,A2,…,AK are assumed to form a multi-agent system for use in searching spaceSearching for a globally optimal solution
Assuming periodic and aperiodic changes in the bio-gradient field over time, respectively, the in vivo objective function to be optimized can be expressed as follows:
Wherein, Representing an externally measurable objective function of kth agent a k at position (x, y, z) at time t; o in (x, y, z; t) represents the in vivo objective function at (x, y, z) at time t, i.e., the dynamic bio-gradient field function; is to overcome A k and Is to interact with each otherAnd a correction term for the fitness introduced; Representing the position coordinates (x, y, z) where A k is located, always in the search space Wherein k=1, 2, …, K;
In the natural computing model, a globally optimal solution The position of (2) will not change during the whole calculation process;
s2) establishment of fitness calculation model
Because the calculation of the fitness of the intelligent agent plays a decisive role in each iteration operation effect in the natural calculation process, the establishment of the fitness calculation model of the intelligent agent is an important link of a natural calculation method, and the establishment of the fitness calculation model comprises the following two steps:
s21) dynamic biological gradient field characterization
Designing a regularized biological gradient field function, wherein the function value of the function gradually decreases along with the approach of the coordinate position to the origin in a parameter space, and a spherical region with the radius of 0.5mm is used as the center of a sphere to represent the position to be solved; the mathematical expression of the regularized biological gradient field function f (x, y, z) is shown in a formula (2) and represents a static biological gradient field function; adding dynamic change factors into the regularized biological gradient field function to obtain two corresponding time-varying models, namely a dynamic biological gradient field function o in (x, y, z; t) meeting periodic sinusoidal change and a dynamic biological gradient field function o in (x, y, z; t) meeting aperiodic Gaussian change, wherein the formula (2) is expressed as follows:
S22) "optimal state reservation" mechanism
Under the influence of the dynamic change of the biological gradient field, the collected local biological gradient field information is in the dynamic change in the intelligent agent tracking period, so that the evaluation of the intelligent agent adaptability is challenged, and in order to solve the problem, an optimal state retaining mechanism is provided, wherein the optimal state retaining mechanism specifically comprises the following operation steps:
i) Calculating the fitness of each intelligent agent at different moments S 1,S2,…,Sm of the tracking period respectively by sampling the intelligent agent A 1,A2,…,AK in the multi-intelligent agent system for a plurality of times in the current tracking period, wherein m represents the total sampling times of one tracking period;
ii) comparing the fitness calculated by the single agent at time S 1,S2,…,Sm, respectively, and selecting the maximum value as the final fitness of the agent;
iii) Respectively selecting the final fitness of all the intelligent agents to form a final fitness sequence of the multi-intelligent-agent system in the current tracking period;
S3) constructing a multi-agent control strategy
In living body calculation, under the action of an external unified control magnetic field, the position update of the intelligent body has consistency, and in each step of iterative update, an optimal position update mode is required to be selected to realize the overall optimal evolution of the multi-intelligent body system, so that a weak person priority evolution strategy is adopted to control the multi-intelligent body to realize the optimal update of the intelligent body position in the living body calculation under the action of the external unified control magnetic field: selecting an intelligent agent corresponding to the minimum fitness in a final fitness sequence of a current tracking period as a selected individual w 1 in the multi-intelligent agent system, and using a position updating mode of w 1 as an updating mode of the whole multi-intelligent agent system to realize state updating of the multi-intelligent agent system under the action of an external unified control magnetic field; according to the control strategy, selecting an individual w 2 with the minimum fitness in the final fitness sequence in the next tracking period of iterative operation, and realizing further updating of the motion condition of the multi-intelligent system;
when the natural computing method is applied to detecting tumor focus positions in contrast agent or drug targeting delivery, the method specifically comprises the following steps:
a. Setting the number of the intelligent agents in the multi-intelligent-agent system and the initial position of each intelligent agent in the search space;
b. calculating the initial fitness of each intelligent agent according to the initial position of each intelligent agent in the search space;
c. Particle swarm optimization algorithm is used as swarm intelligent algorithm to calculate the intelligent body position of' history optimal And "globally optimal" agent locations
D. sampling the randomly moving intelligent agents for a plurality of times in a tracking period, and calculating the final fitness of each intelligent agent in each iterative calculation by using an optimal state retention mechanism;
e. the 'weak person priority' evolution strategy is used for realizing the directional motion control of the multi-agent system in a control period, and the control period and the tracking period form the motion period of the agent;
f. Evaluating fitness and life of each agent, and estimating relative distance between each agent and tumor focus, and removing an agent from the multi-agent system when life of the agent is terminated; when a certain agent reaches the tumor focus position, the agent stops moving;
g. and when the maximum running time of the algorithm is reached, calculating the final position distribution condition of the multi-agent system in the search space, and obtaining the space position of the tumor focus and the detection efficiency of the multi-agent system on the position of the tumor focus.
