CN112677152B - Planning and dynamic supervision control method for multi-robot operation process - Google Patents
Planning and dynamic supervision control method for multi-robot operation process Download PDFInfo
- Publication number
- CN112677152B CN112677152B CN202011483910.1A CN202011483910A CN112677152B CN 112677152 B CN112677152 B CN 112677152B CN 202011483910 A CN202011483910 A CN 202011483910A CN 112677152 B CN112677152 B CN 112677152B
- Authority
- CN
- China
- Prior art keywords
- planning
- robot
- behavior
- alliance
- dynamic
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/02—Total factory control, e.g. smart factories, flexible manufacturing systems [FMS] or integrated manufacturing systems [IMS]
Landscapes
- Manipulator (AREA)
Abstract
The invention discloses a planning and dynamic supervision control method for a multi-robot operation process, which comprises the following steps of 1: several assumptions and associated definitions; step 2: planning and implementing alliance behaviors; and step 3: describing conditions by a federation behavior Petri net; and 4, step 4: planning and realizing the behavior of the alliance structure; and 5: formal language description and dynamic monitoring. On the basis of realizing action and task planning by an artificial intelligence method, the invention establishes the condition that a plurality of Petri networks form a composite network; 4 new character operators are designed, and the purposes of describing the sequence, synchronization and concurrent behaviors of the alliance or alliance structure through a formal language generated by an event set (an action set or a subtask set) are achieved; and a graphic unit symbol corresponding to the character operator is established, and the imaging of the behavior of the analytic form is realized. In the implementation process of the planning result of the alliance structure, in order to reduce the computational complexity existing in the system and adapt to the limitation of the currently available robots, a dynamic monitor is designed.
Description
Technical Field
The invention relates to the technical field of mobile robot planning, in particular to a planning and dynamic supervision control method for a multi-robot operation process.
Background
At present, in the aspect of mobile robot planning technology, the path planning technology of a single robot is abundant in achievement and more in application, and the technologies can be divided into a path planning method oriented to indoor/outdoor, static/dynamic, two-dimensional/three-dimensional and other environments and various action planning technologies; the method for planning multiple robots is relatively limited, and typically, the method includes contents such as collaborative path planning of multiple robots, middleware-based multi-robot collaboration and the like. On the other hand, in a scene that a mobile robot provided with a manipulator is required to cooperatively work, subtask and action planning are involved in the multi-robot working process, in the connection aspect of a planning result and a monitoring layer, a supervision control method based on a formal language and an automaton has the advantage of formal description, the concurrent behavior existing in a system is described by using the concurrent product of the automaton, the state explosion of the system is brought, a Petri network can describe the concurrent behavior but lacks formal expression, and the situation clearly forms an obstacle for establishing the planning result of the formal monitor. Finally, task-level planning and implementation are limited by both coordination and the number and variety of currently available robots, and currently there is no effective solution.
Disclosure of Invention
The invention aims to provide a planning and dynamic supervision control method for a multi-robot operation process, which analyzes the composition of various coordination and restriction relations and establishes formal expression to form a production system; aiming at the planning problems of each subtask and each robot action related to multi-robot cooperative operation, a simple and efficient subtask and action planning method based on an artificial intelligence search algorithm is established; for a multi-robot system with concurrent behaviors, 4 new character operators are designed, corresponding Petri net model elements are established, and the purpose of expanding the Petri net behaviors through formal language description is achieved; the dynamic monitor is designed at the federation architecture level to reduce the computational complexity present in the system to solve the problems set forth in the background above.
In order to achieve the purpose, the invention provides the following technical scheme:
a planning and dynamic supervision control method for a multi-robot operation process comprises the following steps:
step 1: several assumptions and associated definitions
The heterogeneous multi-robot system carries a plurality of discrete objects in a structured environment, the algorithm for forming the robot alliance and alliance structures is known, and the system is of a layered structure;
and 2, step: federation behavior planning and implementation
The first-order predicate logic expresses a collaborative relationship, a state space search algorithm in artificial intelligence solves subtask planning, and supervision control (monitoring) is performed to realize a planning result;
and step 3: conditions for describing alliance behavior Petri net
The operation actions of each robot are respectively planned and converted into a Petri network, then the Petri networks are combined to obtain the overall behavior of the alliance, and the connection condition of the Petri network is established;
and 4, step 4: federation architecture behavior planning and implementation
The task-level planning is not only limited by the cooperative relationship, but also limited by the number and the types of the currently available robots, and a simultaneous planning and implementation method meeting the limiting conditions is established;
and 5: formal language description and dynamic monitoring
4 character operators are defined, the operation behaviors of a plurality of robots expressed by the Petri network are analyzed in a logic level, and a dynamic monitoring method for reducing the computational complexity is designed.
Further, step 1 adopts a layered architecture, the lower layer and the upper layer plan the actions and subtasks, and each plan is performed independently; the lower layer and the upper layer are respectively responsible for implementing the action and subtask planning result, and the system structure of the monitoring system is a layered Petri network.
