CN108910630B - Elevator group control system and method based on multi-agent competition mode - Google Patents

Abstract

The invention discloses a multi-agent competition mode elevator group control system and method. Each elevator control system is an autonomous agent, and the agent that obtains the call signal generates bidding information according to a predetermined agreement and sends it to the coordination range. For other agents, each agent bids against the bidding information, and the agent with the best bid provides services for passengers; each agent includes: a user interface unit for interacting with passengers; a communication unit for interacting with the scope of collaboration The bidding unit is used to generate all the information in the bidding process and bid with other agents within the coordination range; the control unit is used to run the plan and execute the service according to the call form after winning the bid. The present invention adopts a non-centralized control mode that does not rely on additional group control equipment, and changes the unified scheduling of multiple elevators into coordinated services of multiple elevators through a bidding mechanism, thereby improving the reliability of the elevator group control system and reducing the requirements for software and hardware.

Description

Elevator group control system and method based on multi-agent competition

technical field

The invention is related to the dispatching technology of multiple elevators, and in particular relates to an elevator group control system and method based on a multi-agent competition mode.

Background technique

The elevator group control system enables multiple elevators within the dispatching range to operate in coordination through a reasonable dispatch method, thereby improving the overall service efficiency of the elevator group control system. In recent years, more energy-saving and more comfortable dispatching requirements have been generated.

The common implementation method of elevator group control system is to add a group control unit with dispatching function. The group control unit communicates with each elevator within the dispatching range. Each elevator sends group control dispatch instructions, which is collectively referred to as centralized control mode. In the actual operation process, once the group control unit fails, the entire group control system will not be able to run, and the elevator scheduling will be paralyzed. This is a fatal problem of the elevator group control system in the centralized control mode.

In order to solve this problem, the industry has proposed some new improvement schemes, which are mainly divided into the following two types:

(1) The scheme of multi-machine redundancy, that is, a working group control unit and one or more standby group control units are set up in the elevator group control system. Immediately switch to the standby group control unit;

(2) The scheme of semi-centralized control mode, that is to cancel the group control unit, so that each elevator has the group control function. At a certain time, one elevator is used as the group control unit. Once the current group control unit is found to be faulty, the remaining elevators will be One of the other elevators acts as a group control unit according to a pre-agreed priority. For example, CN201210381192.6, CN201610117907.5, CN201610119384.8 and other invention patents adopt this kind of scheme.

Among the above improvement solutions, the first solution will obviously increase the equipment cost, and in order to ensure the high reliability of the elevator system, it is necessary to select multi-level backup, which will further increase the cost, and the group control unit will perform hot switching in the event of a failure. There are also certain technical difficulties. Although the second solution avoids the cost of additional backup, as the group control unit of the elevator group control system, it needs to undertake scheduling responsibilities, which leads to an increase in the requirements for software and hardware and an increase in additional costs. The connection method between them will also become complicated, and the fault monitoring and timely switching when a fault occurs is still complicated.

In the above two improvement schemes, the elevator group control system still belongs to the centralized control mode in essence, and the equipment acting as the group control unit completes the communication with all elevators and obtains its status for analysis, so as to make scheduling decisions, and its workload is relatively large. , the hardware requirements are still high, and the attendant cost is also relatively high.

SUMMARY OF THE INVENTION

The technical problem to be solved by the present invention is to provide an elevator group control system and method based on a multi-agent competition mode, which can solve the problems of low reliability and high requirements on software and hardware of the existing elevator group control system.

In order to solve the above technical problems, the present invention provides a multi-agent competition mode elevator group control system, each elevator control system in the elevator group control system is defined as an autonomous agent, and the agent that obtains the call signal is based on According to the predetermined agreement, bidding information is generated and sent to other agents within the coordination range. Each agent responds to the bidding information and bids, and finally the agent with the best bid obtains the service authority to provide services for passengers;

Each agent includes a user interface unit, a bidding unit, a communication unit, and a control unit, where:

User interface unit for interacting with passengers;

The communication unit is used to transmit information with other agents within the cooperative range;

Bidding unit, which is used to generate all the information in the bidding process and bid with other agents within the cooperative scope;

The control unit is the service execution unit. After the elevator wins the bid, it will run the plan and execute the service according to the call form.

More preferably, the agent further includes a self-learning unit, which performs self-learning according to all the data in the bidding process and the service process and improves the bidding method of the bidding unit.

In the above system, the user interface unit and the communication unit perform bidirectional transmission of information, the bidding unit and the user interface unit and the self-learning unit both perform bidirectional transmission of information, the bidding unit transmits information to the control unit, and the bidding unit Together with the control unit, it transmits information to the self-learning unit.

The present invention also provides an elevator group control method based on a multi-agent competition mode. Each elevator control system in the elevator group control system is defined as an autonomous agent, and the agent that obtains the call signal generates a The bidding information is sent to other agents within the coordination range, and each agent responds to the bidding information and bids, and finally the agent with the best bid obtains the service authority to provide services for passengers.

