KR20120135935A - Robot control system and robot control method using a cognitive architecture - Google Patents

Abstract

PURPOSE: A robot control system using cognitive architecture and a robot controlling method are provided to deduce commands required for a task of a robot by using the cognitive architecture so that the robot can actively cope with environmental variations on a real time basis. CONSTITUTION: A robot control system(1) comprises a robot(200) and cognitive architecture(100). The robot comprises a sensing unit(210), a transceiving unit(220), and a driving unit(230). The sensing unit senses the information of an external environment. The transceiving unit transmits the sensed environmental information to the cognitive architecture and receives an execution command from the cognitive architecture. The driving unit receives the execution command. The architecture receives the external environmental information sensed by the sensing unit, thereby deducing a present condition of the robot and transmits the execution commands to the transceiving unit of the robot by generating the execution command corresponding to a goal. [Reference numerals] (AA) Wire/wireless network

Description

Robot control system and robot control method using cognitive architecture {ROBOT CONTROL SYSTEM AND ROBOT CONTROL METHOD USING A COGNITIVE ARCHITECTURE}

The present invention relates to a robot control system and a robot control method using a cognitive architecture. More particularly, the present invention relates to a robot control system and a robot that recognizes a surrounding environment of a robot and actively controls movement and execution of a robot according to environmental changes. It relates to a control method.

Recently, attempts to use robots in daily life are increasing. In order to use a robot in the human living space, it must be able to cope with various unexpected situations and be able to safely perform work. Existing robots perform tasks manually in a predetermined order, which reduces the stability and the ability to cope with various environmental changes. Therefore, the desire to safely perform tasks in the home with a robot using a cognitive structure that mimics the cognitive ability of a person is increasing.

In order to realize the above object, the present invention provides a robot control system and a robot control method capable of performing a task by actively controlling the robot under various circumstances according to the change of the surrounding environment by using the cognitive architecture.

Robot control system according to an embodiment of the present invention for achieving the above object, a sensing unit for sensing the external environment information, and a transceiver for transmitting the detected environment information to the cognitive architecture and receiving the execution command therefrom and A robot having a driving unit configured to receive and execute an execution command; And a cognitive architecture for receiving external environment information sensed by the sensing unit, inferring the current state of the robot, generating an execution command corresponding to a target to be performed by the robot, and transmitting the generated execution command to the transceiver of the robot. .

The cognitive architecture may include a receiver configured to communicate with the robot through a wired or wireless network; A cognitive model unit configured to generate an execution command for driving the robot through a cognitive inference process based on external environment information of the robot received through the receiver; And a transmission unit configured to transmit an execution command generated by the cognitive model unit to the robot to operate the driving unit of the robot.

The recognition model unit, a recognition unit for recognizing the external environment information of the robot received by the receiving unit; A long term concept memory in which information about a state of the robot is predefined and stored; A concept inference unit that infers a current state of the robot by comparing the information recognized by the recognition unit with information stored in the long-term concept memory; A short-term concept memory that temporarily stores information on the current state of the robot inferred by the concept inference unit and transfers the information to the skill search and selection unit; A short-term target memory for temporarily storing a target to be performed by the robot; A long term skill memory in which information about the operation of the robot is predefined; A skill retrieval and selection unit for retrieving and selecting a skill executable by the robot in the current state of the robot inferred by the concept inference unit from the long-term skill memory to achieve a goal stored in the short-term target memory; And an execution command generation unit configured to generate an execution command for driving the robot based on the skill selected by the skill search and selection unit.

When the skill search and selection unit does not search for a desired skill from the long-term skill memory, the cognitive model unit decomposes the short-term target stored in the short-term target memory into subgoals and corresponds to each sub-target. The method may further include a skill learning unit searching for a skill in the long-term skill memory and storing the found skills as a new skill in the long-term skill memory in association with a detailed goal and a final goal.

The sensing unit may include at least one of a position recognition sensor, a voice recognition sensor, a laser sensor, and a camera sensor.

The robot may include a plurality of robots, and the cognitive architecture may generate a plurality of execution commands for driving the plurality of robots, respectively, to control the plurality of robots.