Further, the steps of solving the subtask planning by applying the state space search algorithm in the artificial intelligence are as follows:
(i) State space definition, wherein robot positions, manipulator states, operation object positions and the like expressing key nodes of a work process are defined as state space variables;
(ii) Action set definition, wherein the actions which are executed by the robot body, the mechanical arm and cause state change are defined as elements of the action set;
(iii) Constructing a production system, expressing the cooperative relationship by using first-order predicate logic, and forming the production system by using all the cooperative relationships;
(iv) And searching the action corresponding to the current state from the action set under the control of the production system.
Further, step 3 needs to consider three typical combinations: sequence, fork and merge.
Further, class 3 constraints in step 4 are: the first type is integrity limitation, i.e. the subtask work cannot damage the integrity of the operated object; the second category is organizational rule restrictions, such as order and coordination of operations among subtasks, which are represented herein as first order predicate logic, that involve operand selection.
Further, in object selection, the following rules should be followed:
(1) If the distance between the current robot and the operation object is different, selecting a subtask from the nearest operation object;
(2) If the operands are stacked together, selecting the subtask from the top-most;
(3) After the subtask operation object is selected and the corresponding action is executed, other objects cannot be damaged.
The third type of limitation is that the selection of the subtasks takes into account the association and interference between subtasks, resource conflicts, etc.
Compared with the prior art, the invention has the beneficial effects that:
1. in the subtask level and task level planning, the cooperation is respectively expressed as first-order predicate logic, and the problem of subtask/task planning is solved through an artificial intelligent search algorithm; the search computation complexity of each cycle is made O (2) by defining new state variables in the task-level planning n m) decreases to O (nm).
2. The problem that a graphical method and a formalization method are independently developed in the planning process is solved. 4 new formal language operators are defined, the attributes and the applicable domains of the operators are given, the graphical expression of the new operators in the Petri network is established, and finally, the operation behaviors of the alliance level and the alliance structure level can be expressed through the formal language and the extended Petri network model.
3. The problem that time is not involved in logic layer planning result research is solved. For the synchronous and concurrent two character (string) operators defined by the invention, the overlapping relation between the two (a plurality of) character (string) execution time sets corresponding to the operators is analyzed.
4. The problems that task-level planning is simultaneously limited by the cooperative relationship and the number and the types of the currently available robots, and the expected behavior of the monitoring system needs to be designed off-line are solved, and an online simultaneous planning and monitoring implementation scheme is provided.
5. The problem that a large-scale multi-robot system lacks an efficient dynamic monitoring strategy in the task-level planning and planning result implementation process is solved. A method for realizing task planning results in an online mode, namely dynamically determining expected behaviors and realizing monitoring is designed, so that the computational complexity is reduced to O (n).
Drawings
FIG. 1 is a flow chart of a task-level simultaneous planning and monitoring algorithm of the present invention;
FIG. 2 is a schematic diagram of a computer handling operation of the present invention;
FIG. 3 is a schematic diagram of a mobile display subtask planning process according to the present invention;
FIG. 4 is a Petri net model for hierarchically monitoring expected behaviors of job tasks according to the invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
A planning and dynamic supervision control method for a multi-robot operation process comprises the following steps:
step 1: several assumptions and associated definitions
Suppose that what heterogeneous multi-robot systems currently need to accomplish is a task of handling several discrete objects in a structured environment, which is time-critical and consists of several sub-tasks, loosely coupled. Each subtask is in turn made up of a number of actions, the actions that the robot can perform being known. The robots required for completing a subtask are collectively called a robot alliance (hereinafter simply referred to as alliance), which may be one or more homogeneous or heterogeneous robots; all robot alliances that accomplish the entire task are called robot alliance structures (hereafter referred to as alliance structures). The robots contained in the alliance structure are only a part of a multi-robot system, and alliance calculation can be achieved by applying the existing algorithm.
In addition, it is assumed that the kind of the cooperative relationship between the respective operations is known. That is, whether a subtask requires one or more robots to complete, these actions are limited by physical laws; in addition, actions are also limited by the operation rules of the objects, such as dependency relationship among the actions; finally, the execution condition of an action is affected by certain robot actions in the same federation or in other federations, such as shared space, resource limitations, and the like. The three types of restrictions jointly form a subtask level cooperative relationship, and subtask planning is to determine an action sequence which satisfies the cooperative relationship and completes a job target.
The invention adopts a layered architecture, the lower layer is also called a coalition layer or a subtask layer, and the upper layer is also called a coalition layer or a task layer. The lower layer and the upper layer plan the actions and the subtasks, and each plan is independently performed; the lower layer and the upper layer are respectively responsible for implementing actions and subtask planning results, and the system structure of the monitoring system is a layered Petri network;
step 2: federation behavior planning and implementation
The method for solving the subtask planning by using the state space search algorithm in the artificial intelligence comprises the following steps:
(i) And (4) defining a state space. The robot position, the manipulator state, the operation object position and the like expressing the key nodes of the operation process are defined as state space variables;
(ii) And defining an action set. The behavior which is executed by the robot body, the mechanical arm and causes the state change is defined as an element of the action set;
(iii) A production system is constructed. Expressing the cooperative relationship by using first-order predicate logic, wherein all the cooperative relationships form a production system;
(iv) And searching the action corresponding to the current state from the action set under the control of the production system.
The pseudo code corresponding to the above planning algorithm is shown in algorithm 1.