In the above method, the trigger condition for the agent to enter the bidding process is that the passenger sends an elevator call signal or reservation signal through the car-hailing device or reservation device connected to the agent, or the agent detects that there is new bidding information on the network where it is located.

Further, the passenger's call signal or reservation signal is transmitted to the user interface unit of the agent in real time through the communication unit of the agent; after the user interface unit receives the call signal, it makes standard service request information according to the prescribed protocol, and sends the information to the user interface. The service request information is transmitted to the bidding unit of the agent; after receiving the service request information, the bidding unit calculates the bidding price according to the service request information and generates bidding information, and then transfers the bidding information to the communication unit via the user interface unit ; After receiving the bidding information, the communication unit sends the bidding information to the network according to the pre-agreed agreement.

Further, the communication unit of the agent always monitors the network where it is located, and after receiving the bidding information, it transmits it to the bidding unit of the agent through the user interface unit of the agent; the bidding unit calculates the bidding price according to the bidding information and generates bidding information, The bidding information is then transferred to the communication unit via the user interface unit; after receiving the bidding information, the communication unit sends the bidding information to the network according to a pre-agreed agreement. Specific steps are as follows:

In the first step, when the communication unit of a certain agent receives the bidding information and transmits it to the bidding unit, the bidding unit judges whether it has the service capability for the bidding information, that is, whether it meets the bidding requirements, and if so, enters the The second step, otherwise, set the bid price to the agreed worst price, and go to the seventh step;

In the second step, the bidding unit generates a bid price according to its own conditions and according to the set bid price algorithm;

In the third step, the bidding unit compares the generated bid price with the current optimal price in the bidding information. If it is better than the current optimal price, it will go to the fourth step, otherwise, it will go to the fifth step;

The fourth step, the bidding unit replaces the bidder with the current best price and the best price, and enters the seventh step;

The fifth step, the bidding unit compares the generated bid price with the current alternate price in the bidding information, and if it is better than the current alternate price, the sixth step is entered;

In the sixth step, the bidding unit replaces the bidder with the current standby price and the standby price, and enters the seventh step;

In the seventh step, the bidding unit updates the bidding information including the bidder list, and transmits the bidding information to the communication unit via the user interface unit, and the communication unit sends the bidding information of the agent to the network where it is located.

In the above method, the termination condition of the bidding process is that all agents within the coordination range have bid or the time interval between the current time and the bidding start time has exceeded the set maximum bidding period. Further, when each agent finds that the bidding meets the termination conditions, it sets the bidding status in the bidding information as the bidding end, and broadcasts it to all agents within the coordination range. Specific steps are as follows:

In the first step, the bidding unit of the agent compares the bidder list in the bidding information with the member list. If all agents have bid, go to the fourth step, otherwise go to the second step;

In the second step, the bidding unit calculates the elapsed time of bidding, that is, the time interval between the current time and the start time of bidding;

In the third step, the bidding unit judges whether the elapsed time for bidding exceeds the longest bidding period, and if so, enters the fourth step; otherwise, the bidding information is made and sent to the network where it is located;

In the fourth step, the bidding unit sets the bidding status in the bidding information as the bidding end sign, and transmits the bidding end information to the communication unit through the user interface unit, and the communication unit sends it to all the agents within the coordination range.

In the above method, after the bidding process is over, the agent confirms whether it is the winning bidder. If it is the winning bidder, the bidding unit of the agent notifies the control unit of the agent of the winning information, and at the same time generates the acceptance information and sends it to the cooperative scope. Each online agent of ; if it is not the winning bidder, further confirm whether it is a candidate winning bidder, if so, start timing and wait for the winning bidder to release the acceptance information, if the bidding unit of the agent fails to receive the bid within the specified time In the case of the successful bidder's acceptance information, it will switch itself to the successful bidder and notify the control unit of the successful bid information, and at the same time, the acceptance information will be released, and the service will be clearly recorded by the candidate successful bidder.

Further, after the control unit of the agent receives the bid-winning information, it forms an execution plan and executes the service. At the same time, all the agents record the bid-winning information in the service registration form held by the bidding unit. , the control unit of the agent notifies the bidding unit, and the bidding unit generates service completion information and sends it to all agents via the user interface unit and the communication unit; the service registration form is used to record all bidding in the elevator group control system information, the ultimate recipient.

Among them, the bidding unit of the agent traverses the service registration table according to a specific period. When the bidding information that can be re-bid is found in the service registration table, the bidding unit generates the re-bidding information, and requests the communication unit via the user interface unit. The undertaker sends the re-bid information; after the agent as the current undertaker receives the re-bid information, its bidding unit re-evaluates the service. If the bid price is still better than the bid price of the re-bidder, the result is to maintain the status quo. Otherwise, the re-bidder wins the bid, and the bidding unit of the current undertaker generates re-bid information according to the result and transmits it to the re-bidder.