2. The population intelligence-based natural computing method of claim 1, wherein the directional motion formula of the agent is represented as follows:
Wherein, c 1,c2 is two different learning factors, and the value is any number in [0,4 ]; r 1,r2 represents a random number between (0, 1), respectively; A blood flow rate at time (t+1); Representing the velocity of agent a j at time t, as selected by the "weak priority" evolution strategy; And The position coordinates of any agent A i at time t and time t+1 are respectively represented; And Respectively representing the movement speeds of the intelligent agent A i at the time t and the time t+1; d 0 represents the movement speed of the agent under the individual control of the external driving field; The motion speed of a i at the time t+1 obtained by applying the "optimal state retention" mechanism and the "weak priority" evolution strategy is represented by i=1, 2, …, K, j=1, 2, …, K; and 2 denotes a 2-norm.
3. The population intelligence based natural computing method of claim 2, wherein K = 20.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111158377.6A CN114021690B (en) | 2021-09-30 | 2021-09-30 | A natural computing method based on swarm intelligence |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111158377.6A CN114021690B (en) | 2021-09-30 | 2021-09-30 | A natural computing method based on swarm intelligence |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114021690A CN114021690A (en) | 2022-02-08 |
| CN114021690B true CN114021690B (en) | 2024-11-22 |
Family
ID=80055359
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111158377.6A Active CN114021690B (en) | 2021-09-30 | 2021-09-30 | A natural computing method based on swarm intelligence |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114021690B (en) |
Family Cites Families (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP1205863A1 (en) * | 2000-11-14 | 2002-05-15 | Honda R&D Europe (Deutschland) GmbH | Multi-objective optimization |
| US20170026305A1 (en) * | 2015-07-23 | 2017-01-26 | Schneider Electric It Corporation | System to place virtual machines onto servers based upon backup runtime constraints |
| CN112488310A (en) * | 2020-11-11 | 2021-03-12 | 厦门渊亭信息科技有限公司 | Multi-agent group cooperation strategy automatic generation method |
| CN112949933B (en) * | 2021-03-23 | 2022-08-02 | 成都信息工程大学 | A traffic organization scheme optimization method based on multi-agent reinforcement learning |
| CN113190018B (en) * | 2021-05-24 | 2022-07-01 | 东南大学 | Intelligent agent path control method based on improved course error rate |
-
2021
- 2021-09-30 CN CN202111158377.6A patent/CN114021690B/en active Active
Non-Patent Citations (1)
| Title |
|---|
| "面向早期肿瘤检测应用的活体计算技术研究";师少龙;《中国博士学位论文全文数据库》;20210115;第3-5章 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114021690A (en) | 2022-02-08 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Yue et al. | Review and empirical analysis of sparrow search algorithm | |
| Xing et al. | Boosting whale optimizer with quasi-oppositional learning and Gaussian barebone for feature selection and COVID-19 image segmentation | |
| JP6078943B2 (en) | Image processing apparatus and method, and program | |
| Tang et al. | An improved multilevel thresholding approach based modified bacterial foraging optimization | |
| CN112231964B (en) | Gas leakage source autonomous searching and positioning method based on deep reinforcement learning | |
| Deetjen et al. | High-speed surface reconstruction of a flying bird using structured light | |
| CN116594289A (en) | Robot gesture pre-adaptation control method and device, electronic equipment and storage medium | |
| Shi et al. | Nanorobots-assisted natural computation for multifocal tumor sensitization and targeting | |
| Shi et al. | Lightweight evolution strategies for nanoswimmers-oriented in vivo computation | |
| Shi et al. | Dynamic in vivo computation for learning-based nanobiosensing in time-varying biological landscapes | |
| CN114021690B (en) | A natural computing method based on swarm intelligence | |
| Pan et al. | A quasi-affine transformation evolutionary algorithm enhanced by hybrid Taguchi strategy and its application in fault detection of wireless sensor network | |
| Song et al. | Real time algorithm for magnet's localization in capsule endoscope | |
| Ali et al. | Self-regulated and co-ordinated smart tumor homing for complex vascular networks | |
| CN114706384A (en) | Multi-machine navigation method, system and medium for maintaining connectivity | |
| CN105913458B (en) | A kind of brain white matter integrity imaging method based on group's tracking | |
| Cao et al. | An improved bacterial foraging algorithm with cooperative learning for eradicating cancer cells using nanorobots | |
| CN105469422A (en) | Target tracking method based on pigeon intelligent optimization Kalman filtering parameters | |
| Ali et al. | Bio-inspired self-regulated in-vivo computation for smart cancer detection | |
| Wang et al. | Reinforcement Learning for Multifocal | |
| Lu et al. | An accurate cell tracking approach with self-regulated foraging behavior of ant colonies in dynamic microscopy images | |
| Ding et al. | A Dynamic In Vivo Computational Strategy for Tumor Targeting | |
| CN120319491B (en) | Nerve regulation and control target positioning method and system based on magnetic resonance and electroencephalogram fusion | |
| Shi et al. | In Vivo Computation for Tumor Sensitization and Targeting at Different Tumor Growth Stages | |
| Li et al. | A biologically inspired model based on a multi-scale spatial representation for goal-directed navigation. |
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 |