In the algorithm, actionList represents an action set, currentState, initialState, nextState, and finnalState represent current, initial, subsequent, and target states, respectively, behavior represents an action sequence, and ε represents an empty action.
When the algorithm is finished, the action sequence behavior is obtained. If a plurality of transition action sequences capable of generating the transition from the initial state to the target state are obtained, the optimal sequence is searched by applying A. In order to accurately realize the planning action on the subtasks, a lower-layer monitor based on a Petri model is designed. The Petri Net is represented here as a 5-tuple.
Pn＝{P,T,L,P 0 ,W L } (1)
Where { P, T, L } is a finite set, P, T, and L are respectively the library set, the transfer set, and the connecting arc set, the 3 variable sets have the relationship P n T = φ,
P∪T≠φ,dom(L)∪cod(L)＝P∪T,P 0 is an initial identity, W L Is the weight of the connecting arc. The discussion herein is limited to deterministic Petri nets, i.e., each output connected to it after a transition occurs gets a token.
If a subtask is completed by a robot, a Petri net can be used to represent all the actions for completing the subtask. In the alliance layer design monitor, firstly, the expected behavior of a closed-loop system is specified, and after concurrent behaviors existing in the operation process are considered, the operation is carried out through a mapping relation: the Petri network model of the operation behavior can be obtained through state-place and action-transfer, and the first mapping relation is that different state values correspond to different place variables; then the control law design of the monitor is carried out, namely only the transition behavior expressed by the Petri net is allowed to occur (output 1), and other transitions are forbidden (output 0).
To ensure the presence of monitors, it is desirable that the behavioral Petri net model should satisfy the reachability of the target library, i.e., p f ∈[p 0 1 in the formula f Is the target depot, p 0 Is the initial library. Furthermore, the Petri net is free of deadlocks, livelocks and loop traps. And step 3: federation behavior Petri netDescription of the conditions
Complex tasks sometimes require several robots of the same federation to be done in concert. For example, an action performed by some robots is a prerequisite for a subsequent action performed by other robots. At the moment, a method of respectively planning and converting each robot into a Petri network and then combining the Petri networks is adopted to obtain the whole behavior Petri network representation of the alliance. Three typical composite structures are considered: the sequence, the bifurcation and the combination, and the corresponding library and the transfer when each Petri net is connected respectively satisfy the following relations:
o 1 ＝t 1 ·∧i 2 ＝·t 2 ∧o 1 ＝i 2 (2)
o 1 ＝t 1 ·∧i 2 ＝·t 2 ∧i 3 ＝·t 3 ∧o 1 ＝i 2 ∧o 1 ＝i 3 (3)
o 1 ＝t 1 ·∧o 2 ＝t 2 ·∧i 3 ＝·t 3 ∧o 1 ＝i 3 ∧o 2 ＝i 3 (4)
in the formula i i 、o i And t i (i =1,K, 3) are respectively Petri nets Pn i (i =1,k, 3) a set of input and output bins, and a set of transitions. The above results can be summarized as: in order to form a complete monitoring behavior model, each sub-Petri netlist should be completed into a workflow diagram, namely, i) i i ∈Pn i ，·i i ＝ε；ii)o i ∈Pn i ,o i = epsilon, where epsilon represents an empty character. The equations (2) to (4) consider the case where the library is used as the ligation point; to shift the basic content of the Petri net as the junction point of these 3 structures.
The process of a subtask operation is reflected in the change of the distribution of the corresponding Petri net tokens, and the invention uses a power set 2 in form P ＝{p i |p i =1 represents the distribution of tokens. For and transition t k Related libraries p i And p j The transfer is allowed to occur under the condition that
Wherein I represents a logical AND operator, I k And O k Are respectively corresponding to t k The input and output set of (a). Let g be k Is the monitor pair transfer t k Control signal of, then t k Is allowed to occur under the condition that
t k Triggered token distribution change
That is, when t is transferred k After this occurs, the tokens are distributed among all output repositories (which may be one or more repositories) to which the transfer corresponds, and the supervisory control process is made up of equations (6) - (8), which is an iterative process.
And 4, step 4: federation architecture behavior planning and implementation
Mission planning is also restricted by co-relationships, consisting of 3 types of restrictions: the first type is integrity limitation, i.e. the subtask job cannot damage the integrity of the operated object; the second category is the organization rule limitation, such as the operation sequence and cooperation among subtasks, which are represented as first-order predicate logic, and the rule relates to the operation object selection, and in the object selection studied by the method, the following rule should be followed:
(i) If the distance between the current robot and the operation object is different, selecting a subtask from the nearest operation object;
(ii) If the operands are stacked, selecting the subtask from the top-most;
(iii) After a subtask operation object is selected and corresponding action is executed, other objects cannot be damaged;
a third type of limitation is that the selection of the subtasks takes into account associations and interferences between subtasks, resource conflicts, etc.
Unlike federation level planning, federation structure level planning is limited by the number and variety of currently available robots in addition to coordination.