Among them, when an agent wins the bid but is unable to perform the service it undertakes, the bidding unit of the agent generates fault information and publishes it to the network where it is located. After each agent receives the fault information, the bidding unit checks the service registration form. All faulty agents are revised as the record of the winning bidder, and the winning bidder and the winning bidder are replaced by the alternate price and the bidder with the alternate price. The newly elected winning bidder notifies the control unit of the service, and the control unit lists the service plan and executes the service.

The beneficial effects of the present invention are:

1. Completely changed the existing centralized control mode or improved centralized control mode, and adopted a completely distributed mode, so the elevator group control system no longer depends on a single group control unit, avoiding the failure of the group control unit. The group control fails. In essence, the group control mode is changed from the scheduling work under limited resources to the cooperative work of independent individual competition tasks, and the scalability of the system is greatly improved;

In the prior art, the number of consoles in a group control system is mainly limited by the number of hardware interfaces and computing capabilities of the group control unit. In the technical solution of the present invention, as long as the network access allows, the group members can be added arbitrarily. There will be no increase in the computational load of the members;

2. Work collaboratively in the form of bidding competition for service rights. At the same time, in order to solve some special specification calls or specific service requirements, a bidding qualification system is set up. Only agents that meet the bidding requirements can bid, which is convenient for the realization of special specifications;

3. The bidding results no longer rely on a single group control unit for service distribution, but each agent decides on its own according to the agreement, and the winning bidder who has obtained the service authority announces the result according to the agreement;

4. Each agent has rules for re-determining the recipients for possible abnormal situations, and at the same time, a mechanism for secondary competition is introduced at any time, and a mechanism for replacing some unsuitable recipients according to changes in conditions makes the group control system. With error correction mechanism;

5. Due to the multi-agent competition method, the computing load is dispersed, and each agent performs independent self-learning according to the scheduling process and operation process, which is easy to implement and update, so as to gradually optimize the expected results and the actual results during bidding. ;

6. Due to the cancellation of the centralized control mode, additional group control unit equipment is no longer required, which saves equipment costs and reduces failure points; at the same time, the services are scattered in each elevator, the operating frequency is relatively scattered, and the group control unit will not be replaced frequently equipment, which is more economically advantageous.

Description of drawings

Fig. 1 is the structural block diagram of the agent in the present invention;

Fig. 2 is the networking schematic diagram of the elevator group of the present invention;

Fig. 3 is the working flow chart of the bidding unit inside the agent in the present invention;

Fig. 4 is the flow chart of bidding information generation in the present invention;

FIG. 5 is a flow chart of judging the end of bidding in the present invention.

Detailed ways

The present invention will be described in further detail below with reference to the accompanying drawings and specific embodiments.

The present invention expands its functions on the basis of the existing independent operating elevator control system, so that it has functional modules that can make corresponding decisions for the received service requests, and each elevator control system can also coordinate with other elevators. Control systems exchange information.

The service request is that the passenger calls the elevator and asks the elevator to come and serve the elevator through the operating device (it can be the landing button, or the floor selection device for the destination floor, or any device that can send a call request, such as a mobile phone) in the landing. Usually, this service request will generate a corresponding call signal (or called a car call signal). In the existing elevator group control system, the call signal will be transmitted to the group control unit through the communication unit of the elevator.

In the elevator group control system based on the multi-agent competition mode of the present invention, each elevator control system is an autonomous intelligent body, and the intelligent body serves the passengers after obtaining the service authority through the competition mode. Generally, the call signal sent by the passenger at the landing is transmitted to an elevator connected to the operating device (ie, the calling device). In the present invention, the car-calling device on each landing is connected to an elevator control system, and the elevator control system that obtains the call signal packs the call signal according to a predetermined agreement, makes bidding information, and sends it to others within the scope of coordination Elevator control system, each elevator control system responds after receiving the bidding information, bids against the call signal, and finally the agent with the best bid obtains the service authority to provide services for passengers.

Each agent includes a user interface unit (ie UI unit), a bidding unit, a communication unit, and a control unit, as shown in Figure 1, where:

The user interface unit is used to interact with passengers, mainly including accepting the passenger's call signal, providing service prompts (informing the passengers through the display screen which elevator serves them, and the status of the elevator that provides the service, such as location, direction, etc.), etc., The user interface unit is directly connected to the landing call device or the car call device and the service prompt device (such as a display screen, etc.) through a wired or wireless manner;

The communication unit is used to transmit information with other agents within the cooperative range;

The bidding unit is used to generate all the information in the bidding process, including bidding information, bidding information, bidding price, etc., and transmit it to the communication unit through the user interface unit and bid with other agents within the coordination range;

The control unit is the elevator controller in the traditional sense, which is the service execution unit. When the bid for an elevator control system as an intelligent body is successful, the control unit in the intelligent body forms an execution plan according to the call form and executes the service to serve passengers. Provide the required services.