The robot control method according to an embodiment of the present invention is a method using a cognitive architecture for generating an execution command corresponding to a target that a robot wants to perform, and receiving external environment information sensed by the robot as the cognitive architecture. ; Generating an execution command for driving the robot in response to the received external environment information through a cognitive inference process; And controlling the robot by transferring the execution command generated by the cognitive architecture to a driving unit of the robot.

In the generating of the execution command through the cognitive inference process, the information about the state of the robot is predefined and compares the information of the long-term concept memory and the external environment information, the current state of the robot in the concept inference unit Inferring; Retrieving and selecting, from the long-term skill memory, a skill executable by the robot in the current state of the robot inferred by the concept inference unit to achieve the goal that the robot is to perform; And generating an execution command for driving the robot based on the selected skill.

In the generating of the execution command through the cognitive inference process, if a desired skill is not retrieved from the long-term skill memory, the robot decomposes a target to be performed into subgoals and then corresponds to each subgoal. The method may further include retrieving a skill to be stored in the long-term skill memory as a new skill in association with the detailed goal and the final goal.

The robot control system of the present invention has the effect of proactively responding to environmental changes in real time by inferring the commands necessary for the robot's work using a cognitive architecture that mimics the human's cognitive ability.

In addition, the robot control system of the present invention, when the robot performs a task or moves in the home or workplace, there is an effect that can safely control the robot by recognizing the surrounding obstacles in real time.

1 is a block diagram of a robot control system according to an embodiment of the present invention.
FIG. 2 is a diagram illustrating the recognition model of FIG. 1.
3 is a flowchart illustrating a robot control method according to an embodiment of the present invention.
4 is a block diagram of a robot control system including one or more robots according to another embodiment of the present invention.

Hereinafter, a robot control system and a robot control method according to a preferred embodiment of the present invention with reference to the accompanying drawings will be described in detail.

1 is a block diagram of a robot control system according to an embodiment of the present invention.

Referring to FIG. 1, the robot control system 1 of the present invention includes a cognitive architecture 100 that controls the robot 200 and a robot 200 that directly performs a task.

In detail, the cognitive architecture 100 includes a receiver 110, a cognitive model 120, and a transmitter 130, and the robot 200 includes a sensing unit 210, a transceiver 220, and a driver 230. It includes. Here, the components shown in FIG. 1 are basic components, and a robot control system having more components may be implemented, and each component is divided into a plurality of elements such as two or more driving units 230. It can also be configured.

The cognitive architecture 100 receives the external environment information from the robot 200, and directly operates the robot 230 to drive the driver 230 included in the robot 200 in order to achieve a goal that the robot 200 intends to perform. After generating, and transmits to the transceiver 220 of the robot 200, to control the operation of the robot 200.

The robot operation signal refers to an execution command signal that is transmitted to the robot 200 to control the driver 230 of the robot 200. When the execution command signal is transmitted, the robot 200 may perform an operation according to the execution command signal through an algorithm implemented in each driver 230.

The receiver 110 communicates with the robot 200 through a wired or wireless network to receive external environment information detected by the sensing unit 210 of the robot 200. The external environment information sensed by the sensing unit 210 of the robot 200 may include location information, obstacle information, object information for cleaning, current state information of the robot 200, etc. necessary for the robot 200 to perform a task. It may include.

The cognitive model unit 120 generates an execution command for driving the robot 200 through a cognitive inference process based on the external environment information received through the receiver 110.

The cognitive architecture 100 previously stores basic operation command information for controlling the driver 230 of the robot 200 required for the robot 200 to achieve a given goal. The cognitive model unit 120 generates an execution command that can control the driving unit 230 of the robot 200 through an inference process.

The transmitter 130 may directly transmit the execution command generated by the recognition model unit 120 to the robot 200 to directly drive one or more drivers 230 included in the robot 200. The receiver 110 and the transmitter 130 of the cognitive architecture 100 communicate with the robot 200 through a wired or wireless network, through which feedback control of the robot 200 may be performed.

The sensing unit 210 acquires external environment information necessary for the cognitive architecture 100 to perform inference using a sensor. The sensing unit 210 may include at least one of a location recognition sensor, a voice recognition sensor, a laser sensor, and a camera sensor.