In principle, on the premise of satisfying the cooperative relationship, two methods are available for realizing task-level job planning and monitoring. The first method is to adopt the same state space definition as that of the subtask-level planning, then realize the planning through a search algorithm, and finally realize the planning result through monitoring. This approach seems to be feasible, but is practically infeasible due to computational cost. For example, if the task set is t 1 ,K,t n }, then the policy exists 2 according to the previous state definition n A state set element, and if there are m synergistic relationships, the computational complexity of selecting a state and verifying a synergistic relationship in each cycle is O (2) n m). When the number of tasks increases, these calculations are very difficult computational tasks for computers installed on robots with relatively limited computational speed and storage capacity. The reason is that the search is performed in a complex state space, each state variable comprising information of all subtasks. Therefore, a new simple state space definition method for independently processing each subtask is established below.
According to the assumed task characteristics, the planning should have flexibility for the number and types of currently participatable working robots while satisfying the collaborative relationship. Therefore, the state space variables are defined according to the completion state of each subtask, so that only n state variables exist, and each state has only 2 values, which respectively represent t i (i =1,k, n) complete or incomplete. Completion of each subtask job results in a change in the corresponding state value. In other words, one is generated (current state) at each cycle based on the cooperative relationship and the number and kinds of available robotsAnd the above-mentioned restrictions do not allow concurrency) or more (the current state and the above-mentioned restrictions allow concurrency) executable tasks (events), the monitor allows this/these events to occur and prohibits other events from occurring.
The method is a method for simultaneously realizing task planning and supervisory control, and the implementation process is as follows. Before the task starts, the league formation algorithm attempts to form a robot league from the list of incomplete subtasks and the available robots. If a federation forms, the algorithm searches for the corresponding subtask and verifies that the corresponding collaborative relationship is satisfied. If the verification result is positive, calling the corresponding bottom-layer Petri net; otherwise the algorithm tries to form a new robot union. To allow for the occurrence of concurrent behavior, a broad search algorithm in artificial intelligence is applied to conduct the planning search. After the subtasks are completed or a new robot arrives at the job site, the algorithm is restarted from the union formation for the incomplete subtasks. This iterative process is aimed at traversing all the subtasks, and the flow of the algorithm is shown in fig. 1, whose computational complexity is O (nm). The operation behavior of the task layer is also represented as a Petri network which is dynamically generated according to two types of limiting conditions, and the transfer expressing the subtask completion process can be expanded into an action relation Petri network, namely, the planning results and the monitoring of the upper layer and the lower layer are integrated in a layered Petri network structure.
Now, the task-level integrated planning and control method is summarized as follows, which includes two parts, the first part is used for determining executable tasks on the premise of satisfying the cooperative relationship and the current available robot; the second part is to achieve the planning result by monitoring, i.e. only selected subtasks are allowed to occur. From a supervisory control perspective, the current desired behavior model of a closed-loop discrete event system is dynamically generated.
And 5: formal language description and dynamic monitoring
By using the concepts such as character strings in formal languages in computer science, a formal analysis method is established in a logic level for the operation behaviors of a plurality of robots (alliances or alliance structures) expressed by a Petri network. For this purpose, events are represented using lower case greek or english letters, defining 4 character operators as shown in table 1.
TABLE 1 character operator Attribute and its applicable Domain
Operators ≧ and ≦ in the table for characters and finite-length character strings, and omitted when expressing a character string made up of a plurality of characters; operator
For string collections, when a collection is degenerated to have only 1 element, the operator degenerates to an operator-; operator + is used for both character strings and character string sets, and the difference from operator ^ is: t is t i ⊕t j Representing an event t i And t j Occur simultaneously (fully synchronized), and t i +t j Representing overlap in execution time of 2 events
And
respectively represent events t i And t j When the time set occurs, ≦ and ≦ respectively have the following time overlap relationship
Further, if s 1 ＝ζ 1 …ζ p ，s 2 ＝η 1 …η q ，p≠q，s 1 +s 2 Simply representing concurrency of two strings, ζ i (i =1, \8230;, p) and η j There need not be [, ] relation between (j =1, \ 8230;, q),the same relationship as in expressions (9) and (10) exists between the upper and lower time limits of the character string.
In order to realize the operators ^ and +, the invention designs two new Petri net structure units, namely circular arcs through a solid line and a dotted line
And
each expressing the relationship between subsequent characters (strings) connected by one library, as shown by the association behavior in the lower half of fig. 4.
Operator
The expansion calculation rule of (2) is calculated by the same Cartesian product, and is used for describing the overall behavior and operator of a new Petri net system after two Petri nets are combined in a sequential form
The combination with + can be used to describe the overall behavior of the system in both cases (3) and (4), respectively. For example, L (C) 1 ) And L (C) 2 ) Respectively represented by federation C 1 And C 2 2 sets of languages are generated, then
Represents C having a sequential relationship 1 And C 2 All of the actions of (a). These operators have the same priority.
According to the planning part of the algorithm shown in FIG. 1, expected behaviors under various available robot conditions are calculated and then converted into a task layer Petri net model. These planning results are also represented as a formal language L K (CS), where CS is a federation structure and the expected behavior K is only L K One element of (CS), namely K ∈ L K (CS). To distinguish from the conventional method, L is referred to herein K (CS) is referred to as the maximum expected behavior set.