The user interface unit here is in a logical sense, including accepting operation parts and information prompting parts, wherein the accepting operation parts include traditional landing buttons, car buttons and new mobile phone applications (APP), credit card machines and even biometric identification The terminal and other devices can essentially output call signals. The information prompt component is a device that feeds back the elevator service information to the user, including the traditional landing direction and floor display, the display screen in the car, and the mobile phone screen. The user interface unit accepts the passenger's operation and generates a call signal to the subsequent unit. During the bidding process of the subsequent unit, the user interface unit displays some currently selected services or other user-friendly information, and displays the bidding result to the passenger after the bidding result is generated. For example, when a passenger sends an elevator call signal and no bidding result is generated, the user interface unit can display the message "You have chosen to go to the N floor, the appropriate elevator is being dispatched, please wait", and the user interface unit can display the bidding result after the result is generated. Elevator service information "Elevator No. m for you is coming, please wait at the side of No. m".

As shown in FIG. 1 , the user interface unit and the communication unit perform bidirectional transmission of information, the bidding unit and the user interface unit perform bidirectional transmission of information, and the bidding unit performs information transmission to the control unit.

In the present invention, multiple elevators that work together are called elevator groups. Correspondingly, multiple elevator control systems form an elevator group control system. The communication mode of the elevator group control system can take various forms, as long as all elevator control systems are satisfied. Complete the bidding process within the specified time. For example, a wired connection (which can be a bus type, a grid type, a star type, etc.) can be used physically, or a wireless connection can be used; the communication form can be event-driven or polling. Those skilled in the art Equipment can be carried out according to actual needs. Usually, the passenger operates the call button at the landing, and the system needs to feedback the result to the passenger within a few seconds. Figure 2 shows the bus-type networking structure.

The elevator group control system responds to the passenger's call signal and feeds back the results, mainly including the bidding process and the bidding execution process. In the above process, it also involves the handling of abnormal situations such as re-tendering and failure of service providers to send. Each process is described in detail below.

In the present invention, the bidding unit of each agent can be triggered to enter the working state (ie the bidding process) by two factors, one factor is that the passenger makes a call (or reservation) through the car-calling (or reservation) device connected to the agent ( or reservation) signal, another factor is that the agent detects new bidding information on the network through the communication unit, and the relevant process is shown in Figure 3.

The specific process for the agent to enter the working state through the above-mentioned first factor is: passengers call (or reserve) the elevator through the car-calling device or reservation device in the landing, for example, using the operation button of the landing or the floor selector or mobile APP, etc. , the passenger's calling operation generates an elevator call signal and transmits it to the user interface unit of the intelligent body corresponding to the car calling device through the communication unit in real time. After receiving the call signal, the user interface unit makes standard service request information (ie, wraps the call signal) according to the prescribed protocol, and transmits the service request information to the bidding unit of the agent. After receiving the service request information, the bidding unit calculates the bid price according to the service request information and generates bidding information, and then transfers the bidding information to the communication unit via the user interface unit. After receiving the bidding information, the communication unit sends the bidding information to the network (the network where all the cooperating agents are located) according to the pre-agreed agreement.

The process of the agent entering the working state through the second factor above is similar to the process of entering the working state triggered by a call, the difference is that the communication unit of the agent always monitors the network where it is located, and transmits it to the bidding unit after receiving the bidding information. , the subsequent process is the same as the call trigger, that is, the bidding unit calculates the bid price, generates bid information and transmits it to the communication unit via the user interface unit and publishes it by the communication unit. The communication unit monitoring the network mainly has two functions: (1) If there is data on the network, receive and transmit it to the subsequent corresponding unit for processing; (2) Obtain the network sending authority to send the data. If the networking adopts a common protocol such as TCP, then conflict avoidance is the default function, and you can also refer to the realization of the conflict avoidance function. This part is already an existing mature technology for network communication, and will not be described in detail here.

The bidding information includes the bidding number, sender, call attributes (departure floor, front/back, direction, destination floor, special specifications), current best price i.e. bidder number, current standby price and bidder number, as shown in Table 1 shown, but not limited to, those skilled in the art can add or subtract relevant information as needed.

Table 1 Bidding Information Structure

The bidding of an agent is the whole process from receiving the bidding information to generating its own bidding price, that is, the link of generating the bidding information in Figure 3. No matter which factor triggers the agent to enter the working state, the bidding starts from the discovery of the bidding information. The difference is that the bidding unit of the agent is triggered by the first factor to enter the working state. The specific process is that it receives the service request information itself. After the bidding information is made, it automatically enters the bidding stage, and the bidding unit of the agent is triggered by the second factor to enter the working state. The specific process is to find the bidding information on the network, and then enter the bidding stage. The specific steps are shown in Figure 4 and described as follows:

In the first step, when the communication unit of a certain agent receives the bidding information and transmits it to the bidding unit, the bidding unit judges whether it has the service capability for the bidding information, that is, whether it meets the bidding requirements, and if so, enters the The second step, otherwise, set the bid price to the agreed worst price, and go to the seventh step;

In the second step, the bidding unit generates a bid price according to its own conditions and according to the set bid price algorithm;