For example, the sensing unit 210 may recognize the position information of the robot 200 such as the current position and the azimuth of the robot 200 through the position recognition sensor, and the robot when an emergency occurs through the voice recognition sensor. Directly to the command 200, it may correspond to a sensor misrecognition, and the like, it is possible to obtain information such as the position and direction, the size of the obstacle appearing in front of the robot 200 through the laser sensor, the robot ( 200 may obtain information about the object being worked on.

The driver 230 receives the execution command transmitted from the cognitive architecture 100 to realize the driving operation of the robot 200. The driving unit 230 may be configured with two or more driving units to ensure the multiple degree of freedom of movement of the robot 200.

The cognitive model unit 120 performs an inference operation by mimicking a human cognitive structure. Referring to FIG. 2, which illustrates the configuration of the cognitive model unit of FIG. 1 in detail, the cognitive model unit 120 includes a recognition unit 121, a long term concept memory 122, a concept inference unit 123, and a short term concept memory 124. A skill search and selection unit 125, a long term skill memory 126, a short term goal memory 127, and an execution command generator 128, and a skill learning for problem solving as needed. The unit 129 may further include.

The recognition unit 121 recognizes external environment information of the robot 200 received by the receiving unit 110. As described above, the external environment information includes position information, obstacle information, object information for cleaning, current state information of the robot 200, etc. necessary for the robot 200 to perform a task.

The long-term concept memory 122 stores predefined general information used to describe the state of the robot 200. Here, the information about the state of the robot 200 may be information related to the position and azimuth angle of each joint of the robot 200, the distance between the robot 200, and the obstacle.

The concept inference unit 123 infers the current state of the robot using the information stored in the long-term concept memory 122 based on the information recognized by the recognition unit 121.

The short-term concept memory 124 temporarily stores information about the current state of the robot 200 inferred by the concept inference unit 123 so that other parts of the cognitive architecture, including the skill search and selection unit 125, may use it. Make sure

In the long-term skill memory 126, general information about the operation of the robot 200 is predefined and stored, and a target that the robot 200 intends to perform is temporarily stored in the short-term target memory 127.

The skill retrieval and selection unit 125 stores the skills that can be executed by the robot 200 in the current state of the robot inferred by the concept inference unit 123 to achieve the target stored in the short-term target memory 127. 126).

The execution command generation unit 128 generates an execution command for driving the robot 200 based on the skill selected by the skill search and selection unit 125.

When the skill search and selection unit 125 does not search for a desired skill from the long-term skill memory 126, the skill learning unit 129 decomposes the short-term goal stored in the short-term goal memory 127 into subgoals. Thereafter, the skills corresponding to each detailed goal are retrieved from the long term skill memory 126, and the retrieved skills are stored in the long term skill memory 126 in association with the detailed goal and the final goal.

Hereinafter, the inference process of the cognitive model unit 120 will be described with specific examples.

First, when the robot 200 is set to stack five red blocks among arbitrary red blocks and blue blocks on the table, the target is temporarily stored in the short-term target memory 127. do.

Subsequently, the sensing unit 210 mounted on the robot 200 detects external environment information around the robot 200, and the detected environment information is received by the receiver 110 of the cognitive architecture 100 through the transceiver 220. Is passed to.

Subsequently, the recognition unit 121 recognizes the external environment information from the information input from the receiving unit 110. Accordingly, information such as the distance between the robot 200 and the table, the positions of the red and blue blocks placed on the table, the position and the azimuth angle of the robot arm, and the like can be recognized.

Subsequently, the concept inference unit 123 infers the current state of the robot using the general information stored in the long-term concept memory 122 based on the information recognized by the recognition unit 121. Accordingly, from the information such as the distance between the robot 200 and the table, the positions of the red and blue blocks placed on the table, the position and the azimuth angle of the robot arm, the higher-order information, that is, the robot arm reaches the specific block. Whether or not, which blocks are on the table, which blocks are on the other, and the like, and the inferred information is stored in the short term concept memory 124.

In the long-term skill memory 126, the robot 200 moves in front of the table avoiding obstacles in a predetermined space, and the robot 200 selects and grabs a red block on the table, and then stacks a stack of skills related to the robot operation. The skill search and selection unit 125 searches for skills for performing the operations and selects the one suitable for the current state of the robot stored in the short term concept memory 124.