The formal language used in the analysis is generated on the basis of subtask character set, in order to express the situation of the robot participating in the operation intuitively, the formal language generated on the basis of the robot label character set is established below, the operation behavior at the alliance structure level is limited by the number and the type of the robots which can be currently available, and the following mapping relation can be obtained according to the task requirement
f:T→CS j (j＝1,...,n CS ) (11)
Where T is the set of tasks, CS j Is a federation structure formed for a certain number and kind of robots, n CS Is a number of federation structural elements, and therefore
CS＝{CS j |j＝1,...,n CS } (12)
And (5) obtaining the robot label set as the formal language description of the character set according to the formula (11).
When planning to implement, designing the monitor according to the task set form language, and if the expected behavior K is relative to L K (CS) is controllable, i.e.
That is, containing an uncontrollable subtask t k Is part of the desired behavior. From the work behavior that has been completed, the currently available robots and the desired behavior K, the dynamic monitor in equation (6) can be expressed as robot r under the current constraints i Performing a subtask t k Of control law, i.e.
Where K is a set of prefixes of K, s k Representing job actions that have been completed, c k Representing the completion of task t k Required robot alliance CS jk Whether or not the shape of the metal layer has already been formed,
the establishment of the formula (5) represents that the subtask does not conflict with other subtasks; r is i ∈CS jk (i＝1,…,n tk )，CS jk ∈CS j ，n tk Is to complete the subtask t k The number of robots required; let delta (s, p) 0 ) From p for a string s 0 Starting library transfer function, then p i ＝δ(s k ,p 0 )，p j ＝δ(s k t k ,p 0 ) In the formula p i And p j Are as defined in formula (7) and formula (8), respectively, transferring t k The distribution of tokens after occurrence is also determined by these two formulas.
Since the maximum expected behavior in the dynamic monitoring method is calculated off-line, the calculation complexity is reduced to O (n), so the method is suitable for application scenarios with large task sets.
To illustrate the application of the method of the present invention, consider the example of FIG. 2 in which a set of computer jobs are carried by a plurality of robots. The job object includes a display, a chassis, and a keyboard. Assuming that the work site has been powered off and the monitor and the keyboard are connected to the main cabinet through wires, respectively, the work targets first cut off the wires, respectively, and then carry 3 objects to target locations, respectively. The task is decomposed for task set T = { T = } i I =1, \8230 |, 5}, in which t 1 、t 2 Representing the cutting subtasks for the keyboard and display, respectively, for the two wires connecting the main chassis, t 3 、t 4 And t 5 The distribution represents moving the chassis, display and keyboard to a designated location; second, each subtask is broken down into several actions.
Next, a subtask level planning method will be described by taking a robot transport display as an example. Firstly, analyzing the limitation of the operation process, if a transfer robot with 2 mechanical arms can complete the subtask, the robot must follow the action sequence of grasping by hands, lifting, moving and stably putting down a display, and the relationship among the actions is called as a physical law; on the other hand, 2 manipulators should simultaneously grasp the display, and if and only after grasping, subsequent actions, such as moving to a target point, etc., can be performed, which belong to operation rule restrictions; in performing the movement, the robot must avoid collisions with other robots or obstacles, which are interferences between the robots. The contents of the above aspect 3 constitute the entire contents of the cooperative relationship.
The process of expressing the synergistic relationship as first-order predicate logic and solving the action sequence plan through an artificial intelligence search algorithm is shown in fig. 3. The rectangles on the right side in the figure represent the collaborative relationship in the form of first-order predicate logic, wherein { lct, goto, avoid } is a predicate set describing the states and actions of the robot body, { ext, rtt, shk } is a predicate set describing the actions of the mechanical arms, and gram describes the states of the mechanical arms, and the physical meanings of the predicates are shown in table 2. w is a ij (i =1, \8230;, 3 j =1, \8230;, 7) is the cost of performing the corresponding state transition action, i is the transfer robot number, and j is the action number. The middle column represents the state transition process, using (x) 1 ,…,x 6 ) Represents a complex state space variable, where x 1 Is the position of the robot (u is the position on the path between a and b, v is the position on the path between b and c), x 2 And x 3 Is the state of two robots (0 for robot retraction, 1 for extension), x 4 Is the target position of the robot, x 5 And x 6 Is the state of the manipulator (0 represents manipulator idle, 1 represents grasping object). The left column of phrases represents actions that cause state transitions. The motion of one robot is influenced by the motion of the other robots, and for the sake of simplicity, collision avoidance behavior during the movement of the robots is given here.
After obtaining the action sequence of the robot to complete the transportation, the planning result can be converted into the Petri net model N for monitoring the subtasks through the mapping relation described above s As shown in the lower part of fig. 4, where o s The meaning of the remaining transfers, representing the completion of the action, has been labeled in fig. 2. The federation behavior model is part of the underlying behavior in the layered expectation model.