In the third step, the bidding unit compares the generated bid price with the current optimal price in the bidding information. If it is better than the current optimal price, it will go to the fourth step, otherwise, it will go to the fifth step;

The fourth step, the bidding unit replaces the bidder with the current best price and the best price, and enters the seventh step;

The fifth step, the bidding unit compares the generated bid price with the current alternate price in the bidding information, and if it is better than the current alternate price, the sixth step is entered;

In the sixth step, the bidding unit replaces the bidder with the current standby price and the standby price, and enters the seventh step;

In the seventh step, the bidding unit updates the bidding information including the bidder list, and transmits the bidding information to the communication unit via the user interface unit, and the communication unit sends the bidding information of the agent to the network where it is located. In order to avoid the conflict between the bids of multiple agents in the network, standard network protocols such as TCP can be used. The network conflict is automatically resolved by the underlying protocol. If there is no network protocol, it can be re-visited by generating a random number and waiting for a time interval. Because the random numbers obtained by different agents will be different, so the waiting interval is different, and network conflicts can be avoided. The above technologies are mature means of network communication, so they will not be described in detail here.

During the bidding process, the bidding unit of each agent judges whether the bidding process should end. In the present invention, the bidding process can be terminated by two conditions, the first is that all the agents in the cooperative range have bid, and the second is that the time interval between the current time and the bidding start time has exceeded the set maximum Long bidding period.

When each agent finds that the bidding meets the termination conditions, it sets the bidding status in the bidding information to the end of the bidding, and broadcasts it to all agents within the coordination range. The process is shown in Figure 5, and the details are as follows:

In the first step, the bidding unit compares the list of bidders with the list of members stored in the bidding unit. If all agents have bid, go to the fourth step, otherwise, go to the second step;

In the second step, the bidding unit calculates the elapsed time of bidding, that is, the time interval between the current time and the start time of bidding;

In the third step, the bidding unit judges whether the elapsed time for bidding exceeds the longest bidding period, and if so, enters the fourth step; otherwise, the bidding information is made and sent to the network where it is located;

In the fourth step, the bidding unit sets the bidding status in the bidding information as the bidding end sign, and transmits the bidding end information to the communication unit through the user interface unit, and the communication unit sends it to all the agents within the coordination range.

When the bidding unit of a certain agent finds that the bidding termination condition is established, it generates bidding end information (that is, the bidding status of the bidding information is set to the bidding end) and sends it to each online agent within the coordination range. The bidder of the price is the winning bidder, and the winning bidder will serve the passengers. When the communication unit of a certain agent receives the bidding end information, it transmits the bidding end information to the bidding unit through the user interface unit, and the bidding unit confirms whether it is the winning bidder (that is, the bidding unit confirms whether the bidder number with the best price is not) Numbering for oneself), if it is the winning bidder, the bidding unit will notify the control unit of the winning bid information, and at the same time generate the acceptance information and send it to each online agent within the coordination scope; if it is not the winning bidder, further confirm whether it is the candidate winning bidder , if so, start timing and wait for the winning bidder to release the acceptance information. If the bidding unit of the agent fails to receive the successful bidder's acceptance information within the specified time, it will switch itself to the winning bidder and notify the control unit of the bid winning information. , and at the same time publish the acceptance information, and clearly indicate that the service will be undertaken by the candidate winning bidder (the candidate whose price is listed in the bidding information).

After the control unit receives the bid-winning information, it forms an execution plan and executes the service. All agents record the winning information in the service registration form held by the bidding unit. The service registration form is held and managed by the bidding unit, which is used to record all bidding information in the elevator group control system, the final undertaker, etc. information.

When an agent completes a service, the control unit notifies the bidding unit, and the bidding unit generates service completion information and sends it to all agents through the user interface unit and the communication unit. All agents receive the information and register the service from the service Delete from the table and move it to the history record (the history record is an independent area in the bidding unit, where each piece of bidding information is stored) for preservation. The service completion information mainly includes data related to the service process, such as the use of the receipt of the bidding information to the start of the service. time, service time, etc.

During the bidding process, some abnormal situations will inevitably occur, such as re-bidding, failure of service providers, etc.

Among them, during the running process, the bidding unit of the agent traverses the service registration table according to a specific cycle, looking for service requirements that are expected to be re-bid. When the bidding information that can be re-bid is found in the service registration form, the bidding unit will generate the re-bid information, and request the communication unit to send the re-bid information to the current recipient of the service via the user interface unit. After the agent communication unit, which is the current undertaker, receives the re-bidding information, it is transmitted to the bidding unit through the user interface unit, and the bidding unit re-evaluates the service. If the bid price is still better than the bid price of the re-bidder, the result is The status quo is maintained, otherwise the rebidder wins the bid, and the bidding unit generates rebidding information according to the result and transmits it to the rebidder. After receiving the re-bidding information, if the re-bidder wins, it will be announced to all agents (in the same way as the bidding results are announced), and all agents will update the service registration form.