Subsequently, the execution command generator 128 moves the robot 200 to the front of the table avoiding obstacles in a predetermined space, and executes a series of skills related to the robot operation to execute a goal of stacking five red blocks. The execution command is generated from and transmitted to the robot 200.

Subsequently, the driving unit 230 of the robot 200 operates to move the robot 200 to the front of the table, and the robot arm moves on the table to select and stack five red blocks.

Here, if the blue block is placed on the red block on the initial table, and the robot 200 cannot catch the red block, the short-term goal of stacking five red blocks cannot be performed. The selector 125 determines that there is no suitable skill for robot operation.

When the short-term goal cannot be performed as described above, the skill learning unit 129 decomposes the short-term goal of picking and stacking five red blocks. In other words, in order to select a red block, first, by removing a blue block, grabbing a red block, and stacking red blocks, the original short-term target is decomposed into detailed targets, and then a skill corresponding to each detailed target is obtained. Is retrieved from the long-term skill memory 126 and stored in the long-term skill memory 126 in association with the detailed goal and the final goal.

Subsequently, the skill search and selection unit 125 retrieves the skills corresponding to each detailed target from the long-term skill memory 126 and delivers the skills to the execution command generator 128, and the execution command generator 128 is blue. Removing the block, catching the red blocks, generates an execution command for performing the step of stacking the red blocks, and transmits to the robot 200.

Subsequently, the driving unit 230 of the robot 200 operates to achieve the goal of stacking five red blocks.

The above example is for explaining the operation of the components of the present invention, the spirit of the present invention is not limited by the description of the above example.

3 is a flowchart illustrating a robot control method according to an embodiment of the present invention.

Referring to FIG. 3, a robot control method using a cognitive architecture for generating an execution command corresponding to a target to be performed by the robot 200, first, the external environment information detected by the robot 200 is a cognitive architecture 100. Is received (S310).

Next, the cognitive architecture 100 generates an execution command for driving the robot 200 in response to the received external environment information through a cognitive inference process (S320).

Here, the process of generating an execution command through the cognitive inference process, based on the external environment information by using the information of the long-term concept memory 122, the information on the state of the robot 200 is stored in advance, the robot ( The robot 200 stored in the short-term concept memory 124 to infer the current state of the current state 200 in the concept inference unit 123 and store in the short-term concept memory 124, and to achieve the target to be performed by the robot. Searching for and selecting a skill that the robot 200 can execute in the current state of the memory from the long-term skill memory 126, and generating an execution command for driving the robot 200 based on the selected skill. have.

In addition, in the process of generating an execution command through the cognitive inference process, if the desired skill is not retrieved from the long-term skill memory 126, the robot 200 decomposes the target to be performed into subgoals. The method may further include retrieving a skill corresponding to each sub-target from the long-term skill memory 126 and storing the retrieved skills in the long-term skill memory 126 in association with the sub-target and the final goal.

Subsequently, the execution command generated by the cognitive architecture 100 is transmitted to the driving unit 230 of the robot 200 to control the robot 200 (S330).

4 is a block diagram of a robot control system including one or more robots according to another embodiment of the present invention.

Referring to FIG. 4, a plurality of robots 200a, 200b, 200c,... Controlled by the cognitive architecture 100 are configured. The plurality of robots 200a, 200b, 200c,... May be configured in the same form or in different forms.

The cognitive architecture 100 previously stores therein robot-specific features for driving different robots 200a, 200b, 200c, ..., or from the robots 200a, 200b, 200c, ..., respectively. In response to each of the robots 200a, 200b, 200c,..., Inferring work may be performed to generate an execution command. Here, the cognitive architecture 100 generates a plurality of execution commands for driving the plurality of robots 200a, 200b, 200c,..., Respectively, to operate the plurality of robots 200a, 200b, 200c,... Each can be controlled.

In addition, the cognitive architecture 100 transmits execution instructions to perform the same operation to each robot 200a, 200b, 200c, ..., or sets different execution instructions to perform different operations. (200a, 200b, 200c, ...).

In addition, the cognitive architecture 100 transmits an execution command to operate the robots 200a, 200b, 200c, ... at the same time or at different times, depending on the settings. 200c, ...) can be controlled.

While the present invention has been described in connection with what is presently considered to be practical exemplary embodiments, it is to be understood that the invention is not limited to the disclosed embodiments, but, on the contrary, Of course.

1: robot control system
100: cognitive architecture
110:
120: cognitive model unit
121: recognition unit
122: long-term conceptual memory
123: Inference Unit
124: Short-Term Concept Memory
125: skill search and selection unit
126: long-term skill memory
127: short-term target memory
128: execution command generation unit
129: skill learning unit
130: transmitter
200: robot
210: sensing unit
220: transceiver
230:

Claims (9)

A robot having a sensing unit which senses external environment information, a transmitting / receiving unit which transmits the sensed environmental information to a cognitive architecture and receives an execution command therefrom, and a driving unit which receives and executes an execution command; And
Receiving an external environment information sensed by the sensing unit, inferring the current state of the robot, generating an execution command corresponding to a target to be performed by the robot, and including a cognitive architecture for transmitting to the transceiver of the robot Robot control system characterized in that. The method of claim 1, wherein the cognitive architecture,
A receiver for communicating with the robot through a wired or wireless network;
A cognitive model unit configured to generate an execution command for driving the robot through a cognitive inference process based on external environment information of the robot received through the receiver; And
And a transmitter configured to transmit an execution command generated by the cognitive model unit to the robot to operate the driving unit of the robot. The method of claim 2, wherein the cognitive model unit,
Recognizing unit for recognizing the external environment information of the robot received by the receiving unit;
A long term concept memory in which information about a state of the robot is predefined and stored;
A concept inference unit that infers a current state of the robot by comparing the information recognized by the recognition unit with information stored in the long-term concept memory;
A short-term concept memory that temporarily stores information on the current state of the robot inferred by the concept inference unit and transfers the information to the skill search and selection unit;
A short-term target memory for temporarily storing a target to be performed by the robot;
A long term skill memory in which information about the operation of the robot is predefined;
A skill retrieval and selection unit for retrieving and selecting a skill executable by the robot in the current state of the robot inferred by the concept inference unit from the long-term skill memory to achieve a goal stored in the short-term target memory; And
And an execution command generation unit configured to generate an execution command for driving the robot based on the skill selected by the skill search and selection unit. The method of claim 3, wherein the cognitive model unit decomposes the short-term target stored in the short-term target memory into subgoals when the skill search and selection unit does not search for a desired skill from the long-term skill memory. And a skill learning unit which searches for the skills corresponding to the detailed targets in the long-term skill memory and stores the found skills in the long-term skill memory in association with the detailed targets and the final targets. The robot control system of claim 1, wherein the sensing unit comprises at least one of a position recognition sensor, a voice recognition sensor, a laser sensor, and a camera sensor. According to claim 1, wherein the robot is composed of a plurality,
The cognitive architecture generates a plurality of execution commands for driving the plurality of robots, respectively, and control the plurality of robots, characterized in that the robot. A robot control method using a cognitive architecture for generating an execution command corresponding to a target that a robot wants to perform,
Receiving external environment information sensed by the robot as the cognitive architecture;
Generating an execution command for driving the robot in response to the received external environment information through a cognitive inference process; And
And transmitting the execution command generated by the cognitive architecture to a driving unit of the robot to control the robot. The method of claim 7, wherein the generating of the execution command through the cognitive inference process,
Inferring a current state of the robot from a concept inference unit by comparing information of a long-term concept memory with information about a state of the robot previously stored and external environment information;
Retrieving and selecting, from the long-term skill memory, a skill executable by the robot in the current state of the robot inferred by the concept inference unit to achieve the goal that the robot intends to perform; And
And generating an execution command for driving the robot based on the selected skill. The method of claim 8, wherein the generating of the execution command through the cognitive inference process comprises:
If the desired skill is not retrieved from the long-term skill memory, the robot disassembles the target to be performed into subgoals, searches for a skill corresponding to each subgoal, and searches the skills for the subgoal and final goal. And storing in the long-term skill memory in association with the robot control method.

Images