The formal language established by the invention is used to analyze the operation behavior of the alliance, and the operation behavior is saved for spaceWriting a Petri network model N s Formal language description representing behavior
α 1 (β 1 +β 2 )(γ 1 ⊕γ 2 )(η 1 ⊕η 2 )α 2 (ζ 1 ⊕ζ 2 )(θ 1 ⊕θ 2 )(ρ 1 ⊕ρ 2 )
The collaborative relationships and behaviors of the federation structures are analyzed below. Assuming that the exploring robot can complete the cutting of any one connecting line and the transfer robot can execute any one of the three transfer subtasks, the task needs at least one exploring robot and one transfer robot, and the corresponding alliance structure is CS min ＝{{r e1 },{r e1 },{r m1 },{r m1 },{r m1 }, in which r e1 And r m1 Respectively representing the numbers of the exploring robot and the carrying robot, at the moment, the task set T and the alliance structure CS min Can be expressed as f = { (t) 1 ,r e1 ),(t 2 ,r e1 ),(t 3 ,r m1 ),(t 4 ,r m1 ),(t 5 ,r m1 ) }; the task needs 2 exploring robots and 3 carrying robots at most, and the corresponding alliance structure is CS max ＝{{r e1 },{r e2 },{r m1 },{r m2 },{r m3 } in the formula, r ei (i =1,2) and r mj (j =1, \8230;, 3) respectively represent the search robot and the transfer robot numbers, and in this case, the mapping equation (11) can be expressed as f = { (t) 1 ,r e1 ),(t 2 ,r e2 ),(t 3 ,r m1 ),(t 4 ,r m2 ),(t 5 ,r m3 )}。
TABLE 2 predicate and predicate variables
The formal language proposed in the present invention is used to describe the behavior of the federation structure, establish the collaboration and implement dynamic monitoring. For reducing the sign of the variables, use is made hereCharacters consistent with the subtasks represent transitions, i.e., events, of the repositories in the Petri net. Integrity constraints in a collaborative relationship can be implemented by position/force control, only considering organizational rule constraints in the following, which is a subset of the task T 1 ＝{t 1 ,t 2 Should be in the subtask set T 2 ＝{t 3 ,t 4 ,t 5 Before starting, all the steps are completed; in addition, the occurrence of subtask elements in the 2 subtask sets is limited only by the currently available robots. T is 1 The required job behavior is the language set L K1 (CS)＝{t 1 t 2 ,t 2 t 1 ,t 1 +t 2 An element of, where + represents the concurrency as defined in table 1, i.e. expression t if and only if events with + all occur 1 +t 2 Is completed. For subtask set T 2 If only one transfer robot arrives after the cutting of the two links is completed, the subsequent operation behavior is t 3 t 4 t 5 (ii) a Otherwise, if one transfer robot arrives after completing one subtask by another transfer robot, the operation behavior is t 3 (t 4 +t 5 )，t 4 (t 3 +t 5 ) Or t 5 (t 3 +t 4 ) (ii) a Otherwise, if 2 transfer robots arrive at the same time, the operation behavior is (t) 3 +t 4 )t 5 ，(t 4 +t 5 )t 3 Or (t) 3 +t 5 )t 4 (ii) a Finally, if 3 transfer robots arrive at the same time, the operation behavior is (t) 3 +t 4 +t 5 ). Task set T 2 The maximum set of expected behaviors of (c) is:
L K2 (CS)＝{t 3 t 4 t 5 ,t 3 (t 4 +t 5 ),t 4 (t 3 +t 5 ),t 5 (t 3 +t 4 ),(t 3 +t 4 )t 5 ,(t 4 +t 5 )t 3 ,(t 3 +t 5 )t 4 ,(t 3 +t 4 +t 5 )}
expression of the overall behavior of a federation structure as a set of languages
Its elements include t 1 t 2 t 3 t 4 t 5 ，…，t 1 t 2 (t 3 +t 4 +t 5 )，…，(t 1 +t 2 )t 3 t 4 t 5 ，…，(t 1 +t 2 )(t 3 +t 4 +t 5 ) That is, there are 24 character strings (behavior elements) in total.
According to the mapping relation established by the formula (11), the formal language description of the league structure behavior generated by using the robot label as the character set can be obtained. For economy, only the behavior of robots to form two extreme forms of federation structures is given here: when forming CS min The temporal behavior description language is r e1 r e1 r m1 r m1 r m1 }; when forming CS max The temporal behavior description language is
Since operator + does not satisfy the distribution law, this equation is already the simplest form.
The established dynamic monitoring method realizes L according to the condition of the available robots K (CS) one of the behaviors, namely, the task level monitor, is responsible for selecting and implementing one of the upper level behaviors of the transfer from the initial repository ini to the target repository fi in FIG. 4, according to the progress of the work process. For example, when there are only 1 exploring robot and 1 handling robot, the upper level behavior achieved by the dynamic monitoring method of equations (14) - (15) is t 1 t 2 t 3 t 4 t 5 . This result is consistent with the behavior of the integrated approach implementation of fig. 1, and the advantage of the dynamic monitoring approach is to circumvent the complex calculations on-line. The planning action corresponding to each subtask in the lower layer is not changed, so that the specific action monitoring is consistent after the upper layer monitoring action is determined.
The above description is only for the preferred embodiment of the present invention, but the scope of the present invention is not limited thereto, and any person skilled in the art should be considered as the technical solutions and the inventive concepts of the present invention within the technical scope of the present invention.