When an agent wins the bid but cannot perform the service it undertakes, the bidding unit generates fault information and publishes it to the network through the user interface unit and the communication unit. The fault information includes the number of the fault person and the fault attribute. After the unit receives the fault information, it informs the bidding unit through the user interface unit. The bidding unit checks the service registration table, revises all the faulty agents in the table as the record of the winning bidder, and replaces the winning bidder and the winning bidder with the bidder with the alternate price and the bidder with the alternate price. , the candidate field information remains unchanged, that is, the successful bidder and the candidate for these readjusted services are the same after modification. The newly elected winning bidder (ie, the original candidate) notifies the control unit of the service, and the control unit schedules the service and executes the service. In addition, when the agent finds that the services it undertakes are not suitable to be provided by itself, it will also release the information of re-bidding, and the process is the same as normal bidding.

In the aforementioned bidding process, the bidding price of an agent for a service request is calculated according to the current state of the agent and the characteristics of the service according to a certain bidding price algorithm, and the calculation result reflects the quality of the service provided by the agent expectations. Usually, there are two situations for the bid price. The first is the waiting time index. The lower the value, the better, such as the elevator ride time, the average waiting time of passengers during elevator operation, etc. It can also be these services or waiting time indicators. The second is the reward obtained by the agent for providing services. In this case, the higher the value, the better. In this case, the party with the highest reward in the bidding process wins. Usually, some indirect indicators are used to encourage more agents. If the order is accepted and the order completion time is lower than the currently known average period of the system (all agents are tied or their own historical average), the product of the difference and the total number of services undertaken will be rewarded to the agent according to a certain proportional coefficient.

The algorithm for calculating the bid price can be an explicit function calculation process, or it can be a computational model such as a neural network (including a deep neural network). As a specific implementation solution, reference may be made to the solution published in the invention patent application with the application number of 201310080795.7.

On the basis of the above-mentioned elevator group control system, in addition to the aforementioned bidding behavior, a better feature of the intelligent behavior of the agent is that it can learn the bidding process and service process, and continuously improve its own bidding price calculation ability. Therefore, the agent also includes a self-learning unit, which can perform self-learning and improve the bidding method of the bidding unit according to all the data in the bidding process and the execution service process. As shown in Figure 1, the bidding unit and the control unit jointly Self-learning unit for information transmission.

In the present invention, the self-learning unit of each agent adopts the method of reinforcement learning to independently learn according to the bidding data and service results. In the self-learning process, the agent has two states, namely idle state and service state. The idle state means that the current running plan of the agent does not have any outstanding service requests and is in a static state, which belongs to the waiting task state. Service status means that there are unfinished service requests in the agent operation plan or although there are no unfinished service requests, but a service process is being executed. The process of entering the service state from the idle state and completing all the service plans to enter the idle state again is called an episode. After each scenario, the self-learning unit of the agent conducts self-learning on the data during the scenario.

The goal of the agent's self-learning is to make the predicted data of the bid price when bidding as consistent as possible with the actual data of the final service completion, that is, through self-learning, the predicted service quality when bidding is more and more accurate, so that the best bid price when bidding. It's the best service you can get.

The self-learning results in one scenario of the agent are temporarily stored in the self-learning unit, and the bidding price calculation rules are usually not updated immediately. After the scenarios experienced by the agent reach a certain number (for example, 50), the bidding price calculation rules of all agents are updated uniformly, that is, the bidding price calculation rules of all agents are updated synchronously with the result learned by one agent. When the self-learning unit of an agent has undergone the self-learning of a specified number of scenarios, it will send a learning notification in a broadcast mode (sent by the self-learning unit, sequentially through the bidding unit, the user interface unit, and finally broadcast by the communication unit to the network where it is located) Prepare to publish the learning results to other agents within the coordination range. After the agent receives the learning notification, it reports that the publisher is ready to receive data. After the publisher publishes the learning results, each agent in the group updates its own bidding price calculation rules, and the update is successful. After feedback to the publisher the results, the publisher records the list and version of the successful updaters. It does not matter if an agent fails to update successfully, because the update of the learning result directly overwrites the old parameters, and the intermediate stage can be skipped to update to the latest result.

Each learning update has a unique version number. For example, the version number is 0 at the beginning of the system, and increases by 1 for each update. When an agent is ready to publish the learning results, it will automatically generate a new version number based on the existing records when issuing the learning notification, and publish it to all members in the group. Each member will update the internal records and maintain the latest version number. Regardless of whether the learning result update is successful or not, the version number cannot be revoked once it is published. If the published update version is lower than the current latest version, the update is ignored.

Although in the above description, all the agents within the cooperative scope jointly update the bidding price calculation rule, the present invention is not limited to this, and the agents can also update independently. In addition, in the present invention, although all the data information generated by the bidding unit is transmitted to the communication unit through the user interface unit, and then published by the communication unit, those skilled in the art can easily make improvements according to the above principles, so that the information generated by the bidding unit Publishing directly through the communication unit, without going through the user interface unit, is hereby stated.

