CN105426986A - High reliability control method and high reliability control system in multi-robot system - Google Patents

Abstract

The present invention provides a high-reliability control method and system in a multi-robot system. The method includes: robot topology design, fault prediction design, and fault recovery design, wherein: robot topology design: robot topology includes working robots The working group, and the standby working group composed of spare robots, the position between the standby robot and the working robot can be replaced; failure prediction design: combined with the health status of the robots in the system, the Markov model prediction method is used to predict the failure of the working robot ; Fault recovery design: When a working robot fails, control the standby robot to replace the working robot. The results of the simulation system show that the present invention can successfully coordinate and complete the production and processing tasks of the device in both the no-fault situation and the fault situation. When a certain robot in the system is in a fault state, the backup robot is allowed to take over the faulty robot to complete the operation, realizing the processing and production of the device. seamless connection.

Description

High reliability control method and system in multi-robot system

technical field

The present invention relates to the field of automatic control, in particular, to a high-reliability control method and system in a multi-robot system, which is applied to a multi-robot system, and through fault prediction, topological arrangement, and troubleshooting control of a multi-robot system, Realize high-reliability control of multi-robot systems.

Background technique

Since the 1970s, industrial robots have become an important technology in the industrial field. Worldwide, the application of industrial robots has set off a climax. With the continuous increase of labor costs and the rising labor costs of enterprises, industrial robots have objective development needs. At the same time, the application fields of industrial robots have gradually expanded from the automotive industry to electronics manufacturing, food and drug and plastic industries. In order to complete complex processes, industrial robots work together in many cases. Therefore, industrial robots often work in the form of multi-robot systems.

Multiple robots in industrial production usually work in series on the assembly line and cooperate with each other to complete production tasks. During the entire production process, if a certain robot fails, the entire production line will be paralyzed. In the high-speed manufacturing industry, the failure of the production line will undoubtedly cause the reduction of production efficiency and the loss of economic benefits.

At present, there is no relatively good control method to improve reliability in multi-robot systems in industrial production. High-reliability control technology essentially belongs to the category of fault-tolerant control, but it is difficult to achieve a fault-tolerant controller in the strict sense, especially in industrial robot systems. There are two reasons for this. First, industrial robots are generally standardized products, and fault-tolerant control of a single robot cannot be implemented. Second, the organizational structure of a multi-robot system is complex and changeable. Even if a high-quality fault-tolerant controller is designed, It is also difficult to meet the robustness requirement. A more reasonable approach is to use a compromise between cost and complexity.

The pure fault-tolerant control method should have more reliable theoretical performance. By identifying the faulty system and adjusting the control rate, high-reliability control can be achieved very economically. This fault-tolerant control method uses system components as units to identify and remodel, and the elastic variable control rate actually indirectly improves the life and reliability of system components. In many industrial robot application scenarios, the overall reliability of a multi-robot system is often more important than the service life of components, because the replacement cost of robot components is lower than the failure cost of the system. Therefore, it is a good solution to use robots as units for high-reliability control.

Contents of the invention

Aiming at the defects in the prior art, the object of the present invention is to provide a highly reliable control method and system in a multi-robot system controlled by robots.

To achieve the above object, the present invention adopts the following technical solutions:

According to the first aspect of the present invention, a high reliability control method in a multi-robot system is provided, the method comprising:

Robot topology design: The robot topology includes a working group composed of working robots and a standby working group composed of standby robots. The position between the standby robot and the working robot can be replaced;

Fault prediction design: Combined with the health status of the robot in the system, use the Markov model prediction method to predict the fault of the working robot;

Failure recovery design: When a working robot fails, control the standby robot to replace the working robot.

Preferably, the robot topology design, wherein the working group is a group of multiple robots (working robots or standby robots) with the same hardware structure and close physical locations, the control programs of multiple robots can be the same or different , each robot in the working group is physically replaceable, and the software program can be changed according to needs. At the same time, the working position between the standby robot and the working robot can be replaced, and when the standby robot is replaced to the faulty station, it can be fully automatically realized without other intervention.

Preferably, the fault prediction design, wherein the health status of the robot is implemented by fault diagnosis (system identification), that is, the health status of the robot in the system is evaluated and classified, the output information is the health rating of the robot, and the health status of the robot is defined situation. Health status can be set according to actual needs.

More preferably, the health rating is five discrete levels, and the health status of the robot is defined as the following five states: normal state, slightly degraded state, moderately degraded state, highly degraded state, and faulty state.

Preferably, the failure prediction design, wherein the Markov model prediction method is used to predict the failure of the working robot, is as follows:

S1, Prediction object state division: Corresponding to the health status of the robot, the five states of "normal state", "slightly degraded state", "moderately degraded state", "highly degraded state" and "faulty state" are taken as Marr The object state of the Cove model;

S2, calculate the initial probability p _i :

For a new robot that passes the performance test, it is considered to be in a "normal state", the initial probability of "normal state" is approximately equal to 1, and the initial probability vector is {1,0,0,0,0};

S3, calculate the down transition probability p _ij of each state

Use the frequency of mutual transition between states to describe its probability approximately, and obtain the state transition probability matrix;

S4, predicting according to the transition probability matrix and the initial probability.

More preferably, in S3, the transition probability matrix is calculated using the EM (Expectation Maximization) algorithm in the Markov model.

More preferably, in S4, when realizing the state prediction, the historical "health status" data of the robot is connected with five possible states, and five state sequences are obtained as a result, and the state sequences are brought into the Markov model to calculate the likelihood Then compare the size of several probabilities, and then select the end state corresponding to the sequence with the highest probability, that is, the prediction state, and this end state is the prediction result.

Preferably, in the fault recovery design, the working robot and the standby robot are equipped with their own controllers, and when working normally, the working robot on the station communicates with the main controller to complete the production operation, and when the working robot on a certain station When the robot fails, the heavy-duty controller of the standby robot replaces the faulty robot and communicates with the main controller to complete the processing task of the workpiece.

More preferably, the main controller and the robots on each station are each equipped with a transmitter and a receiver, and the standby robot is equipped with a receiver; the initial value of the main controller transmitter channel is the first value, and the initial value of the receiver channel is The value is the second value; the initial value of the robot transmitter channel on each station is the second value, the initial value of the receiver channel is the first value; the initial value of the standby robot receiver channel is the second value;

Communication between multi-robots when there is no failure: When the working robots on the station can work normally, the communication is concentrated between the main controller and the working robots on each station, the transmitter of the main controller and the receiver of all working robots Both establish communication through the same first numerical channel; when the sensor of the main controller detects the arrival of workpieces on the conveyor belt, the main controller sends a message through the transmitter to inform the working robots that they are ready to start their respective tasks;

Communication when a working robot fails: When a working robot on the station fails, the working robot first transmits the second numerical channel to the main controller and the receiver of the standby robot through the second numerical channel of the transmitter. Send a message to notify it that a fault has occurred; after receiving the message, the standby robot moves to the faulty robot station and completes the adjustment of the initial pose; after the main controller receives the message, it pauses for a period of time so that the standby robot can reach the faulty station, And the adjustment of the initial pose is completed; then, the standby robot sets the value of the receiver channel to the first value, and when the workpiece passes by next time, the standby robot can receive the message sent by the main controller through the first value channel; finally, the fault The receiver channel of the robot is set as an idle channel not used by the system, and the communication between the main controller and the faulty robot is disconnected;

Multi-robot communication when the standby robot is working: After the adjustment is completed, the transmitter and receiver channels of each robot are as follows: the transmitter channel value of the main controller is the first value, and the receiver channel value is the second value; the transmitter channel value of the faulty robot is the second value, the value of the receiver channel is the set value; the value of the transmitter channel of other working robots is the second value, and the value of the receiver channel is the first value; the value of the receiver channel of the spare robot replacing the faulty robot is the first value; At this time, the transmitter of the main controller establishes communication with the standby robot and the receivers of other working robots through the same first numerical channel, while the faulty robot disconnects the communication connection with other robots.

According to a second aspect of the present invention, a high reliability control system in a multi-robot system is provided, the system comprising:

Robot topology system: including a working group composed of working robots and a standby working group composed of standby robots, the position between the standby robot and the working robot can be replaced;

Fault prediction system: Combined with the health status of the robots in the system, the Markov model prediction method is used to predict the faults of the working robots;

Fault recovery system: According to the prediction results of the fault prediction system, when a working robot fails, the backup robot that controls the robot topology system replaces the working robot.

Compared with the prior art, the present invention has the following beneficial effects:

The present invention starts from the design of the controlled system itself to realize high reliability control: the division of robot working groups is adopted, and the working groups are composed of some physically replaceable individual robots; the division of working groups can reduce the reliability of the system significantly. Under the premise of reducing costs.

The fault prediction design of the present invention can avoid the production of defective products, workpiece damage and even safety accidents. The fault prediction technology can realize that robot faults occur according to "known time", and then fault recovery operations can occur in advance. At the same time, fault prediction can realize flexible reliability control, and the adjustment of the stability margin of the robot system can be realized by changing the definition of the fault state, so as to realize the dynamic control of product quality; further, Markov is used in the fault prediction of the present invention. Prediction method, this is generally more natural thinking.

In the fault recovery design of the present invention, through the design of the internal communication mode and interaction process of the robot working group, it can meet the change of the system working mode of the robot in the normal state, fault state and the whole process after fault recovery.

The method of the present invention is not limited to a certain robot system, nor is it limited to a specific application scenario. The result of the multi-robot system simulation carried out by using the invention shows that the production and processing tasks of devices can be successfully coordinated and completed in both the no-fault situation and the fault situation, and relatively satisfactory results are obtained.

Description of drawings

Other characteristics, objects and advantages of the present invention will become more apparent by reading the detailed description of non-limiting embodiments made with reference to the following drawings:

Fig. 1 is the flowchart of fault prediction in an embodiment of the present invention;

Fig. 2 is a schematic diagram of communication among multi-robots when there is no fault in an embodiment of the present invention;

Fig. 3 is a schematic diagram of communication when a station robot fails in an embodiment of the present invention;

4 is a schematic diagram of communication between multiple robots when the standby robot is working in an embodiment of the present invention;

Fig. 5 is a flow chart of master controller control in an embodiment of the present invention;

Fig. 6 is a block diagram of the system structure in an embodiment of the present invention.

detailed description

The present invention will be described in detail below in conjunction with specific embodiments. The following examples will help those skilled in the art to further understand the present invention, but do not limit the present invention in any form. It should be noted that those skilled in the art can make several modifications and improvements without departing from the concept of the present invention. These all belong to the protection scope of the present invention.

The present invention is mainly a high-reliability control technology for multi-robot systems. Firstly, it proposes multi-robot system working group division and physical topology design for the controlled system, which is the root of high-reliability systems. After the robot has a physical structure, it needs to be controlled by a controller. Based on the above idea, the present invention proposes a high-reliability control method and system in a multi-robot system.

A high reliability control method in a multi-robot system, the method comprising:

Robot topology design: The robot topology includes a working group composed of working robots and a standby working group composed of standby robots. The position between the standby robot and the working robot can be replaced;

Fault prediction design: Combined with the health status of the robot in the system, use the Markov model prediction method to predict the fault of the working robot;

Failure recovery design: When a working robot fails, control the standby robot to replace the working robot.

Correspondingly, the control system matching the above method includes:

Robot topology system: including a working group composed of working robots and a standby working group composed of standby robots, the position between the standby robot and the working robot can be replaced;

Fault prediction system: Combined with the health status of the robots in the system, the Markov model prediction method is used to predict the faults of the working robots;

Fault recovery system: According to the prediction results of the fault prediction system, when a working robot fails, the backup robot that controls the robot topology system replaces the working robot.