Claims (10)
1. A planning and dynamic supervision control method for a multi-robot operation process is characterized by comprising the following steps:
step 1: several assumptions and associated definitions
The heterogeneous multi-robot system carries a plurality of discrete objects in a structured environment, the algorithm for forming the robot alliance and alliance structures is known, and the system is of a layered structure;
step 2: federation behavior planning and implementation
The first-order predicate logic expresses a synergistic relation, a state space search algorithm in artificial intelligence solves subtask planning, and a planning result is realized through supervision control;
and step 3: conditions for describing alliance behavior Petri net
The operation actions of each robot are respectively planned and converted into a Petri network, then the Petri networks are combined to obtain the overall behavior of the alliance, and Petri network connection conditions are established;
and 4, step 4: federation architecture behavior planning and implementation
The task-level planning is not only limited by the cooperative relationship, but also limited by the number and the types of the currently available robots, and a simultaneous planning and implementation method meeting the limiting conditions is established;
and 5: formal language description and dynamic monitoring
Defining 4 character operators, which are sequential product ·, concurrent ++, synchronous |, and sequential Cartesian product
And analyzing the operation behaviors of the robot alliance structure expressed by the Petri network at a logic level, and designing a dynamic monitoring method for reducing the computational complexity.
2. The multi-robot operation process planning and dynamic supervision and control method according to claim 1, when applying the state space search algorithm in artificial intelligence to solve the action and subtask planning respectively, characterized in that the cooperative relationship of the action level and the subtask level is expressed as first-order predicate logic respectively, and a processing method when the alliance level structure operation is limited by the number and kind of currently available robots is designed.
3. The method for multi-robot operation process planning and dynamic supervision control according to claim 1, wherein step 3 considers three typical Petri net composite structures: sequence, fork and merge, characterized by the conditions that should be met when 3 structures are established to be connected through the library.
4. A method for multi-robot process planning and dynamic supervisory control as claimed in claim 1, when describing the operation behaviour of a multi-robot system, characterized by establishing the attributes, applicability and priority of 4 operators.
5. The method for planning and dynamically supervising the operation of a plurality of robots according to claim 1, wherein the relationship between upper and lower limits of the execution time of the event in the analytic form is established
And
respectively represent events t i And t j Set of occurrence times, then
And t i +t j The time overlap relationship is as follows
Further, if s 1 ＝σ 1 …σ p ，s 2 ＝η 1 …η q ，p≠q，s 1 +s 2 Representing only the concurrency of two strings, σ i And η j Need not exist in
Relation, σ i I =1, \ 8230;, p; eta j J =1, \ 8230and q in (1) have the same relationship as in expressions (9) to (10) between the upper and lower time limits of the character string.
6. A method for multi-robot procedure planning and dynamic supervision control according to claim 1, characterized in that for realizing operators
And +, two new Petri net structural units are designed, namely circular arcs through solid lines and broken lines
And
respectively, expressing the relationship between subsequent characters connected by one library.
7. A method for multi-robot process planning and dynamic supervisory control as claimed in claim 1, wherein task sets are mapped with robot federation structures
f:T→CS j (j＝1,...,n CS ) (11)
Where T is the task set, CS j Is a federation structure formed for a certain number and kind of robots, n CS Is a number of elements of a federation structureIts relationship to the robot federation structure:
CS＝{CS j |j＝1,...,n CS } (12)
and establishing a second formal language job behavior description method according to the mapping relation (11), namely, a formal language generated based on the robot label character set.
8. A method for multi-robot process planning and dynamic supervisory control as claimed in claim 1 wherein the federation configuration level is such that executable tasks are determined on the premise that the coordination relationship with the currently available robots is satisfied; from a supervisory control perspective, the current expected behavior model of a closed-loop discrete event system is dynamically generated.
9. A method for multi-robot procedure planning and dynamic supervision control according to claim 1, characterized in that all planning results of a federation infrastructure layer are represented as a formal language L K (CS), where CS is a federation structure and the dynamically implemented desired behavior K is only L K One element of (CS), namely K ∈ L K (CS); to distinguish from the conventional method, L is referred to herein K (CS) is referred to as the maximum expected behavior set.
10. A method for planning and dynamic supervision control of a multi-robot working process according to claim 1, characterized in that the dynamic supervisor in equation (6) is expressed as robot r under current limit conditions, depending on the working behaviour that has been completed, the currently available robots and the desired behaviour K, when planning is implemented i Performing a subtask t k Of control law, i.e.