The invention completely changes the existing centralized control mode or improved centralized control mode, and adopts a completely distributed mode, so the elevator group control system no longer depends on a single group control unit, and avoids the failure of the group control unit. The failure of group control, in essence, changes the group control mode from the scheduling work under limited resources to the cooperative work of independent individual competing tasks, and the scalability of the system is greatly improved. In the prior art, the number of consoles in a group control system is mainly limited by the number of hardware interfaces and computing capabilities of the group control unit. In the technical solution of the present invention, as long as the network access allows, the group members can be added arbitrarily. There will not be any increase in the computational load of the members.

The present invention works cooperatively in the form of bidding competition for service rights, and meanwhile, in order to solve some special specification calls or specific service requirements, a bidding qualification system is set, and only intelligent bodies that meet the bidding requirements can bid, which facilitates the realization of special specifications; The bidding results no longer rely on a single group control unit for service allocation, but each agent decides on its own according to the agreement, and the winning bidder who has obtained the service authority announces the results according to the agreement.

At the same time, each agent of the present invention has rules for re-determining the recipients for possible abnormal situations, and at the same time, a mechanism for secondary competition is introduced at any time, and some unsuitable recipients can be replaced according to changes in conditions. The group control system has an error correction mechanism; due to the multi-agent competition method, the independent self-learning of each agent is easy to achieve, and it can learn and update independently according to its own operation process, so as to gradually optimize the prediction when bidding. The results are closer to the actual results.

Since the centralized control mode is canceled, the present invention does not need additional group control unit equipment, saves equipment costs, and reduces failure points; at the same time, the service is scattered in each elevator, the operating frequency is relatively scattered, and the group control unit will not be replaced frequently equipment, which is more economically advantageous.

The present invention has been described in detail above through specific embodiments, which are only preferred embodiments of the present invention, and are not intended to limit the present invention. Without departing from the principles of the present invention, equivalent replacements and improvements made by those skilled in the art shall be deemed to be within the technical scope protected by the present invention.

Claims (14)

1. the elevator group control system of a multi-agent competition mode is characterized in that, each elevator control system in the described elevator group control system is defined as an autonomous agent, and the agent that obtains the call signal according to predetermined Agreement, generate bidding information and send it to other agents within the scope of coordination, each agent responds to the bidding information and bids, and finally the agent with the best bid obtains the service authority to provide services for passengers; Each agent includes a user interface unit, a bidding unit, a communication unit, and a control unit, where: User interface unit for interacting with passengers; The communication unit is used to transmit information with other agents within the cooperative range; Bidding unit, which is used to generate all the information in the bidding process and bid with other agents within the cooperative scope; The control unit is the service execution unit. After the elevator wins the bid, it will run the plan and execute the service according to the call form. 2. The elevator group control system of the multi-agent competition mode according to claim 1, wherein the agent further comprises a self-learning unit, which performs self-learning and improvement according to all data in the bidding process and the service process The bidding method of the bidding unit. 3. The elevator group control system of the multi-agent competition mode according to claim 1, wherein the user interface unit and the communication unit carry out bidirectional transmission of information, and the bidding unit and the user interface unit carry out bidirectional transmission of information, The bidding unit transmits information to the control unit. 4. the elevator group control system of multi-agent competition mode according to claim 2, is characterized in that, described user interface unit and communication unit carry out bidirectional transmission of information, described bidding unit and user interface unit and self-learning unit are all The bidirectional transmission of information is performed, the bidding unit transmits information to the control unit, and the control unit transmits information to the self-learning unit. 5. An elevator group control method based on a multi-agent competition mode, wherein each elevator control system in the elevator group control system is defined as an autonomous agent, and the agent that obtains the call signal is in accordance with a predetermined agreement. , generate bidding information and send it to other agents within the scope of collaboration, and each agent responds to the bidding information and bids; The termination condition of the bidding process is that all agents within the coordination range have already bid or the time interval between the current time and the bidding start time has exceeded the set maximum bidding period, and each agent finds that the bidding meets the termination conditions. , set the bidding status in the bidding information to the end of bidding, and broadcast it to all agents within the coordination range; After the bidding process is over, the agent confirms whether it is the winning bidder. If it is the winning bidder, it obtains the service authority to provide services for the passengers; if it is not the winning bidder, it further confirms whether it is the candidate winning bidder. If so, it starts timing. , waiting for the successful bidder to release the acceptance information. If the agent fails to receive the acceptance information from the successful bidder within the specified time, it will switch itself to the successful bidder and provide services for passengers. 6 . The elevator group control method based on multi-agent competition mode according to claim 5 , wherein the triggering condition for the agent to enter the bidding process is that the passenger sends a call through a car-calling device or a reservation device connected to the agent. 7 . ladder signal or appointment signal, or the agent detects that there is new tender information on its network. 7. The elevator group control method based on multi-agent competition mode according to claim 6, is characterized in that, The passenger's call signal or reservation signal is transmitted to the user interface unit of the agent in real time through the communication unit of the agent; After receiving the call signal, the user interface unit makes standard service request information according to the prescribed protocol, and transmits the service request information to the bidding unit of the agent; After the bidding unit receives the service request information, it calculates the bid price according to the service request information and generates bidding information, and then transfers the bidding information to the communication unit via the user interface unit; After receiving the bidding information, the communication unit sends the bidding information to the network according to the pre-agreed agreement. 8. The elevator group control method based on multi-agent competition mode according to claim 6, is characterized in that, The communication unit of the agent always monitors the network where it is located, and when the bidding information is received, it is transmitted to the bidding unit of the agent through the user interface unit of the agent; The bidding unit calculates the bidding price according to the bidding information and generates bidding information, and then transfers the bidding information to the communication unit through the user interface unit; After receiving the bidding information, the communication unit sends the bidding information to the network according to the pre-agreed agreement. 9. the elevator group control method based on multi-agent competition mode according to claim 7 or 8, is characterized in that, concrete steps are as follows: In the first step, when the communication unit of the agent receives the bidding information and transmits it to the bidding unit, the bidding unit judges whether it has the ability to serve the bidding information, that is, whether it meets the bidding requirements. step, otherwise, set the bid price to the agreed worst price, and go to step 7; In the second step, the bidding unit generates a bid price according to its own conditions and according to the set bid price algorithm; In the third step, the bidding unit compares the generated bid price with the current optimal price in the bidding information. If it is better than the current optimal price, it will go to the fourth step, otherwise, it will go to the fifth step; The fourth step, the bidding unit replaces the bidder with the current best price and the best price, and enters the seventh step; The fifth step, the bidding unit compares the generated bid price with the current alternate price in the bidding information, and if it is better than the current alternate price, the sixth step is entered; The sixth step, the bidding unit replaces the current standby price and the standby price bidder, and enters the seventh step; In the seventh step, the bidding unit updates the bidding information including the bidder list, and transmits the bidding information to the communication unit via the user interface unit, and the communication unit sends the bidding information of the agent to the network where it is located. 10. The elevator group control method based on multi-agent competition mode according to claim 5, is characterized in that, concrete steps are as follows: In the first step, the bidding unit of the agent compares the bidder list in the bidding information with the member list. If all agents have bid, go to the fourth step, otherwise go to the second step; In the second step, the bidding unit calculates the elapsed time of bidding, that is, the time interval between the current time and the start time of bidding; In the third step, the bidding unit judges whether the elapsed time for bidding exceeds the longest bidding period, and if so, enters the fourth step; otherwise, the bidding information is made and sent to the network where it is located; In the fourth step, the bidding unit sets the bidding status in the bidding information as the bidding end sign, and transmits the bidding end information to the communication unit through the user interface unit, and the communication unit sends it to all the agents within the coordination range. 11. The elevator group control method based on a multi-agent competition mode according to claim 5, wherein after the bidding process is over, when the agent confirms that it is the winning bidder, it generates acceptance information and sends it to each team within the coordination scope. An online agent; if it is not the bid winner, it will further confirm whether it is a candidate bid winner. If it is and fails to receive the bid winner's acceptance information within the specified time, it will switch itself to the bid winner and notify the control unit of the bid winning information. , and at the same time publish the undertaking information, and clearly record that the service will be undertaken by the candidate winning bidder. 12. The elevator group control method based on multi-agent competition mode according to claim 11, characterized in that, after the control unit of the agent receives the bid-winning information, it forms an execution plan and executes the service, and simultaneously all the agents send the bid-winning information It is recorded in the service registration form held by the bidding unit. After the agent executing the service completes a service, the control unit of the agent notifies the bidding unit, and the bidding unit generates the service completion information and sends it through the user interface unit and the communication unit. For all agents; the service registration form is used to record all bidding information and final undertakers in the elevator group control system. 13. The elevator group control method based on the multi-agent competition mode according to claim 12, wherein the bidding unit of the agent traverses the service registration table according to a specific period, and when a tender that can be re-bid is found in the service registration table information, the bidding unit generates re-bidding information, and requests the communication unit to send the re-bidding information to the current undertaker of the service via the user interface unit; after the agent as the current undertaker receives the re-bidding information, its bidding unit sends the re-bidding information to the service Re-evaluation is performed. If the bid price is still better than the bid price of the re-bidder, the result is to maintain the status quo; otherwise, the re-bidder wins the bid. 14. The elevator group control method based on a multi-agent competition method according to claim 12, wherein when a certain agent wins the bid but cannot perform the service it undertakes, the bidding unit of the agent generates fault information and publishes After arriving at the network where each agent receives the fault information, the bidding unit checks the service registration table, revises all the faulty agents in the table as the record of the winning bidder, and replaces the winning bidder and the winning bidder with the candidate price and the candidate price. The newly elected winning bidder informs the control unit of the service, the control unit schedules the service and executes the service.

Images