In order to facilitate understanding of the present invention, the details from the above three aspects are described in detail below:

1. Robot topology design:

Robot topology design is a foundation of this technique. In other words, the topology design of the robot is part of the design of the assembly line. Different from the general production line design, the topology design in the present invention emphasizes the design of the backup line. According to the basic common sense of system reliability, there is a parallel relationship between the normal line and the backup line. In system reliability analysis, the system forms a linear superposition of reliability. When the robot system is about to fail, the system switches between the normal line and the faulty line, which can ensure the recovery of the system in a short time. In addition, the restoration of faulty lines is also a part of the present invention. It is undoubtedly very important that the regular repair of the fault line can make the system realize the normal operation around the clock.

At the same time, the robot topology design must also consider the cost of the system. If each robot is designed with a backup individual, the total cost of the robotic system is almost doubled. This result is unacceptable in most cases. Therefore, the present invention proposes a concept of a working group and a standby working group, the working group is composed of working robots, and the standby working group is composed of standby robots. Each working group is a group of multiple robots with the same hardware structure and close physical locations. Of course, their control programs can be the same or different. That is to say, the robots in the working group are physically replaceable, while the software programs can be changed according to requirements. The second level of "physically replaceable" mentioned here means that the working position between the standby robot and the working robot can be replaced. When the backup robot is replaced to the faulty station, it can be realized completely automatically without other intervention.

2. Fault prediction design:

In order to realize the high-reliability control of multi-robots, the present invention further designs a fault prediction method on the basis of the above-mentioned hardware improvements. The practical significance of fault prediction is to ensure that the robots in the system have a certain stability margin. During the life cycle of a robot, its health is in a process of change. And its changing law is a statistical law. If failure prediction is not carried out, defective products, workpiece damage and even safety accidents may occur. This is undoubtedly a technique worthy of attention in this method. Fault prediction technology is based on fault diagnosis (or system identification). Fault diagnosis (system identification) is the assessment and classification of the health status of the robots in the system. The output of the system identification process is the health rating of the robot. The health rating is five discrete levels, and the present invention defines the "health status" of the robot as the following states: normal state, slightly degraded state, moderately degraded state, highly degraded state, and faulty state. Of course, these states can also be adjusted or changed according to actual needs, which does not affect the realization of the object of the present invention.

The invention uses the Markov model prediction method to realize the failure prediction of the robot. Markov model forecasting is a method of using probability to establish a random time series model for forecasting. Markov prediction method is an application of Markov process and Markov chain in the field of prediction. This method predicts the future state of things by classifying the state of things, studying the initial probability of each state and the transition probability between states. Change trends to predict the future of things.

If both time and state parameters are discrete Markov processes with no aftereffect, this random process is a Markov chain. No aftereffect can be specifically expressed as if the time parameter t _s of the random variable sequence {Y(t), t∈T} is taken as "now", then t>t _s means "future", and t<t _s means "past ", then, under the condition that the current situation Y(t _s ) of the system is known, the situation of Y(t) at the next moment in the "future" has nothing to do with the situation in the "past". This characteristic of the random process is called without aftereffect.

Forecasting model: S ^(k+1) = S ^(k) P where: S ^(k) is the state vector of the predicted object at t=k time; P is the one-step transition probability matrix; S ^(k+1) is the predicted object at The state vector at t=k+1, the predicted result.

According to the above prediction model, it can be obtained: S ⁽¹⁾ = S ⁽⁰⁾ P

S ⁽²⁾ = S ⁽¹⁾ P

S ^(k+1) = S ⁽⁰⁾ P ^(k+1)

Prediction model: where: S ⁽⁰⁾ is the initial state vector of the prediction object, which is a vector composed of the initial probability of the state. For the Markov chain, its probability at any time t can be determined by the initial probability initial state vector and one-step transition probability.

Applicable conditions: The prediction model is only applicable to time series with Markov properties. During the forecast period, the state transition probability at each moment remains stable, and they are all one-step transition probabilities. If the state transition probability of time series changes with different moments, this method should not be used. This method is generally suitable for short-term forecasting.

The state transition probability matrix P comprehensively describes the relationship between the changes of the predicted object in each state, and plays an important role in the prediction. It not only determines the state of the predicted object, but also determines the change trend and final result of the predicted object.

The steps of the Markov model prediction method are as follows:

1), prediction object state division

Predicting the state of an object has the same meaning as the "health" of the robot described above. In this step, we take the five states of "normal state", "slightly degraded state", "moderately degraded state", "highly degraded state" and "faulty state" as the object states of the Markov model.

2), calculate the initial probability p _i

The initial probability refers to the probability of a state appearing. When the theoretical distribution of the state probability is unknown, if the sample size is large enough, the theoretical distribution of the state can be approximately described by the sample distribution. Therefore, the frequency of state occurrence can be used to approximate the probability of state occurrence. Assuming that the forecast object has E _i (i=1,2,...,n) states, in the known historical data, the number of occurrences of E _i state is M _i ; then the frequency of E _i occurrence F _i =M _i /N . In the application scenario of robot work, a new robot that passes the performance test is usually considered to be in a "normal state", which is a very reasonable setting. It is equivalent to that among the five states of the predicted object, the initial probability of "normal state" is approximately equal to 1, that is, the new robot cannot be in other states at the initial moment. So the initial probability vector is {1,0,0,0,0}.

3) Calculate the down transition probability p _ij of each state

Like the initial probability of the state, the theoretical distribution of the state transition probability is unknown. When the sample size is large enough, the probability can also be approximately described by the frequency of mutual transition between states. Assuming that the number of transitions from state E _i to E _j is M _i , then

p _ij ＝P(E _i →E _j )＝P(E _j |E _i )≈M _ij /M _i (i=1,2,L,n)(j=1,2,L,n)

One-step transition probability matrix (state transition probability matrix) is obtained:

$\mathrm{<span\; class="notranslate">\; P</span>}<span\; class="notranslate">\; =</span>\left[\begin{array}{cccc}{\mathrm{<span\; class="notranslate">\; p</span>}}_{\mathrm{<span\; class="notranslate">\; 11</span>}}& {\mathrm{<span\; class="notranslate">\; p</span>}}_{\mathrm{<span\; class="notranslate">\; 12</span>}}& {\mathrm{<span\; class="notranslate">\; p</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}\mathrm{<span\; class="notranslate">\; j</span>}}& {\mathrm{<span\; class="notranslate">\; p</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}\mathrm{<span\; class="notranslate">\; no</span>}}\\ {\mathrm{<span\; class="notranslate">\; p</span>}}_{\mathrm{<span\; class="notranslate">\; twenty\; one</span>}}& {\mathrm{<span\; class="notranslate">\; p</span>}}_{\mathrm{<span\; class="notranslate">\; twenty\; two</span>}}& {\mathrm{<span\; class="notranslate">\; p</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}\mathrm{<span\; class="notranslate">\; j</span>}}& {\mathrm{<span\; class="notranslate">\; p</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}\mathrm{<span\; class="notranslate">\; no</span>}}\\ {\mathrm{<span\; class="notranslate">\; p</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}\mathrm{<span\; class="notranslate">\; 1</span>}}& {\mathrm{<span\; class="notranslate">\; p</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}\mathrm{<span\; class="notranslate">\; 2</span>}}& {\mathrm{<span\; class="notranslate">\; p</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}\mathrm{<span\; class="notranslate">\; j</span>}}& {\mathrm{<span\; class="notranslate">\; p</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}\mathrm{<span\; class="notranslate">\; no</span>}}\\ {\mathrm{<span\; class="notranslate">\; p</span>}}_{\mathrm{<span\; class="notranslate">\; no</span>}\mathrm{<span\; class="notranslate">\; 1</span>}}& {\mathrm{<span\; class="notranslate">\; p</span>}}_{\mathrm{<span\; class="notranslate">\; no</span>}\mathrm{<span\; class="notranslate">\; 2</span>}}& {\mathrm{<span\; class="notranslate">\; p</span>}}_{\mathrm{<span\; class="notranslate">\; no</span>}\mathrm{<span\; class="notranslate">\; j</span>}}& {\mathrm{<span\; class="notranslate">\; p</span>}}_{\mathrm{<span\; class="notranslate">\; no</span>}\mathrm{<span\; class="notranslate">\; no</span>}}\end{array}\right]$

P ₁₁ , P ₂₂ , ... P _nn on the main diagonal of the matrix represent the probability that they are still in the original state after one step of transfer. Using this method to calculate the state transition probability is relatively simple, and the operation is relatively basic. However, this method also has obvious shortcomings. Simple statistical calculations ignore the order relationship between state sequences, which leads to a lot of information in the original data not being utilized. A better way is to use the EM (Expectation Maximization) algorithm in the Markov model to calculate the state transition probability matrix.

To understand the EM algorithm intuitively, it can also be regarded as a successive approximation algorithm: the parameters of the model are not known in advance, a set of parameters can be randomly selected or a certain initial parameter λ0 is roughly given in advance, and the corresponding set of parameters can be determined. The most probable state of the parameters, calculate the probability of the possible results of each training sample, and then correct the parameters by the sample in the current state, re-estimate the parameter λ, and re-determine the state of the model under the new parameters, so, by Multiple iterations, until a certain convergence condition is met, can make the parameters of the model gradually approach the real parameters. The use of EM algorithm is the key to the realization of fault prediction algorithm. For a Markov model, its key parameters mainly include the state, the initial vector (the probability of each state when the system is initialized), and the state transition probability matrix. The first two are done in 1) and 2), and only need simple calculations. The calculation of the state transition probability matrix (EM algorithm) in 3) is an iterative calculation, which is a key step in model training.

4) Prediction is made according to the state transition probability matrix and the initial probability. The final step requires calculations using historical operating data and a Markov model. The historical operation data is a state sequence, which is the data of the "health status" of the individual robot in the previous working cycle. The Markov model is essentially a probability model. Given a state sequence, its corresponding probability can be calculated according to the state transition probability matrix. When implementing state prediction, the robot's historical "health" data is concatenated with the five possible states, resulting in five state sequences. Finally, the state sequence is brought into the model to calculate the likelihood probability and compare several probabilities, and then select the end state corresponding to the sequence with the highest probability (that is, the predicted state). This end state is the predicted result. The entire fault prediction process is shown in Figure 1.

3. Fault recovery technology:

The working robot and standby robot on the station have their own controllers. When the equipment is working normally, the working robot on the station communicates with the main controller to complete the production operation; when the working robot on a certain station fails, the backup robot overload controller replaces the failed working robot and communicates with the main controller, Complete the processing task of the workpiece.

The following example introduces the communication between multiple robots and the control process of each robot.

Communication among multiple robots: The main controller and the robots on each station are equipped with transmitters and receivers, and the backup robot is equipped with transmitters. The initial value of the transmitter channel of the main controller is 0, and the initial value of the transmitter channel is 1; the initial value of the transmitter channel of each robot on each station is 1, and the initial value of the receiver channel is 0; the initial value of the standby robot receiver channel is 1 .

1) Communication between multiple robots when there is no failure

When the working robots on the stations can work normally, the communication is concentrated between the main controller and the working robots on each station. As shown in Figure 2, robot A, robot B, robot C, and robot D are working robots. The transmitter of the main controller and the receivers of robot A, robot B, robot C, and robot D all establish communication through the same channel 0. When the sensor of the main controller detects the arrival of workpieces on the conveyor belt, the main controller sends a message through the transmitter to inform robot A, robot B, robot C, and robot D that they are ready to start their respective tasks.

2) Communication when a working robot fails

When a working robot on the station breaks down (it may be assumed that robot A breaks down), at this time, robot A first sends a message to the main controller and the transmitter channel 1 of the standby robot through channel 1 of the transmitter to notify them A failure has occurred. After receiving the message, the standby robot moves to robot A station and completes the adjustment of the initial pose. After the main controller receives the message, it pauses for a period of time so that the standby robot can reach the robot A station, and completes the adjustment of the initial pose. After completion, the standby robot sets the receiver channel to 0, so that when the next workpiece passes , the message sent by the main controller through channel 0 can be received by the standby robot. Finally, set the receiver channel of robot A to an idle channel not used by the system (it may be set to channel 9), and disconnect the communication between the main controller and robot A. The whole process is shown in Figure 3.

3) Communication between multiple robots when the standby robot is working

After the adjustment is completed, the transmitter and receiver channels of each robot are as follows: the transmitter channel value of the main controller is 0, and the receiver channel value is 1; the transmitter channel value of robot A is 1, and the receiver channel value is 9; robot B, The transmitter channel value of robot C and robot D is 1, and the receiver channel value is 0; the receiver channel value of the standby robot is 0. At this time, the transmitter of the main controller and the receivers of the standby robot, robot B, robot C, and robot D all establish communication through the same channel 0, while robot A disconnects the communication connection with the rest of the robots, as shown in Figure 4.

4), the overall control process

Under the environment of Webots, the simulation experiment of the present invention is carried out, and the more important overall control process is shown in FIG. 5 .

0. Set the receiver channel to 1, which is used to receive the information sent from the malfunctioning robot; set the transmitter channel to 0, to send messages to the normal working robot.

1. The program is controlled by a loop, and the waiting unit controls the time.

2. Detect whether there is a message in the fault message queue, if there is a message in the queue, go to 3a, otherwise go to 3b.

3a. Check whether the delay is over, if not, delay for a certain time and go to 1; otherwise, it means that the standby robot has arrived at the station and the adjustment of the initial pose is completed, go to 4.

3b. Conveyor belt motion control, check the operation status, if it still cannot run, wait for the robot on the station to process the workpiece, go to 1; if it can run, go to 4.

4. Obtain the detection value of the position sensor on the conveyor belt. If the detection value is less than the threshold, it means that the position sensor has not detected the arrival of the workpiece, and go to 1; otherwise, go to 5.

5. Check whether the message is sent, if it is sent, go to 1; otherwise, send the message and record it, and go to 1 at the same time.

The results of the simulation system show that the production and processing tasks of the device can be successfully coordinated and completed in the case of no failure and failure. When a robot in the system fails, let the backup robot take over the faulty robot to complete the operation, realizing the seamless processing and production of the device. butt.

Specific embodiments of the present invention have been described above. It should be understood that the present invention is not limited to the specific embodiments described above, and those skilled in the art may make various changes or modifications within the scope of the claims, which do not affect the essence of the present invention.

Claims (10)

1. the High-reliability Control method in multi-robot system, is characterized in that, described method comprises:

Robot topology design: robot topology comprises the working group be made up of working robot, and the working group for subsequent use be made up of standby machine people, and the position between standby machine people and working robot is replaceable;

Failure prediction designs: the health status of robot in coupling system, uses Markov model predicted method to carry out failure prediction to working robot;

Fault recovery designs: when certain working robot is broken down, and controls standby machine people and substitutes this working robot.

2. the High-reliability Control method in multi-robot system according to claim 1, it is characterized in that, described robot topology design, wherein working group is one group of multiple robot that hardware configuration is identical, physical location is close, the control program of multiple robot is identical or different, each robot in working group is that physics is interchangeable, and software program can change according to demand; Working position simultaneously between standby machine people and working robot is replaceable, and standby machine people can realize the working robot replacing failure stations automatically, does not need other intervention.

3. the High-reliability Control method in multi-robot system according to claim 1, it is characterized in that, described failure prediction design, wherein the health status of robot adopts fault diagnosis to realize, namely the health status of the robot in system assessed and classify, output information is the health grading of robot, and defines the health status of robot.

4. the High-reliability Control method in multi-robot system according to claim 3, it is characterized in that, described health grading is five discrete grades, and the health status of robot is defined as following five states: normal state, slight degenerate state, gently degraded state, high degradation state, fault case.

5. the High-reliability Control method in multi-robot system according to claim 4, is characterized in that, described failure prediction design, wherein uses Markov model predicted method to carry out failure prediction to working robot, specific as follows:

S1, forecasting object state demarcation: corresponding to the health status of robot, by " normal state ", " slight degenerate state ", " gently degraded state ", " high degradation state ", " fault case " these five states as the Obj State of Markov model;

S2, calculates probability p _i:

For the new engine people by Performance Detection, think that it is in " normal state ", the probability of " normal state " approximates 1, and probability vector is exactly { 1,0,0,0,0};

S3, calculates the lower transition probability p of each state _ij

Describe its probability approx by the frequency of transfer mutually between state, obtain state transition probability matrix;

S4, predicts according to transition probability matrix and probability.

6. the High-reliability Control method in multi-robot system according to claim 5, is characterized in that, in S3, uses the expectation-maximization algorithm in Markov model to carry out the calculating of transition probability matrix.

7. the High-reliability Control method in multi-robot system according to claim 5, it is characterized in that, in S4, when realizing status predication, robot history " health status " data and five possible states are coupled together, and result obtains five status switches, status switch is brought in Markov model into the size calculating likelihood probability more several probability respectively, then the end state that the sequence pair that select probability is maximum is answered and predicted state, namely this end state predicts the outcome.

8. the High-reliability Control method in the multi-robot system according to any one of claim 1-7, it is characterized in that, described fault recovery design, wherein working robot, standby machine people are equipped with the controller of self, working robot when working properly on station and master controller carry out communication and complete production operation, when the working robot on certain station is broken down, standby machine people heavy duty controller substitutes failed machines people and master controller communication, completes the processing tasks to workpiece.

9. the High-reliability Control method in multi-robot system according to claim 8, is characterized in that, the robot on described master controller and each station is all equipped with transmitter and receiver separately, and standby machine people is equipped with receiver; Master controller transmitter channel initial value is the first numerical value, and receiver channel initial value is second value; Robot transmitter channel initial value on each station is second value, and receiver channel initial value is the first numerical value; Standby machine people receiver channel initial value is second value;

Communication between multirobot during non-fault: when the working robot on station can normally work, communication concentrates between the working robot on master controller and each station, and the transmitter of master controller and the receiver of all working robot are all by the first identical numerical value Path Setup communication; When the sensor of master controller detect travelling belt has workpiece to arrive time, master controller sends message by transmitter, informs that working robot prepares to start respective task;

Communication when certain working robot breaks down: when a certain working robot on station breaks down, now first this working robot sends message by the second value passage of transmitter to the receiver second value passage of master controller and standby machine people, notifies that it breaks down; Move toward the artificial displacement of failed machines after standby machine people receives message, and complete the adjustment of initial pose; After master controller receives message, time-out a period of time makes standby machine people arrive failure stations, and completes the adjustment of initial pose; Then, receiver channel value is set to the first numerical value by standby machine people, and when workpiece is through out-of-date next time, the message standby machine people that master controller is sent by the first numerical value passage just can receive; Finally the receiver channel of failed machines people is set to the idle channel that system does not use, disconnects the communication between master controller and failed machines people;

Communication between standby machine man-hour multirobot: after completing adjustment each robot transmitter and receiver channel as follows: master controller transmitter channel value is the first numerical value, and receiver channel value is second value; The transmitter channel value of failed machines people is second value, and receiver channel value is setting value; Other working robot's transmitter channel values are second value, and receiver channel value is the first numerical value; The standby machine people receiver channel value of replacing failed machines people is the first numerical value; Now the transmitter of master controller and the receiver of standby machine people, other working robots are all by the first identical numerical value Path Setup communication, and failed machines people and all the other robots then disconnect communication and are connected.

10. for realizing the High-reliability Control System in the multi-robot system of method described in any one of the claims 1-9, it is characterized in that, comprising:

Robot topological system: comprise the working group be made up of working robot, and the working group for subsequent use be made up of standby machine people, the position between standby machine people and working robot is replaceable;

Failure prediction system: the health status of robot in coupling system, uses Markov model predicted method to carry out failure prediction to working robot;

Fault recovery system: predicting the outcome according to failure prediction system, when certain working robot is broken down, the standby machine people of control topological system substitutes this working robot.