In the formula
Set of prefixes of K, s k Representing job actions that have been completed, c k Represents the completion of task t k Required robot alliance CS jk Whether or not it has already formed, I k And O k Are respectively corresponding to t k The input and output set of (a);
the establishment of the formula (5) represents that the subtask does not conflict with other subtasks; r is i ∈CS jk ；r i Wherein i =1, \8230, n tk ，n tk Is to complete the subtask t k The number of robots required; let delta (s, p) 0 ) From p for a string s 0 Starting library transfer function, then p i ＝δ(s k ,p 0 )，p j ＝δ(s k t k ,p 0 ) In the formula p i And p j Has the same meanings as those of formula (7) and formula (8), respectively, transferring t k The distribution of tokens after occurrence is also determined by these two formulas;
wherein for and transition t k Related libraries p i And p j Conditions under which the transfer is allowed to occur
let g k Is the monitor pair transfer t k Control signal of (2), then k Is allowed to occur under the condition that
t k Triggered token distribution change
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011483910.1A CN112677152B (en) | 2020-12-16 | 2020-12-16 | Planning and dynamic supervision control method for multi-robot operation process |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011483910.1A CN112677152B (en) | 2020-12-16 | 2020-12-16 | Planning and dynamic supervision control method for multi-robot operation process |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112677152A CN112677152A (en) | 2021-04-20 |
| CN112677152B true CN112677152B (en) | 2023-01-31 |
Family
ID=75448166
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011483910.1A Active CN112677152B (en) | 2020-12-16 | 2020-12-16 | Planning and dynamic supervision control method for multi-robot operation process |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112677152B (en) |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108510049A (en) * | 2018-03-23 | 2018-09-07 | 山东大学 | The service autonomous cognitive approach of robot based on emotion-space time information and robot |
| CN108958202A (en) * | 2018-07-27 | 2018-12-07 | 齐齐哈尔大学 | A kind of method that multirobot collaboration is explored |
| CN111113409A (en) * | 2019-11-21 | 2020-05-08 | 东南大学 | Multi-robot multi-station cooperative spot welding operation planning method based on step-by-step optimization |
| CN111273664A (en) * | 2020-02-11 | 2020-06-12 | 西北工业大学深圳研究院 | Multi-robot path collaborative planning method for on-orbit assembly of space truss structure |
| CN111399462A (en) * | 2020-04-01 | 2020-07-10 | 浙江大学 | Discrete and continuous mixed product operation control logic digital twin virtual-real synchronization method |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US7730063B2 (en) * | 2002-12-10 | 2010-06-01 | Asset Trust, Inc. | Personalized medicine service |
-
2020
- 2020-12-16 CN CN202011483910.1A patent/CN112677152B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108510049A (en) * | 2018-03-23 | 2018-09-07 | 山东大学 | The service autonomous cognitive approach of robot based on emotion-space time information and robot |
| CN108958202A (en) * | 2018-07-27 | 2018-12-07 | 齐齐哈尔大学 | A kind of method that multirobot collaboration is explored |
| CN111113409A (en) * | 2019-11-21 | 2020-05-08 | 东南大学 | Multi-robot multi-station cooperative spot welding operation planning method based on step-by-step optimization |
| CN111273664A (en) * | 2020-02-11 | 2020-06-12 | 西北工业大学深圳研究院 | Multi-robot path collaborative planning method for on-orbit assembly of space truss structure |
| CN111399462A (en) * | 2020-04-01 | 2020-07-10 | 浙江大学 | Discrete and continuous mixed product operation control logic digital twin virtual-real synchronization method |
Non-Patent Citations (1)
| Title |
|---|
| 基于离散事件系统的多机器人协调搜索控制研究;沙丽娟,等.;《齐齐哈尔大学学报(自然科学版)》;20120930;第1-5页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112677152A (en) | 2021-04-20 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Balas | Machine sequencing via disjunctive graphs: an implicit enumeration algorithm | |
| Balas et al. | Guided local search with shifting bottleneck for job shop scheduling | |
| Sapaty | Holistic analysis and management of distributed social systems | |
| Zha | An object-oriented knowledge based Petri net approach to intelligent integration of design and assembly planning | |
| Ding et al. | Robotic task oriented knowledge graph for human-robot collaboration in disassembly | |
| Fischer et al. | Unifying control in a layered agent architecture | |
| Lei et al. | Hybrid heuristic search approach for deadlock-free scheduling of flexible manufacturing systems using Petri nets | |
| Akbari et al. | Knowledge-oriented task and motion planning for multiple mobile robots | |
| Cailhol et al. | Multi-layer path planning control for the simulation of manipulation tasks: Involving semantics and topology | |
| Qian et al. | An assembly timing planning method based on knowledge and mixed integer linear programming | |
| Gan et al. | Propagation path optimization of product attribute design changes based on petri net fusion ant colony algorithm | |
| Karami et al. | Task allocation for multi-robot task and motion planning: A case for object picking in cluttered workspaces | |
| Surynek | On Satisfisfiability Modulo Theories in Continuous Multi-Agent Path Finding: Compilation-based and Search-based Approaches Compared. | |
| Eswaran et al. | Optimal layout planning for human robot collaborative assembly systems and visualization through immersive technologies | |
| CN112677152B (en) | Planning and dynamic supervision control method for multi-robot operation process | |
| Luo et al. | Learning rational subgoals from demonstrations and instructions | |
| Zheng et al. | Survey on design approaches for robotic manufacturing systems in SMEs | |
| Mogali et al. | Scheduling for multi-robot routing with blocking and enabling constraints | |
| Zeigler et al. | A simulation environment for intelligent machine architectures | |
| Surynek et al. | Mutex propagation for SAT-based multi-agent path finding | |
| Chen et al. | Development and composition of robot architecture in dynamic environment | |
| Winter et al. | Semi-automatic task planning of virtual humans in digital factory settings | |
| De | A knowledge-based approach to scheduling in an FMS | |
| Zhu et al. | Graphical composite modeling and simulation for multi-aircraft collision avoidance | |
| Chowdhary et al. | Automated planning |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |