CN117689022A

CN117689022A - Determination method of optimal rule base for real-time navigation of clinical exploration robot

Info

Publication number: CN117689022A
Application number: CN202410154602.6A
Authority: CN
Inventors: 陈虎; 涂燕晖; 姚振杰; 陈一昕
Original assignee: Jiangsu Future Network Group Co ltd
Current assignee: Jiangsu Future Network Group Co ltd
Priority date: 2024-02-04
Filing date: 2024-02-04
Publication date: 2024-03-12
Anticipated expiration: 2044-02-04
Also published as: CN117689022B

Abstract

The embodiment of the application provides a method for determining an optimal rule base for real-time navigation of a clinical exploration robot, which comprises the following steps: acquiring distance parameters of the robot and the obstacle and non-obstacle boundary in multiple directions through a radar ranging sensor of the robot; processing the distance parameters to generate the final linear speed and the final angular speed of the robot when the robot is right to travel and avoid the obstacle; the robot is controlled to further record the current coordinates of the robot according to the final linear speed and the final angular speed, and the fitness value scoring is carried out on a rule base currently used by the robot based on the current coordinates; the operation is circulated until the robot navigation failure condition is triggered, and the final score of the rule base is recorded; traversing all rule bases, generating the adaptability corresponding to each rule base, and taking the rule base with the highest adaptability value as the optimal rule base. The embodiment of the application can replace manual determination of the optimal rule base for real-time navigation of the robot.

Description

Determination method of optimal rule base for real-time navigation of clinical exploration robot

Technical Field

The application relates to the technical field of automation, in particular to a method for determining an optimal rule base for real-time navigation of a clinical exploration robot.

Background

Serious patients in hospitals are easily affected by bacteria, viruses and external environment interference, so that the illness is aggravated, and the family members of the serious patients are generally not allowed to enter the patients. Remote clinical visits are the best choice for alleviating the jeopardy between family and patient. With the strong development of deterministic networks and 5G technology in recent years, it is expected that clinical exploration services with remote real-time holographic imaging, language communication, limb contact, palpation, kissing, etc. functions will become reality. In order to enable patients to enjoy the remote visit service, the general idea is to install a remote visit device at the bedside of each patient, so that personalized visit service can be provided for the patient, however, the visit device not only occupies the space of other bedside monitoring devices, but also is only started under specific conditions (the body of the patient meets the visit requirement and the patient and family members have visit needs) and wastes medical resources, so that the installation of the device for each bed is unnecessary. In view of the above, it is most reasonable to provide all-weather patient visit services using a clinical visit machine that is capable of adapting to all-terrain. The doctor only needs to send the position information and the visit time of the patient to the robot, and the clinical visit robot can automatically bypass the obstacle and reach the bedside of the patient in the shortest path, automatically return after the visit service is finished, does not occupy the ward space, and reduces the medical resource idling. In addition to critically ill patients, infectious patients and even general patients may use the exploratory services of a clinical exploratory robot.

The heart of the clinical exploration robot behavior decision is an autonomous navigation algorithm. Conventional algorithms for autonomous navigation include proportional-integral-derivative (PID) algorithms, which are also widely used in the field of industrial control. The document IEEE Trans Cybern (Navigation of a Fuzzy-Controlled Wheeled Robot Through the Combination of Expert Knowledge and Data-Driven Multiobjective Evolutionary Learning) uses a PID algorithm to control the direction and speed of the robot respectively, automatically completes path planning, and simultaneously reduces the distance between the departure place and the destination to the greatest extent while realizing the functions of obstacle avoidance and the like. The documents Neuron (Deep Reinforcement Learning and Its Neuroscientific Implications), IEEE Trans Cybern (Deep Reinforcement Learning With Quantum-Inspired Experience Replay) and Med Image animal (Deep reinforcement learning in medical imaging: A literature review) also consider that the PID can handle simple navigation tasks, and for a complex navigation system with multiple inputs and outputs, the PID cannot meet the requirements, and a Deep Reinforcement Learning (DRL) algorithm is required to establish the input-output relationship of the complex navigation system, so that the machine learns and makes a decision in practice. However, the DRL algorithm performs path selection through multiple agents, once the number and the scale of the intelligent agents are increased, the scale of the joint behavior space and the joint observation space becomes extremely huge, the robot calculation pressure is increased in level, the real-time navigation capability of the robot is greatly reduced, and in this way, documents IEEE Trans Cybern (Fuzzy Control Design of Nonlinear Time-Delay Parabolic PDE Systems Under Mobile Collocated Actuators and Sensors), bioresour Technol (A novel cascaded fractional fuzzy approach for control of fermentation process) and ISA Trans (Adaptive fuzzy fractional-order sliding-mode control of LCL-interleaved grid-connected converter with reduced-order) shift the research center of gravity to a Fuzzy Inference System (FIS), and the FIS also has the capability of processing complex input and output, and has fast response time and good real-time performance. The navigation performance of the FIS mainly depends on the structure of the fuzzy controller, a used fuzzy control rule base, a comprehensive reasoning algorithm, a fuzzy decision method and other factors. The fuzzy control rule base is a FIS behavior decision brain, and establishes a mapping relation between rule precursors and intelligent agent behaviors, and is usually obtained by artificial long-term practice and debugging and experience. However, in a complex navigation system, a rule base is often trapped in a local optimal solution by means of manual debugging and experience accumulation, and a heuristic algorithm is urgently needed to replace manual determination of the optimal rule base.

Disclosure of Invention

The embodiment of the application provides a method for determining an optimal rule base for real-time navigation of a clinical exploring robot, which can replace manual determination of the optimal rule base for real-time navigation of the robot.

The embodiment of the application provides a method for determining an optimal rule base for real-time navigation of a clinical exploration robot, which comprises the following steps:

acquiring distance parameters of the robot and the obstacle and non-obstacle boundary in multiple directions through a radar ranging sensor of the robot;

the distance parameter is used as an input variable of the robot right-to-the-road driving and obstacle avoidance, the input variable is subjected to fuzzification treatment, and the final linear speed and the final angular speed of the robot right-to-the-road driving and obstacle avoidance are generated through a fuzzy inference engine and an anti-fuzzy operation carried by the robot according to a rule base;

controlling the robot to further record the current coordinates of the robot according to the final linear speed and the final angular speed, and scoring the fitness value of a rule base currently used by the robot based on the current coordinates;

the operation is circulated until the robot navigation failure condition is triggered, and the final score of the rule base is recorded;

Traversing all rule bases, generating fitness corresponding to each rule base, and obtaining a first fitness set formed by all the fitness and a first rule base set corresponding to the first fitness set;

normalizing rule back parts of all rules in the first rule base set to obtain a normalized second rule base set;

selecting a better rule base from the second rule base set by using a roulette algorithm, forming a third rule base set, performing a cross operation on the third rule base set to generate a fourth rule base set, and performing a mutation operation on the fourth rule base set to generate a fifth rule base set;

judging whether the genetic iteration times exceeds a preset evolution algebra, if so, taking the rule base with the highest fitness value in the first rule base set as an optimal rule base.

In the method for determining an optimal rule base for real-time navigation of a clinical exploration robot according to the embodiment of the present application, the obtaining, by a radar ranging sensor of the robot itself, distance parameters of the robot and an obstacle and non-obstacle boundary in multiple directions includes:

Acquiring the shortest distances between the robot and the obstacle in a plurality of first preset directions through a radar ranging sensor of the robot to obtain a plurality of first shortest distances, and acquiring the shortest distances between the robot and the boundary of the non-obstacle in a plurality of second preset directions to obtain a plurality of second shortest distances;

and taking the plurality of first shortest distances and the plurality of second shortest distances as distance parameters of the robot and the boundaries of the obstacle and the non-obstacle in a plurality of directions.

In the method for determining the optimal rule base for real-time navigation of the clinical exploring robot, the robot is a spherical robot, and the radar ranging sensor is installed in the center of mass of the robot.

In the method for determining an optimal rule base for real-time navigation of a clinical exploration robot according to the embodiment of the present application, the acquiring, by a radar ranging sensor of the robot itself, shortest distances between the robot and an obstacle in a plurality of first preset directions to obtain a plurality of first shortest distances, and acquiring shortest distances between the robot and a non-obstacle boundary in a plurality of second preset directions to obtain a plurality of second shortest distances includes:

Acquiring first distances between the radar ranging sensor and an obstacle in a plurality of first preset directions through the radar ranging sensor of the spherical robot, and acquiring second distances between the radar ranging sensor and a non-obstacle boundary in a plurality of second preset directions;

and subtracting the radius of the spherical robot from the first distances and the second distances to obtain the first shortest distances and the second shortest distances.

In the method for determining an optimal rule base for real-time navigation of a clinical exploring robot according to the embodiment of the present application, the step of using the distance parameter as an input variable for right-to-robot travel and obstacle avoidance, performing a blurring process on the input variable, and generating a final linear velocity and a final angular velocity for right-to-robot travel and obstacle avoidance by a fuzzy inference engine and an anti-blurring operation carried by the robot according to the rule base, includes:

the distance parameter is used as an input variable of the robot to right and avoid obstacle, the input variable is subjected to fuzzification processing, and a membership function of each input variable is generated;

Inputting all the input variables and all the membership functions into a fuzzy inference engine of the robot, converting the input variables into a combination of a plurality of groups of sets through the fuzzy inference engine according to a rule base, giving corresponding activation energy, and generating special linear speed and special angular speed of a rule precursor;

and generating a final linear speed and a final angular speed of the robot to right and avoid the obstacle based on the activation energy, the linear speed and the angular speed.

In the method for determining an optimal rule base for real-time navigation of a clinical exploring robot according to the embodiment of the present application, the generating the final linear velocity and the final angular velocity of the robot right-to-travel and obstacle avoidance based on the activation energy, the linear velocity and the angular velocity includes:

generating a weighted angular velocity and a weighted linear velocity of the robot right-to-travel and obstacle avoidance based on the activation energy, the linear velocity and the angular velocity;

obtaining the membership function corresponding to the input variable of the robot for making obstacle avoidance and right-to-travel function selection, and generating the final linear speed and the final angular speed of the robot for right-to-travel and obstacle avoidance based on the membership function, the weighted angular speed and the weighted linear speed.

In the method for determining an optimal rule base for real-time navigation of a clinical exploring robot according to the embodiment of the present application, after determining whether the number of genetic iterations exceeds a preset evolution algebra, the method further includes:

if not, carrying out inverse normalization processing on all rule bases in the fifth rule base set;

traversing the rule base after inverse normalization, and then continuing to execute the normalization, crossing and mutation operations of the rule base until the number of genetic iterations exceeds a preset evolution algebra;

and taking the rule base with the highest fitness value in the first rule base set as the optimal rule base.

The method for determining the optimal rule base for real-time navigation of the clinical exploration robot in the embodiment of the application further comprises the following steps:

and when the moving step number of the robot exceeds the maximum allowable step number, the robot fails to navigate.

The embodiment of the application also provides a device for determining the optimal rule base for the real-time navigation of the clinical exploration robot, the device comprises a processor and a memory, wherein a computer program is stored in the memory, and the processor is used for executing the method for determining the optimal rule base for the real-time navigation of the clinical exploration robot, which is described in any embodiment, by calling the computer program stored in the memory.

The embodiment of the application also provides a computer readable storage medium, wherein a computer program is stored in the computer readable storage medium, and when the computer program runs on a computer, the computer is caused to execute the method for determining the optimal rule base for real-time navigation of the clinical exploration robot.

According to the method for determining the optimal rule base for the real-time navigation of the clinical exploring robot, distance parameters of the robot, obstacle and non-obstacle boundaries in multiple directions are obtained through the radar ranging sensor of the robot, then the distance parameters are used as input variables of robot right-following driving and obstacle avoidance, fuzzy processing is carried out on the input variables, and final linear speed and final angular speed of robot right-following driving and obstacle avoidance are generated through a fuzzy inference engine and anti-fuzzy operation carried out by the robot according to the rule base; then controlling the robot to further record the current coordinates of the robot before the robot is controlled to a final linear speed and a final angular speed, and scoring the fitness value of a rule base currently used by the robot based on the current coordinates; then, the operation is circulated until the robot navigation failure condition is triggered, and the final score of the rule base is recorded; traversing all rule bases, generating the fitness corresponding to each rule base, and obtaining a first fitness set formed by all fitness and a first rule base set corresponding to the first fitness set; then, carrying out normalization processing on rule back parts of all rules in the first rule base set to obtain a second rule base set after normalization processing; selecting a better rule base from the second rule base set by using a roulette algorithm, forming a third rule base set, performing cross operation on the third rule base set to generate a fourth rule base set, and performing mutation operation on the fourth rule base set to generate a fifth rule base set; and finally judging whether the genetic iteration times exceed a preset evolution algebra, if so, taking a rule base with the highest fitness value in the first rule base set as an optimal rule base, so that the optimal rule base for real-time navigation of the robot can be determined instead of manual work.

Drawings

In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings that are required to be used in the description of the embodiments will be briefly described below. It is obvious that the drawings in the following description are only some embodiments of the present application, and that other drawings may be obtained from these drawings without inventive effort for a person skilled in the art.

Fig. 1 is a flowchart of a method for determining an optimal rule base for real-time navigation of a clinical exploration robot according to an embodiment of the present application.

Fig. 2 is another flow chart of a method for determining an optimal rule base for real-time navigation of a clinical exploration robot according to an embodiment of the present application.

Fig. 3 is a membership function of each input variable for realizing right-to-right driving of the robot turtlebot3 provided in the embodiment of the present application on an XY two-dimensional plane.

Fig. 4 is a membership function of each input variable of the robot turtlebot3 for realizing the obstacle avoidance function on the XY two-dimensional plane provided in the embodiment of the present application.

Fig. 5 is a membership function of the robot turtlebot3 provided in the embodiment of the present application to make two function selections of input variables for avoiding obstacles and leaning to travel on the XY two-dimensional plane.

Fig. 6 is a cross and variance diagram of rule base (a) and rule base (b) provided in an embodiment of the present application.

Fig. 7 is a transient diagram of the robot turtlebot3 traveling in the three-dimensional simulation environment provided in the embodiment of the present application.

Fig. 8 is a graph of correspondence between genetic iteration optimization times and rule base fitness values in a specific embodiment provided in an embodiment of the present application.

Fig. 9 is a schematic structural diagram of a determining device of an optimal rule base for real-time navigation of a clinical exploration robot according to an embodiment of the present application.

Detailed Description

The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. It will be apparent that the described embodiments are only some, but not all, of the embodiments of the present application. All other embodiments, which can be made by a person skilled in the art without any inventive effort, are intended to be within the scope of the present application based on the embodiments herein.

The embodiment of the application provides a method for determining an optimal rule base for real-time navigation of a clinical exploration robot, which is jointly realized by a fuzzy reasoning module and a genetic algorithm module.

Referring to fig. 1 and 2, the method may include the steps of:

Step 101, acquiring distance parameters of the robot and the obstacle and non-obstacle boundaries in multiple directions through a radar ranging sensor of the robot.

The robot is provided with a radar ranging sensor, and the radar ranging sensor is used for acquiring distance parameters of the robot and surrounding obstacles and non-obstacle boundaries in multiple directions. The non-obstacle boundary refers to a non-obstacle boundary on the right side of the robot, such as a road boundary, when the robot is right-to-travel.

Wherein the robot is a clinical exploration robot, but is not limited to a clinical exploration robot.

In some embodiments, the acquiring, by the radar ranging sensor of the robot itself, distance parameters of the robot and the obstacle and non-obstacle boundary in multiple directions includes:

Wherein the plurality of first shortest distances includes a shortest distance between the robot left and the obstacle boundary, a shortest distance between the robot right and the obstacle boundary, and a shortest distance between the robot middle and the obstacle boundary. The plurality of second shortest distances includes a shortest distance between a right front of the robot and a right boundary (non-obstacle boundary) and a shortest distance between a right rear of the robot and the right boundary (non-obstacle boundary).

In some embodiments, the robot is a spherical robot, the centroid of the robot being mounted with the radar ranging sensor.

In some embodiments, the acquiring, by the radar ranging sensor of the robot itself, the shortest distance between the robot and the obstacle in a plurality of first preset directions to obtain a plurality of first shortest distances, and acquiring, by the radar ranging sensor of the robot, the shortest distance between the robot and the non-obstacle boundary in a plurality of second preset directions to obtain a plurality of second shortest distances includes:

Step 102, taking the distance parameter as an input variable of the robot right-to-travel and obstacle avoidance, carrying out fuzzification processing on the input variable, and generating a final linear speed and a final angular speed of the robot right-to-travel and obstacle avoidance through a fuzzy inference engine and anti-fuzzy operation carried by the robot according to a rule base.

In some embodiments, the step of using the distance parameter as an input variable of the robot right-to-travel and obstacle avoidance, and performing fuzzification processing on the input variable, and generating a final linear velocity and a final angular velocity of the robot right-to-travel and obstacle avoidance through a fuzzy inference engine and an anti-fuzzy operation carried by the robot according to a rule base, includes:

Blurring of input variables:

among them, the objective environment of the navigation system is complex and has uncertainty, and it is generally impossible to express the relationship between input and output in the objective environment using an accurate parsing scheme. However, in the course of developing, debugging, maintaining and iterating navigation systems, experts have obtained a great deal of experience in decision making of navigation devices in various complex scenarios, which are often recorded in ambiguous expressions. To fully exploit these experiences, the input variables need to be obfuscated, i.e., the membership functions of the input variables are determined first.

The all-terrain is a space curved surface, the three-dimensional space curved surface is projected to three two-dimensional planes XY, YZ and ZX, and the navigation performance of the robot on the three-dimensional curved surface can meet the requirements as long as the robot is ensured to meet the navigation requirements on the three two-dimensional projection planes.

As shown in fig. 3, fig. 3 is a membership function of each input variable of the robot turtlebot3 for realizing right-to-right travel on an XY two-dimensional plane. the input variables of the turbolebot 3 are (shortest distance (second shortest distance) from the right boundary in front of the turbolebot 3), dfre_xy, and (shortest distance (second shortest distance) from the right boundary behind the turbolebot 3), dbre_xy, respectively. The projection of radar scanning area on XY plane is fan-shaped, and the invention firstly uses radian as The sector radar scanning area in the range completes the measurement of the shortest distance between the barycenter of the turtlebot3 and the right boundary, and then the shortest distance is used for subtracting the radius of turtlebot3 to obtain DFRE_XY,>representing half of the corresponding scan area of DFRE XY. In the same way, the radian is +.>The sector radar scanning area in the range completes the measurement of the shortest distance between the barycenter of the turtlebot3 and the right boundary, and then the DBRE_XY can be obtained by subtracting the radius of turtlebot3 from the shortest distance>Representing DBRE_XY correspondenceHalf of the scan area. In fig. 3, a divides DFRE variables into five sets, i.e., a "lower", "low", "medium", "high" and "higher" sets from small to large, and there is an overlap between the sets, and in order to determine the degree to which dfre_xy variables are affiliated to each set, we take dfre_xy as a0_1 as an example, and the degree (confidence) to which it is affiliated to the "lower" set and the "low" set is 0.42 and 0.58, respectively, and the degree to which it is affiliated to the other sets is 0. The dfre_xy variable consists of membership functions for these five sets, where the "lower" and higher "membership functions are trapezoids and the remaining set membership functions are triangles. The membership function is adopted to represent the relation among the sets, so that the change process of changing the quantity to the quality change among the sets can be represented, and when the problem of the set attribution of the DFRE_XY in the overlapping area is solved, the error operation that only one of the sets can be selected is avoided. The specific values of these five super-parameters are determined by the navigational behavior of turtlebot3 in the simulation environment. Similarly, the meaning of the membership function of DBRE_XY can be explained. The YZ plane and the ZX plane contain dfre_yz and dbre_yz and dfre_zx and dbre_zx, respectively. The interpretation and membership function modeling of DFRE_YZ and DFRE_ZX are identical to those of DFRE_XY; the interpretation and membership function modeling of DBRE_YZ and DBRE_ZX are exactly the same as DBRE_XY.

Fig. 4 is a membership function of each input variable of the robot turtlebot3 for realizing the obstacle avoidance function on the XY two-dimensional plane. the input variables of the turtlebot3 are (shortest distance (first shortest distance) from the obstacle boundary to the left of the turtlebot 3), dleo_xy, (shortest distance (first shortest distance) from the obstacle boundary in the middle of the turtlebot 3), dmeo_xy, and (shortest distance (first shortest distance) from the obstacle boundary to the right of the turtlebot 3), dreo_xy, respectively. Using radian asThe sector radar scanning area in the range completes the measurement of the shortest distance between the barycenter of the turtlebot3 and the left boundary of the obstacle, and then the radius of turtlebot3 is subtracted from the shortest distance to obtain DLEO_XY,>representing half of the corresponding scan area of DLEO XY. The radian is +.>The sector radar scanning area in the range completes the measurement of the shortest distance between the barycenter of the turtlebot3 and the upper boundary of the obstacle, and then the radius of turtlebot3 is subtracted from the shortest distance to obtain DMEO_XY,>representing half of the corresponding scan area of dmeo_xy. The radian is +.>Andthe sector radar scanning area in the range completes the measurement of the shortest distance between the barycenter of the turtlebot3 and the right boundary of the obstacle, and then the DREO_XY can be obtained by subtracting the radius of turtlebot3 from the shortest distance >The DREO_XY is shown corresponding to half of the upper right and lower right scan areas, respectively. The membership of dl eo_xy, dmeo_xy and d eo_xy are exactly the same. The input variables for the YZ plane and the ZX plane include dleo_yz, dmeo_yz and dreo_yz, respectively, and dleo_zx, dmeo_zx and dreo_zx. The interpretation and membership function modeling of DLEO_YZ and DLEO_ZX are exactly the same as that of DLEO_XY; the interpretation and membership function modeling of dmeo_yz and dmeo_zx are exactly the same as dmeo_xy; the interpretation and membership function modeling of DREO_YZ and DREO_ZX are exactly the same as that of DMEO_XY.

Fig. 5 is a membership function of the robot turtlebot3 in the XY two-dimensional plane for two function choices of avoidance obstacle and right-to-travel. The membership function consists of a set for avoiding obstacles and a set for right-to-travel, and the two sets also have overlapping areas. the input variable of the turtlebot3 is (shortest distance of turtlebot3 from obstacles and boundaries, deo_xy). Using radian as、/>And->The sector radar scanning area within the range completes the measurement of the shortest distance between the mass center of the turtlebot3 and the boundary between the obstacle and the non-obstacle, and then the DEO_XY can be obtained by subtracting the radius of the turtlebot3 from the shortest distance, < ->And indicates that DEO XY corresponds to half of the right, upper and left scan areas, respectively. Similarly, the input variables of the YZ plane and the ZX plane include deo_yz and deo_zx, respectively, and the interpretation and membership function modeling of deo_yz and deo_zx are exactly the same as deo_xy.

Rule base:

the rule base is a decision center of fuzzy reasoning, each rule is composed of a rule front part and a rule back part, the rule front part executes ' and ' or ' operation by a plurality of conditions, and the rule back part is a specific value of an output variable of the robot under the rule front part.

The robot realizes the right-to-travel function on the XY plane, two input variables, namely DFRE_XY and DBRE_XY, are required to be input to the robot at the same time, so that the two input variables only have a 'and' relation, and because each input variable has five sets of 'lower', 'low', 'medium', 'high', and 'higher', all input combinations (rule front pieces) of the two variables are 5 multiplied by 5=25, and only the angular speed and the linear speed of the robot in the XY plane, which are respectively corresponding to the 25 rules, are required to be determined, namely the rule rear pieces of the rules on the XY plane are determined. Similarly, 25 rules are also included in the YZ and ZX planes.

The robot needs three input variables, i.e., dleo_xy, dmeo_xy, and dreo_xy, respectively, each input variable having five sets, so that all combinations of these three variables have a total of 5×5×5=125 rules. Similarly, the robot can realize that the capability of avoiding obstacle also comprises 125 rules on the YZ and ZX planes respectively.

Fuzzy inference engine:

taking right-to-the-right traveling on the XY plane as an example, when dfre_xy and dbre_xy take values of a0_1 and a2_3, respectively, the vertical lines a0_1 and "lower" intersect at two points (a0_ 1,0.42) and (a2_ 3,0.58), respectively, and the vertical lines a2_3 and "high" and "medium" intersect at two points (a2_ 3,0.34) and (a2_ 3,0.64), respectively (fig. 3). All rules involved in dfre_xy=a0_1 and dbre_xy=a2_3 are:

rule 1: front piece: lower in height&High→rule back part:；

rule 2: front piece: low and low&High→rule back part:；

rule 3: front piece: lower in height&Medium→rule back part:；

rule 4: front piece: low and low&Medium→rule back part:；

wherein, rule 1 front lower & high indicates that the dfre_xy fuzzy set value is "lower" and the dbre_xy fuzzy set value is "high"; the upper corner mark min (0.42,0.34) =0.34 of the rule 1 back part linear velocity and angular velocity, as the activation energy (output membership function value) of the rule on the linear velocity and angular velocity, the lower and higher corner mark indicates the exclusive linear velocity and angular velocity of the rule front part. Rules 2-4 can be interpreted similarly.

Without loss of generality, we use mathematical formulas to represent the fuzzy inference engine principle of the invention. A total of n variables are input to the fuzzy system, each variable having its own membership function. When inputting for a certain time, the values of the variables are The intersection point (ordinate is greater than 0) of the vertical line and the respective membership function respectively determines that the set and membership value are +.>And->Representing the number of intersections of the vertical line of the variable with the set, < >>Respectively represent the set of the vertical lines of the variable intersecting with the membership function from bottom to top, +.>Representing the variable->The front piece is +.>The activation energy (output membership function value) of the rule and rule back are +.>And. V represents a linear velocity and G represents an angular velocity. The fuzzy inference machine (comprising XY, YZ and ZX planes) for realizing right-to-the-way and obstacle avoidance functions can be expressed by using the mathematical formulas.

In some embodiments, the generating the final linear velocity and the final angular velocity of the robot right-to-travel and obstacle avoidance based on the activation energy, linear velocity, and angular velocity comprises:

Inverse modeling of input variables:

without loss of generality, defuzzification is to solve for the weighted linear velocity of the robot outputAnd weighted angular velocity。

Weighted linear velocityAnd weighted angular velocity>The calculation formula is general and is based on +.>Andcan calculate the weighted linear velocity and the weighted angular velocity of obstacle avoidance and approach travel on different planes respectively as +.>And->Wherein the F takes the values of 'right-to-travel' and 'obstacle avoidance'; the value of P is XY, YZ or ZX.

When the robot selects the obstacle avoidance function and the right-to-travel function in a specific scene, the fuzzy set (membership function) is used for expressing the relation between the obstacle avoidance set and the right-to-travel set, so that the existence of overlapping and the transformation of the quantity between the obstacle avoidance set and the right-to-travel set can be more embodied (figure 5). The vertical line of DEO_P (P is XY, YZ or ZX) and the vertical coordinate of intersection point of obstacle avoidance set and right-to-travel set of the membership function of DEO_P (P is XY, YZ or ZX) are. The final linear velocity of the robot on each plane is:wherein, the value of P is XY, YZ or ZX.

The final angular velocity of the robot on each plane is:。

and 103, controlling the robot to further record the current coordinates of the robot according to the final linear speed and the final angular speed, and scoring the fitness value of a rule base currently used by the robot based on the current coordinates.

Wherein the genetic algorithm is a computational model of the biological evolution process simulating the natural selection and genetic mechanism of the Darwin biological evolutionary theory. The method consists of six parts including genetic algorithm initialization, calculation of fitness of a rule base, selection of the rule base, intersection of the rule base, variation of the rule base and determination of an optimal rule base.

Initializing a genetic algorithm:

each rule base is called an individual, and the invention initializes a population to p_num in size, i.e., a population contains p_num individuals. The cross probability and the variation probability of all the rules in the invention are respectivelyAnd->The evolution algebra is initialized to g_num, i.e. the genetic iteration is stopped when the evolution algebra is exceeded.

Wherein, the person skilled in the art can set P_num according to the actual situation,、/>And a value of g _ num,the present invention is not particularly limited herein.

Calculating the adaptability of the rule base:

the simulation environment of the robot is a three-dimensional space, in order to check the advantages and disadvantages of each rule base (individual), the robot needs to clear all space coordinate information before entering the simulation environment initially, and after entering the simulation environment, the initial starting position is calibrated to be (0, 0), and the robot needs to record all past space coordinates. A radar sensor on the robot acquires distance parameters (input variables) and inputs the distance parameters (input variables) to the robot, and the robot fuzzifies the input variables and outputs the input variables through a fuzzy inference engine and anti-fuzzy operation according to a rule base And->，/>And->Controlling the linear and angular speeds of the robot in different planes, the robot moving to the next spatial position +.>And the position space coordinates are recorded, and if the position is not recorded in the past, the fitness value is increased by 1.

And 104, circulating the operation until the robot navigation failure condition is triggered, and recording the final score of the rule base.

Wherein navigation fails if the robot comes into contact with an obstacle or non-obstacle boundary, at which point the final score of the rule base currently used by the robot is remembered.

Wherein if the robot is in the middleThe radar sensor on the robot acquires distance parameters (input variables) and inputs the distance parameters to the robot, and the distance parameters are continuously acquired +.>And->To control the robot to continue moving in space until a navigation failure condition is triggered, the score of the rule base is the value of the final fitness.

In some embodiments, the robot navigation fails when the number of robot movements exceeds a maximum allowed number of steps.

Wherein, in order to prevent the robot from falling into the space dead loop, when the moving step number of the robot exceeds the maximum allowable step number max_step, the navigation fails.

And 105, traversing all rule bases, and generating the fitness corresponding to each rule base, so as to obtain a first fitness set formed by all the fitness and a first rule base set corresponding to the first fitness set.

For example, p_num rule bases are traversed to obtain a set H (first fitness set) of p_num fitness values. The p_num rule bases constitute a set S0 (first rule base set).

And 106, carrying out normalization processing on rule back parts of all rules in the first rule base set to obtain a second rule base set after normalization processing.

The genetic algorithm needs to normalize rule base back-parts, which is beneficial to accelerating the optimization speed. Without losing generality, arrangeThe maximum allowable values of the linear velocity and the absolute value of the angular velocity of all regular back parts are respectively expressed as the maximum allowable values of the robot on the P plane for realizing the F function. The invention specifies that the robot cannot be retracted. The linear and angular speeds of all regular back-pieces are normalized to +.>。

Wherein,representing the original values of the linear velocity and the angular velocity of the regular back part respectively; />The normalized values of the regular back part linear velocity and the angular velocity are shown, respectively. We use +.>Representing rule base->For the number of rules in the rule base, r_num is the number of rules in the rule base. And (3) normalizing all rule back parts in S0 to obtain a rule base set S1 (second rule base set).

And 107, selecting a better rule base from the second rule base set by using a roulette algorithm, forming a third rule base set, performing a crossover operation on the third rule base set to generate a fourth rule base set, and performing a mutation operation on the fourth rule base set to generate a fifth rule base set.

The preferred rule base includes rule bases with fitness values reaching preset fitness values, and those skilled in the art can set the preset fitness values according to actual situations, which are not limited herein.

Selection of a rule base:

the selection of the rule base is accomplished by using a roulette rule. The probability that each individual (rule base) is selected as the next generation evolution is。

Wherein,for numbering rule base>The expression number is->A value of fitness of a rule base of (a). According to->The rule base S2 (third rule base set) is formed by sampling from the rule base set S1 until the total number of samples to be sampled is p_num.

Intersection of rule base:

as shown in fig. 6, two specific rule bases (rule base (a) and rule base (b)) are shown using fig. 6. The ordinate of FIG. 6 is the normalized linear velocity corresponding to the rule base widget The abscissa is the normalized angular velocity corresponding to the regular back part。/>Rules for rule base (a); />Is a rule of rule base (b). The rule number 2 of rule base (a) is according to the crossover probability +.>Realize->And->I.e. rule base (a) and rule base (b) are exchanged according to probability +.>The regular allelic exchange numbered 2 is achieved. Other numbers are similarly given according to the cross probability +.>And realizing equipotential cross exchange. If P_num is odd, then the method is randomly repeatedAnd (3) preparing a rule base, wherein the number of the rule base sets S2 is even, randomly dividing the rule base sets into two equal teams, pairing the rule base sets two by two, and completing the intersection of the rule bases according to the intersection principle of the rule base (a) and the rule base (b) to obtain a new rule base set S3 (fourth rule base set).

Variation of rule base:

normalized rule back-pieceThe value after mutation is +.>。

Wherein,representing compliance mean +.>A normal distribution with variance 1; x is->Sampling to obtain the product. />Representing compliance mean +.>A normal distribution with variance 1; y is->Sampling to obtain the product.

In correspondence with the figure 6 of the drawings,values of the abscissa after mutation, +.>The value of the ordinate after the mutation, the rule number 2 of rule base (a) is according to the mutation probability +.>By->Become->. Each rule in the rule base set S3 is according to a probability The mutation is completed, and a rule base set S4 (fifth rule base set) is obtained. />

And step 108, judging whether the genetic iteration times exceed a preset evolution algebra, and if so, taking the rule base with the highest fitness value in the first rule base set as an optimal rule base.

The preset evolution algebra is an initialized evolution algebra g_num.

In some embodiments, after determining whether the number of genetic iterations exceeds the preset evolution algebra, the method further includes:

If the evolution algebra is not completed, the following inverse normalization is performed on each rule back part in the library set S4:。

the rule back part in S4 is a normalized value, and the original values of the linear velocity and the angular velocity are needed to be used by the robot in the simulation environment, so that inverse normalization of the rule back part is needed. And each rule base after the inverse normalization is completed is required to be iteratively executed in the steps 101 to 102 and the calculation of the fitness is completed, and in the steps 103 to 107, until the genetic evolution frequency exceeds the set evolution algebra g_num, stopping the genetic evolution, and taking the rule base with the highest fitness value in the S0 as the optimal rule base of the autonomous navigation algorithm of the robot (clinical exploration robot).

From the above, in a complex environment, a rule base for fuzzy reasoning for robot navigation is difficult to create by manual experience (expert knowledge), but an optimal rule base can be determined in a simulation environment by using a genetic algorithm. Firstly, initializing a rule base (individual) with a certain scale, and carrying out random sampling assignment in a [0,1] interval on the linear speed and the angular speed of a rule back part in the rule base. In order to evaluate the quality of each rule base, the robot needs to use the rule bases to autonomously navigate in the all-terrain simulation environment one by one until the condition of failure in navigation is triggered. Before the robot moves, the distance between the edges of the obstacle and the non-obstacle is measured by the radar ranging sensor of the robot, and the distance parameter is used as a variable for right-keeping running and obstacle avoidance. Because the rule base is a decision base based on fuzzy sets, the input variables are subjected to fuzzification to obtain membership functions of the input variables. The robot-carried fuzzy inference engine converts input variables into a combination of a plurality of groups of sets and gives corresponding activation energy, and then the output values of the linear speed and the speed for right-to-hand driving and obstacle avoidance on each plane can be obtained through anti-fuzzy operation. However, the right-side travel and obstacle avoidance are not split, and overlap (obstacle avoidance can be completed while right-side travel is completed), so that the membership functions of right-side travel and obstacle avoidance selection are used to determine weights using respective linear speeds and angular speeds, and then final linear speeds and angular speed values of the robot on each plane are obtained by weighting, and the robot is further arranged in front of the linear speeds and the angular speeds. If the advancing space coordinate robot does not count, the adaptability value is increased by 1 score, otherwise, the adaptability value is not added, and when the robot contacts with an obstacle or a non-obstacle or the advancing max_step is still not scored, the navigation failure condition is triggered, and the robot exits from the simulation environment. When the robot traverses all rule bases, a fitness set H and a corresponding rule base set S0 are obtained. Before using the genetic algorithm, the population scale (the number of rule bases), the crossover and mutation probability of the rule bases and the genetic evolution algebra need to be initialized. In addition, the rule base S1 is obtained by normalizing the back part of each rule in S0, so as to speed up the optimization of the genetic algorithm. Then, a roulette algorithm is used for selecting a better rule base from the S1 and forming a rule base set S2, then a cross operation is carried out on the S2 to obtain a rule base set S3, and then a mutation operation is carried out on the S3 to obtain a rule base set S4. If the evolution algebra is not exceeded, an inverse normalization operation is needed to be executed on the S4, and the original values of the linear velocity and the angular velocity in the rule base are obtained. Traversing the rule base after inverse normalization, performing iteration to execute the steps 101 to 105 to finally obtain a fitness set H, and then continuously performing rule base normalization, intersection and mutation until the number of genetic iterations exceeds evolution algebra, stopping the algorithm, and taking the rule base (individual) with the highest fitness in S0 as an optimal rule base.

The present application creates a robot (clinical exploration robot) autonomous navigation algorithm using Genetic Algorithm (GA) mentioned in literature Nano Lett (A genetic algorithm approach to probing the evolution of self-organized nanostructured systems) and Sensors (Path Planning of Mobile Robots Based on a Multi-Population Migration Genetic Algorithm) in combination with FIS, which maps the space to XY, YZ and ZX planes, and autonomously learns decision rule base (optimal rule base) on the three planes, respectively, for real-time navigation of the robot.

From the above, for the three-dimensional curved surface topography environment, the exploring robot can learn in real time to obtain a rule base, and is applied to navigation of the exploring robot, and the exploring service of a specific patient (serious patient, infectious patient and the like) is remotely completed. The fuzzy reasoning is a control method based on fuzzy theory, can solve the problems of uncertainty and ambiguity in the real world, and the basic idea is to realize the control of human experience by using a computer.

Specific examples:

the membership functions a 0-a 4 related to the application take values of 0.2m,0.4m,0.6m,0.8m and 1.01m respectively (figures 3-5); the variables b, c, d, e, f, g, h1, h2 and h3 related to the radar scanning area are respectively taken; radius value of Turtlebot3 is 0.2m; for right-to-drive and obstacle avoidance, the input variable n is 2 and 3 respectively; a population size p_num of 100; taking the cross probability to be 0.1; the variation probability is 0.1; the evolution algebra g_num takes 20,000; taking 3000 of the maximum allowable step number max_step of the robot; in order to realize the F function, the maximum allowable value of the linear speed of all regular back parts of the robot on the P plane is 0.2m/s, and the maximum allowable value of the absolute value of the angular speed is 0.15m/s2; each rule base contains r_num rules, r_num being 450 (25×3+125×3=450).

The construction of the simulation environment and the algorithm programming were completed using ubuntu 16.04, ROS2, gazebo, turtlebot3, and python 3.6. First, a robot turtlebot3 is created in a gazebo environment, which is a spherical structure with radar inside that can transmit radar waves in XY, YZ and ZX planes to measure the distance of an obstacle or boundary from the robot edge. Fig. 7 is a diagram showing the moment that the turtlebot3 travels in the three-dimensional simulation environment, nine white pillars in the diagram simulate obstacles, a hexagonal enclosure simulates the boundary of a road, and a robot is required to travel right to the simulation environment and can automatically avoid the obstacles. The genetic algorithm performs 20,000 iterations, each iteration can obtain fitness values of 100 rule bases, the 100 values are arranged in order from large to small, and the minimum value, the average value and the maximum value of the fitness are taken to determine whether the rule bases are optimized (fig. 8). The abscissa Epoch of FIG. 8 is in units of 10≡2, and the ordinate is also in units of 10≡2. As can be seen from FIG. 8, when the number of iterations of genetic optimization reaches 17500, the model rule base is already optimized, the maximum fitness value is 20,000, and when the number of iterations is 17500, the rule base with the highest fitness is the optimal rule base.

Any combination of the above optional solutions may be adopted to form an optional embodiment of the present application, which is not described herein in detail.

In particular, the present application is not limited by the order of execution of the steps described, and certain steps may be performed in other orders or concurrently without conflict.

As can be seen from the above, according to the method for determining an optimal rule base for real-time navigation of a clinical exploring robot provided by the embodiment of the application, distance parameters of boundaries of the robot, an obstacle and a non-obstacle in multiple directions are obtained through a radar ranging sensor of the robot, then the distance parameters are used as input variables of right-to-robot driving and obstacle avoidance, fuzzy processing is performed on the input variables, and final linear speeds and final angular speeds of right-to-robot driving and obstacle avoidance are generated through a fuzzy inference engine and anti-fuzzy operation carried out by the robot according to the rule base; then controlling the robot to further record the current coordinates of the robot before the robot is controlled to a final linear speed and a final angular speed, and scoring the fitness value of a rule base currently used by the robot based on the current coordinates; then, the operation is circulated until the robot navigation failure condition is triggered, and the final score of the rule base is recorded; traversing all rule bases, generating the fitness corresponding to each rule base, and obtaining a first fitness set formed by all fitness and a first rule base set corresponding to the first fitness set; then, carrying out normalization processing on rule back parts of all rules in the first rule base set to obtain a second rule base set after normalization processing; selecting a better rule base from the second rule base set by using a roulette algorithm, forming a third rule base set, performing cross operation on the third rule base set to generate a fourth rule base set, and performing mutation operation on the fourth rule base set to generate a fifth rule base set; and finally judging whether the genetic iteration times exceed a preset evolution algebra, if so, taking a rule base with the highest fitness value in the first rule base set as an optimal rule base, so that the optimal rule base for real-time navigation of the robot can be determined instead of manual work.

The embodiment of the application also provides a device for determining the optimal rule base for the real-time navigation of the clinical exploration robot, and the device for determining the optimal rule base for the real-time navigation of the clinical exploration robot can be integrated in the robot.

Referring to fig. 9, fig. 9 is a schematic structural diagram of a determining device of an optimal rule base for real-time navigation of a clinical exploration robot according to an embodiment of the present application. The determination means 30 of the optimal rule base for real-time navigation of the clinical exploration robot comprises a memory 120, one or more processors 180, and one or more application programs, wherein the one or more application programs are stored in the memory 120 and configured to be executed by the processor 180; memory 120 may be used to store applications and data. The memory 120 stores application programs including executable code. Applications may constitute various functional modules. The processor 180 executes various functional applications and data processing by running application programs stored in the memory 120.

In particular, in this embodiment, the memory 120 stores a computer program, and the processor 180 executes the method for determining the optimal rule base for real-time navigation of the clinical exploration robot according to any one of the above embodiments by calling the computer program stored in the memory 120, where the processor includes a fuzzy inference module and a genetic algorithm module.

The embodiment of the application also provides a computer readable storage medium, wherein a computer program is stored in the computer readable storage medium, and when the computer program runs on a computer, the computer executes the method for determining the optimal rule base for real-time navigation of the clinical exploration robot.

It should be noted that, for the method for determining an optimal rule base for real-time navigation of a clinical exploration robot according to the present application, it will be understood by those skilled in the art that all or part of the process for implementing the method for determining an optimal rule base for real-time navigation of a clinical exploration robot according to the embodiments of the present application may be implemented by controlling related hardware through a computer program, where the computer program may be stored in a computer readable storage medium, such as a memory of a wearable device, and executed by at least one processor in the wearable device, and the execution process may include the process of the embodiment of the method for determining an optimal rule base for real-time navigation of a clinical exploration robot. The storage medium may be a magnetic disk, an optical disk, a Read Only Memory (ROM), a random access Memory (RAM, random Access Memory), or the like.

The method for determining the optimal rule base for real-time navigation of the clinical exploration robot is described in detail. The principles and embodiments of the present application are described herein with specific examples, the above examples being provided only to assist in understanding the methods of the present application and their core ideas; meanwhile, those skilled in the art will have variations in the specific embodiments and application scope in light of the ideas of the present application, and the present description should not be construed as limiting the present application in view of the above.

Claims

1. A method for determining an optimal rule base for real-time navigation of a clinical exploration robot, the method comprising:

2. The method for determining an optimal rule base for real-time navigation of a clinical exploration robot according to claim 1, wherein said obtaining distance parameters of the robot from obstacle and non-obstacle boundaries in multiple directions by a radar ranging sensor of the robot itself comprises:

3. The method for determining an optimal rule base for real-time navigation of a clinical exploration robot according to claim 2, wherein the robot is a spherical robot, and a centroid of the robot is equipped with the radar ranging sensor.

4. The method for determining an optimal rule base for real-time navigation of a clinical exploration robot according to claim 3, wherein the acquiring, by a radar ranging sensor of the robot itself, the shortest distances between the robot and the obstacle in a plurality of first preset directions to obtain a plurality of first shortest distances, and acquiring the shortest distances between the robot and the non-obstacle boundary in a plurality of second preset directions to obtain a plurality of second shortest distances includes:

5. The method for determining an optimal rule base for real-time navigation of a clinical exploring robot according to claim 1, wherein the step of using the distance parameter as an input variable for right-to-the-robot travel and obstacle avoidance, and performing a blurring process on the input variable, and generating a final linear velocity and a final angular velocity for right-to-the-robot travel and obstacle avoidance by a fuzzy inference engine and an anti-fuzzy operation carried by the robot according to the rule base comprises the steps of:

6. The method of claim 5, wherein generating a final linear velocity and a final angular velocity of the robot right-to-travel and obstacle avoidance based on the activation energy, linear velocity, and angular velocity comprises:

7. The method for determining an optimal rule base for real-time navigation of a clinical exploration robot according to claim 1, wherein the determining whether the number of genetic iterations exceeds a preset evolution algebra further comprises:

8. The method of determining an optimal rule base for real-time navigation of a clinical exploration robot of claim 1, further comprising:

9. A determination device of an optimal rule base for real-time navigation of a clinical exploration robot, characterized in that the device comprises a processor and a memory, wherein the memory stores a computer program, and the processor is used for executing the determination method of an optimal rule base for real-time navigation of a clinical exploration robot according to any one of claims 1 to 7 by calling the computer program stored in the memory.

10. A computer-readable storage medium, characterized in that the computer-readable storage medium has stored therein a computer program which, when run on a computer, causes the computer to perform the method for determining an optimal rule base for real-time navigation of a clinical exploration robot according to any one of claims 1 to 7.