On-Device Based Method and System for Robot Control Combining Classical Computing and Quantum Computing

Abstract

A robot control method for controlling a robot using on-device-based quantum computing according to an exemplary embodiment may include: an input step in which a first server disposed within the robot obtains state information from at least one sensor module within the robot; a modeling step in which the first server generates quantum feature data based on the state information; an interpretation step in which a second server disposed outside the robot receives the quantum feature data and performs a quantum neural network (QNN) operation to generate quantum interpretation data; a conversion step in which the first server receives the quantum interpretation data and converts it into control signal data; and an output step in which the first server generates a control command for the robot based on the control signal data and applies it to the robot to control the robot.

Description

Technology Field

[0001] An exemplary embodiment relates to a robot control method and system combining on-device classical computing and quantum computing. Background Technology

[0002] Modern robotic technology is advancing in various fields, such as autonomous driving, industrial automation, and medical services, and research is continuing, particularly with the goal of application in complex environments. Here, a "complex environment" can refer to a space containing hazardous elements or complex paths that are difficult for conventional robots to navigate. For example, complex environments may include natural terrain that is difficult for humans to access directly, or areas where disasters such as earthquakes, fires, or building collapses have occurred.

[0003] In modern robotics technology, active research is being conducted to develop robot path optimization and control techniques in the aforementioned complex environments by utilizing sensor fusion technologies, such as LiDAR, Inertial Measurement Units (IMUs), and cameras, along with Artificial Intelligence (AI). While these technologies perform computations based on conventional classical computing, they exhibit limitations in terms of computational efficiency and energy consumption in tasks such as non-linear optimization problems and large-scale data processing.

[0004] On the other hand, quantum computing applies computational principles different from conventional classical computing, making it highly likely to demonstrate superior performance in parallel processing and solving nonlinear problems. For example, the Quantum Approximate Optimization Algorithm (QAOA) is advantageous for solving complex optimization problems, and Grover's Algorithm can dramatically improve data retrieval speed.

[0005] However, quantum computing technology has the following limitations. First, there may be limitations in real-time computation. Currently commercialized quantum hardware is restricted in computational speed and accessibility, making it difficult to apply to robot control where real-time processing is essential. Second, input / output data bottlenecks may occur. This could lead to performance degradation during the process of encoding data into quantum states and decoding results. Third, there may be hardware constraints. Since quantum computers require cryogenic cooling systems and large-scale equipment, their mobility is limited. Therefore, it may be difficult to integrate quantum computing systems directly into the robot itself. Fourth, there may be issues with noise and reliability. Quantum computations are susceptible to noise, making it difficult to guarantee the accuracy and reliability of the results.

[0006] Therefore, in the field of modern robotics technology, much research is being conducted to overcome the limitations of quantum computing technology as described above. The problem to be solved

[0007] A robot control method and system combining on-device-based classical computing and quantum computing according to the technical concept of the present invention aims to overcome the limitations of real-time computation, eliminate input / output data bottlenecks, overcome hardware constraints, and solve noise and reliability problems in robot control methods.

[0008] The objectives of the present invention are not limited to those mentioned above, and various other goals that the present invention aims to achieve in addition to those described will be clearly understood from the description below. means of solving the problem

[0009] A robot control method combining on-device-based classical computing and quantum computing according to an exemplary embodiment may include: an input step in which a first server disposed within the robot obtains state information from at least one sensor module within the robot; a modeling step in which the first server generates quantum feature data based on the state information; an interpretation step in which a second server disposed outside the robot receives the quantum feature data and performs a quantum neural network (QNN) operation to generate quantum interpretation data; a conversion step in which the first server receives the quantum interpretation data and converts it into control signal data; and an output step in which the first server generates a control command for the robot based on the control signal data and applies it to the robot to control the robot.

[0010] The sensor module includes an Inertial Measurement Unit (IMU) configured to acquire dynamic information regarding the movement of the robot, and the state information may include the dynamic information.

[0011] The above dynamic information is represented as a vector, and the quantum feature data includes IMU qubit data. In the modeling step, the first server generates the IMU qubit data based on the dynamic information, and the IMU qubit data may include at least one qubit to which the dynamic information is mapped to a quantum rotation state.

[0012] The sensor module includes a camera sensor configured to acquire distance information for a specific target, and the state information may include the distance information.

[0013] The distance information is represented as a vector, the quantum feature data includes camera qubit data, and in the modeling step, the first server generates the camera qubit data based on the distance information, and the camera qubit data may include at least one qubit to which each sample point of the distance information is mapped.

[0014] The above robot comprises: a main body part including the sensor module; and a driving part responsible for the movement of the main body part, and The first server above can control the robot by controlling the speed and direction of the drive unit in the output stage.

[0015] The first server above may include a Field-Programmable Gate Array (FPGA).

[0016] A robot control method combining on-device-based classical computing and quantum computing according to an exemplary embodiment may include: an input step in which a first server disposed within the robot obtains state information from at least one sensor module within the robot; a modeling step in which the first server generates quantum feature data based on the state information; an interpretation step in which a second server disposed outside the robot receives the quantum feature data and performs a quantum neural network (QNN) operation to generate quantum interpretation data; a conversion step in which a third server disposed outside the robot receives the quantum interpretation data and converts it into control signal data; and an output step in which the first server receives the control signal data, generates a control command for the robot, and applies it to the robot to control the robot.

[0017] In the above modeling step, the first server can generate data converted into rotation angles of quantum rotation gates (RY, RX, RZ) based on the state information and provide the converted data to the second server.

[0018] In the above interpretation step, the second server can perform the QNN operation by applying a Deep Quantum Neural Network (DQNN) structure. Effects of the invention

[0019] A robot control method combining on-device-based classical computing and quantum computing according to the technical concept of the present invention can control the robot by having a first server deployed within the robot receive state information from first to third sensors to generate quantum feature data, a second server receive the quantum feature data to generate quantum interpretation data, and the first server generate control commands based on the quantum interpretation data. As a result, since the control of the robot can be performed based on on-device-based computation, the real-time control performance of the robot can be improved.

[0020] Furthermore, the first server responsible for classical computing and the second server responsible for quantum computing can be separated. Consequently, during the process of generating signals to control the robot, the computational process is carried out by combining the first server related to classical computing and the second server related to quantum computing; thus, the computational speed can increase dramatically compared to when only the second server is used. Therefore, real-time path optimization and obstacle avoidance performance can be improved, and data processing efficiency can be enhanced.

[0021] In addition, the reliability of quantum computing can be enhanced by applying noise-resistant quantum algorithms.

[0022] For the reasons mentioned above, a robot control method combining on-device classical computing and quantum computing can be utilized in various fields such as autonomous driving, industrial automation, disaster relief, and welfare robots, and in particular, the stability and efficiency of robots can be maximized in complex and unpredictable environments. Brief explanation of the drawing

[0023] FIG. 1 is a drawing showing a robot according to an exemplary embodiment. FIG. 2 is a conceptual diagram illustrating a robot control system combining on-device-based classical computing and quantum computing according to an exemplary embodiment. FIG. 3 is a conceptual diagram illustrating a robot control system combining on-device-based classical computing and quantum computing according to an exemplary embodiment. FIG. 4 is a plan view showing a robot and an obstacle placed in front according to an exemplary embodiment. FIG. 5 is a plan view showing a robot and an obstacle placed in front according to an exemplary embodiment. FIG. 6 is a diagram illustrating a robot control method combining on-device-based classical computing and quantum computing according to an exemplary embodiment. FIG. 7 is a diagram illustrating a robot control method combining on-device-based classical computing and quantum computing according to an exemplary embodiment. Specific details for implementing the invention

[0024] Hereinafter, exemplary embodiments will be described in detail with reference to the attached drawings. In the drawings below, the same reference numerals denote the same components, and the size of each component in the drawings may be exaggerated for clarity and convenience of explanation. Meanwhile, the embodiments described below are merely exemplary, and various modifications are possible from these embodiments. In the following, terms such as "upper" or "upper" may include not only those located directly above, below, left, or right in contact, but also those located above, below, left, or right in non-contact. A singular expression includes a plural expression unless the context clearly indicates otherwise. Furthermore, when a part is described as "comprising" a component, this means that, unless specifically stated otherwise, it does not exclude other components but may include additional components. The use of the term "above" and similar descriptive terms may apply to both the singular and plural. Unless there is an explicit description of the order of the steps constituting the method, these steps may be performed in a suitable order and are not necessarily limited to the described order. Additionally, terms such as "...part," "module," etc., as used in the specification refer to a unit that performs at least one function or operation, which may be implemented in hardware or software, or a combination of hardware and software. The connections of lines or connecting members between components depicted in the drawings are exemplary representations of functional connections and / or physical or circuit connections, and may be replaced or additionally represented as various functional connections, physical connections, or circuit connections in the actual device. The use of all examples or exemplary terms is merely intended to describe the technical concept in detail and, unless limited by the claims, the scope is not limited by such examples or exemplary terms.

[0025] Hereinafter, a robot control method and system combining classical computing and quantum computing according to exemplary embodiments will be described in more detail.

[0026] FIG. 1 is a drawing showing a robot according to an exemplary embodiment. FIG. 2 is a conceptual drawing showing a robot control system combining on-device-based classical computing and quantum computing according to an exemplary embodiment.

[0027] Referring to FIGS. 1 and 2, a robot control system (100) combining on-device-based classical computing and quantum computing according to an exemplary embodiment may be provided to control the operation of a robot (1).

[0028] The robot (1) may include a main body (11) and a driving unit (12). The main body (11) of the robot (1) may include first to third sensors (21, 22, 23). The first to third sensors (21, 22, 23) may perform the function of acquiring dynamic information about the robot (1) itself and information about an external object (3). A detailed description of the first to third sensors (21, 22, 23) will be provided later.

[0029] The drive unit (12) of the robot (1) can perform the function of moving the robot. The drive unit (12) may operate in contact with the ground by including wheels, or may fly while being spaced vertically away from the ground. If the robot (1) can fly, the robot (1) may be a flying vehicle such as a drone, but is not limited thereto.

[0030] A robot (1) according to an exemplary embodiment may include a first sensor (21). The first sensor (21) may be configured to detect dynamic information of the robot (1) in real time. The dynamic information may include information regarding the movement and posture of the robot (1).

[0031] Information regarding the movement of the robot (1) may include information regarding the velocity and / or acceleration in the three axes (x-axis, y-axis, z-axis) of the robot (1). For example, the first sensor (21) may include an Inertial Measurement Unit (IMU).

[0032] For example, the first sensor (21) can obtain information about the speed and / or acceleration of the robot (1) in the three-axis direction through an IMU. However, the first sensor (21) can also obtain information about the speed and / or acceleration of the robot (1) in the three-axis direction through a device other than the IMU. For example, the first sensor (21) may include x-axis, y-axis, and z-axis gyroscopes.

[0033] As another example, the first sensor (21) can acquire information regarding the yawing, rolling, and pitching of the robot (1). The yawing, rolling, and pitching, etc., may be information regarding the posture of the robot (1). The robot (1) can transmit the information regarding the movement and posture acquired by the first sensor (21) in real time to the first server (4) described later.

[0034] A robot (1) according to an exemplary embodiment may include a second sensor (22). The second sensor (22) may be configured to acquire visual information (220) about an object (3). For example, the visual information (220) may include visual information about an object (3) located along a path where the robot (1) moves.

[0035] The second sensor (22) may include an image sensor, but is not limited thereto. The robot (1) can transmit visual information (220) about an object (3) acquired by the second sensor (22) in real time to the first server (4) described later.

[0036] Visual information (220) can be represented as a pixel-based state vector. For example, visual information (220) can contain color information of each pixel that is part of an image recognized through the second sensor (22). For example, visual information (220) can include the color channel values ​​of the pixels. Visual information (220) can include a vector composed of three components: a red channel, a green channel, and a blue channel. Each of the red channel, the green channel, and the blue channel can have a value from 0 to 255. A vector combining the red channel, the green channel, and the blue channel can also be represented as an RGB color value.

[0037] A robot (1) according to an exemplary embodiment may include a third sensor (23). The third sensor (23) may be configured to acquire distance information (230) in real time. The distance information (230) may include information indicating the distance between the robot (1) and the object (3).

[0038] For example, the third sensor (23) may include a LiDAR sensor configured to acquire distance information (230) about an object (3), but is not limited thereto. Alternatively, for example, the third sensor (23) may include a camera sensor configured to acquire distance information (230) about an object (3). The robot (1) can transmit the distance information (230) about the object (3) acquired by the third sensor (23) in real time to the first server (4) described later.

[0039] Dynamic information, visual information (220), and distance information (230) can be collectively referred to as state information (41). That is, state information (41) can be transmitted in real time from the first to third sensors (21, 22, 23) of the robot (1) to the first server (4).

[0040] A robot control system (100) combining classical computing and quantum computing according to an exemplary embodiment may be configured to control a robot (1). Here, the robot (1) may mean being controlled in real time.

[0041] A robot control system (100) combining on-device-based classical computing and quantum computing can control a single robot (1) so that it can move efficiently along a certain path, but can also control multiple robots (1) so that multiple robots (1) can move efficiently along a certain path. This may vary depending on the design of the robot control system (100) combining on-device-based classical computing and quantum computing.

[0042] A first server (4) may be placed within the robot (1). The first server (4) may be a device that performs calculations using classical computing methods within the robot (1). The first server (4) may be, for example, a Field-Programmable Gate Array (FPGA), but is not limited thereto.

[0043] The state information (41) acquired by the first to third sensors (21, 22, 23) of the robot (1) can be transmitted to the first server (4) placed within the robot (1). That is, the first to third sensors (21, 22, 23) can transmit the state information (41), including dynamic information, visual information (220), and distance information (230), to the first server (4).

[0044] By placing the first server (4) within the robot (1), the speed at which the first server (4) receives status information (41) from the first to third sensors (21, 22, 23) can be improved. That is, the delay time when the first server (4) receives status information (41) from the first to third sensors (21, 22, 23) can be minimized.

[0045] According to an exemplary embodiment, the first server (4) can generate quantum feature data (42) processed into a form of data that can be processed by the second server (5) positioned outside the robot (1), based on dynamic information, visual information (220), and distance information (230) for each of the robots (1). That is, the quantum feature data (42) may be data preprocessed into a form suitable for the second server (5). In other words, the first server (4) can generate quantum feature data (42) by preprocessing state information (41). The quantum feature data (42) may be generated by converting the state information (41) into a quantum state through a quantum feature mapping algorithm.

[0046] Quantum feature data (42) may include dynamic qubit data, distance qubit data, and time qubit data.

[0047] Dynamic qubit data may be generated from dynamic information regarding the movement and posture of the robot (1). For example, dynamic qubit data may include IMU qubit data. IMU qubit data may be quantum feature data (42) processed into a form of data that can be processed by the second server (5) based on dynamic information acquired by the first server (4). In this case, the dynamic information acquired by the first server (4) may be information regarding the velocity and / or acceleration of the robot (1) obtained when the first sensor (21) is an IMU.

[0048] The first server (4) can generate IMU qubit data based on dynamic information. The IMU qubit data may include at least one qubit in which the dynamic information is mapped to a quantum rotation state. Here, the dynamic information may include information about velocity and / or acceleration, and the velocity and acceleration information may be represented as vectors. The velocity and acceleration vectors may be classical data. The mapping may be appropriately encoding the classical data into a quantum state.

[0049] The first server (4) can acquire dynamic information and generate at least one qubit mapped to the rotation angle of a quantum rotation gate. Through this, the first server (4) can model the dynamic information into IMU qubit data.

[0050] Specifically, the first server (4) can convert state information (41) into the rotation angle of the quantum rotation gates (RY, RX, RZ) during the modeling process. The first server (4) can analyze the state information (41) transmitted by the first to third sensors (21, 22, 23), convert it into appropriate rotation angle data, and include it in the quantum feature data (42).

[0051] The first server (4) can generate visual qubit data based on visual information (220). The visual qubit data may include qubits in which the pixel color channel values ​​of the visual information (220) are encrypted with complex coefficients of a quantum state vector. That is, RGB color values ​​are mapped to the qubit state using a phase encoding method, and brightness values ​​are mapped to the qubit state using an amplitude encoding method.

[0052] The first server (4) can convert the time information (220) into qubits encrypted with complex coefficients of a quantum state vector and map them. Through this, the first server (4) can model the time information (220) into time qubit data.

[0053] The visual qubit data may be quantum feature data (42) processed into a form of data that can be processed by the second server (5) based on the visual information (220) acquired by the first server (4).

[0054] Distance qubit data may be generated from distance information (230) between the robot (1) and the object (3). For example, distance qubit data may include LiDAR qubit data. LiDAR qubit data may be quantum feature data (42) processed into a form of data that can be processed by the second server (5) based on distance information (230) acquired by the first server (4).

[0055] For example, distance qubit data may include camera qubit data. The camera qubit data may be quantum feature data (42) processed into a form of data that can be processed by the second server (5) based on distance information (230) acquired by the first server (4).

[0056] The first server (4) can generate LiDAR qubit data based on distance information (230). The LiDAR qubit data may include at least one qubit to which each sample point of the distance information (230) is mapped. Here, the distance information (230) may be expressed as a distance vector from the robot (1) to the object (3). The distance vector may be classical data.

[0057] The first server (4) can generate camera qubit data based on distance information (230). The camera qubit data may include at least one qubit to which each sample point of the distance information (230) is mapped. Here, the distance information (230) may be expressed as a distance vector from the robot (1) to the object (3). The distance vector may be classical data.

[0058] The first server (4) can normalize each sample point of distance information (230) and map it to a quantum state. The sample points may be distance information for each of a plurality of objects (3). Through this, the first server (4) can model the distance information (230) as LiDAR qubit data or camera qubit data.

[0059] The second server (5) can receive quantum feature data (42) from the first server (4) and perform a quantum neural network operation to generate quantum interpretation data (51).

[0060] The QNN can learn quantum feature data (42) by applying parameterized quantum gates and generate quantum interpretation data (51) based on this. The learning process of the QNN aims to minimize the loss function, and the learned parameters can be updated in real time by reflecting the change in state information (41) transmitted by the first to third sensors (21, 22, 23). Through this, the QNN can adaptively generate optimal control values ​​according to changes in the environment in which the robot (1) moves.

[0061] The second server (5) can apply a Quantum Approximate Optimization Algorithm (QAOA) for the training and optimization of the QNN. QAOA is an algorithm advantageous for solving combinatorial optimization problems and can operate by minimizing a cost function within a quantum circuit.

[0062] For example, the first server (4) can generate quantum feature data (42) based on state information (41) transmitted by the first to third sensors (21, 22, 23) and then transmit it to the second server (5). The second server (5) can set the quantum feature data (42) to an initial state and construct a cost function by applying a quantum Hamiltonian. Two Hamiltonians may be used in the QAOA, the first being a cost Hamiltonian representing the difference between the current state of the robot (1) and the target state, and the second being a mixer Hamiltonian that transforms the initial state into an optimized state.

[0063] The second server (5) can update the parameterized quantum gate by repeatedly applying cost Hamiltonian and mixed Hamiltonian. Through this, the optimal movement path, obstacle avoidance strategy, and energy-efficient movement pattern of the robot (1) can be derived. Finally, the quantum interpretation data (51) derived through QAOA can represent the most optimized state and behavior of the robot (1). And based on the quantum interpretation data (51), control commands for the robot (1) can be generated from the first server (4).

[0064] Quantum feature data (42) can be transmitted to a second server (5) and used as input values ​​for a QNN. This method can efficiently connect on-device-based classical data and quantum data.

[0065] The second server (5) can perform computations by applying a Deep Quantum Neural Network (DQNN) structure. The DQNN has a deep structure containing multiple quantum gates, which allows for the optimization of the robot's (1) movements even in complex environments. In particular, by utilizing the structural characteristics of the DQNN, the quantum feature data (42) and the robot's (1) environmental changes can be analyzed more precisely, and quantum interpretation data (51) can be generated accordingly.

[0066] Quantum interpretation data (51) can be transmitted to the first server (4). The first server (4) may be included in the classical computing domain. Conversely, the second server (5) may be included in the quantum computing domain. The first server (4) and the second server (5) interact with each other, and a robot control system (100) combining classical computing and quantum computing can be operated.

[0067] The first server (4) can correct quantum interpretation data (51) by utilizing a classical neural network. The first server (4) can generate first control signal data by analyzing the quantum interpretation data (51) derived through the QNN and verifying it in a classical manner. That is, the first server (4) can receive the quantum interpretation data (51) from the second server (5) and convert it into first control signal data.

[0068] The first server (4) generates a control command for the robot (1) based on the first control signal data, and the movement of the robot (1) can be controlled through the control command. Due to the control command, the speed and direction of the drive unit (12) of the robot (1) can be adjusted, thereby controlling the robot (1). That is, the first server (4) can control the robot (1) by adjusting the speed and direction of the drive unit (12).

[0069] A first server (4) can be placed within the robot (1), and the first server (4) can receive quantum interpretation data (51) and convert it into first control signal data. The first server (4) can generate control commands for the robot (1) based on the first control signal data, and the movement of the robot (1) can be controlled through the control commands. As a result, the robot (1) and the second server (5) can form a robot control system that combines on-device-based classical computing and quantum computing.

[0070] On-device means that specific computations or data processing are performed on the robot (1) itself rather than on a cloud or external server. That is, in the on-device method, a first server (4) placed inside the robot (1) directly performs classical computing operations, and through this, the first server (4) can process the status information (41) transmitted by the first to third sensors (21, 22, 23) in real time and convert it into quantum feature data (42).

[0071] An on-device based computational structure may have the advantage of minimizing data transmission delays with the cloud or external server and improving the real-time control performance of the robot (1). For example, status information (41) obtained from the first to third sensors (21, 22, 23) of the robot (1) can be processed directly at the first server (4) inside the robot (1) without passing through a network, thereby generating first control signal data and immediately transmitting it to the drive unit (12) of the robot (1).

[0072] In addition, the on-device method according to the present embodiment can be designed with consideration for integration with quantum computing. That is, the first server (4) can perform classical computing inside the robot (1), convert state information (41) into quantum feature data (42), and transmit it to the second server (5) located externally. By linking classical data and quantum data in this way using an on-device method, real-time performance and computational optimization can be achieved.

[0073] Accordingly, a robot control system (100) combining on-device-based classical computing and quantum computing can enable rapid data processing within the robot (1) and autonomous operation of the robot (1), thereby allowing multiple robots (1) to be controlled to move along a more efficient path.

[0074] FIG. 3 is a conceptual diagram illustrating a robot control system combining classical computing and quantum computing according to an exemplary embodiment. Except for the items described below, there may be overlap with the description made with reference to FIG. 1 and FIG. 2, and such overlapping descriptions will be omitted.

[0075] Referring to FIGS. 1 and 3, a robot control system (100) combining classical computing and quantum computing may include a robot (1), a second server (5), and a third server (6). The robot (1) may include a first server (4), first to third sensors (21, 22, 23), a main body (11), and a driving unit (12).

[0076] State information (41) obtained from the first to third sensors (21, 22, 23) can be transmitted to the first server (4). The first server (4) can generate quantum feature data (42) based on the state information (41).

[0077] The second server (5) can receive quantum feature data (42) from the first server (4) and perform QNN operations to generate quantum interpretation data (51).

[0078] Quantum interpretation data (51) can be transmitted to a third server (6). The third server (6) can be placed outside the robot (1). The first server (4) and the third server (6) can be included in the classical computing area. The first server (4) and the third server (6) can be different servers. On the other hand, the second server (5) can be included in the quantum computing area. The first server (4), the second server (5), and the third server (6) interact with each other, and a robot control system (100) combining on-device-based classical computing and quantum computing can be operated.

[0079] The third server (6) can correct the quantum interpretation data (51) by utilizing a classical neural network. The third server (6) can generate the second control signal data (43) by analyzing the quantum interpretation data (51) derived through the QNN and verifying it in a classical manner. That is, the third server (6) can receive the quantum interpretation data (51) from the second server (5) and convert it into the second control signal data (43).

[0080] The second control signal data (43) may be a concept similar to the first control signal data described with reference to FIGS. 1 and FIGS. 2. However, while the first control signal data is a signal generated by the first server (4) itself, the second control signal data (43) may be a signal generated by the third server (6).

[0081] The second control signal data (43) can be transmitted to the first server (4) of the robot (1). Based on the second control signal data (43), a control command can be generated at the first server (4) located within the robot (1), and the movement of the robot (1) can be controlled through the control command. Due to the control command, the speed and direction of the drive unit (12) of the robot (1) are adjusted, thereby controlling the robot (1).

[0082] FIG. 4 is a plan view showing a robot and an obstacle placed in front according to an exemplary embodiment. FIG. 5 is a plan view showing a robot and an obstacle placed in front according to an exemplary embodiment.

[0083] Referring to FIGS. 1, 2, 4 and 5, the first to third sensors (21, 22, 23) according to an exemplary embodiment can exchange information with the first server (4) in real time. The first to third sensors (21, 22, 23) can transmit status information (41) to the first server (4), and the first server (4) can generate first control signal data. The first server (4) can control the movement of the robot (1) through the first control signal data.

[0084] As shown in FIG. 4, the robot (1) may be moving in a target direction (61). Here, the target direction (61) of the robot (1) may refer to the direction in which the robot (1) intends to move. Also, there may be an object (3) on the path of the robot (1)'s target direction (61). The object (3) may be an obstacle to the robot (1) moving. In order for the robot (1) to continue moving in the target direction (61), it may be necessary to bypass the object (3).

[0085] The first to third sensors (21, 22, 23) of the robot (1) can transmit status information (41) to the first server (4). Subsequently, through interaction between the first server (4) and the second server (5), first control signal data can be generated at the first server (4). Based on the first control signal data, a control command can be generated at the first server (4). Then, the movement of the robot (1) can be controlled through the control command. That is, the first server (4) can generate and apply a control command for the robot (1) based on the first control signal data. Due to the control command, the speed and direction of the drive unit (12) of the robot (1) are adjusted, thereby controlling the robot (1).

[0086] The robot (1) can move along a detour path (62) by means of a control command. For example, the robot (1) can move to the side of the object (3) by moving in a direction perpendicular to the target direction (61) and then moving back to the target direction (61). In this case, the robot (1) can continue to advance toward the target direction (61) without colliding with the object (3).

[0087] The first control signal data generated by the first server (4) may contain information regarding an optimized avoidance path resulting from the calculation results of the first server (4) and the second server (5). Through the first control signal data, the robot (1) can select the optimal path among the paths to move to a certain destination. The optimal path may refer to a path that takes the least amount of time to move from a certain point to the destination. Alternatively, the optimal path may refer to a path that minimizes encounters with obstacles such as the target object (3). However, the present invention is not limited thereto, and the optimal path may be defined in various ways. The avoidance path included in the first control signal data may vary depending on the definition of the optimal path.

[0088] FIG. 6 is a diagram illustrating a robot control method combining on-device-based classical computing and quantum computing according to an exemplary embodiment.

[0089] Referring to FIGS. 1, FIGS. 2, and FIGS. 6, a robot control method (200) combining on-device-based classical computing and quantum computing according to an exemplary embodiment may be provided to control a robot (1).

[0090] A robot control method (200) combining on-device-based classical computing and quantum computing may include an input step (S1) in which a first server (4) placed within the robot (1) obtains state information (41) from at least one sensor module within the robot (1); a modeling step (S2) in which the first server (4) generates quantum feature data (42) based on the state information (41); an interpretation step (S3) in which a second server (5) placed outside the robot (1) receives the quantum feature data (42) and performs a quantum neural network (QNN) operation to generate quantum interpretation data (51); a conversion step (S4) in which the first server (4) receives the quantum interpretation data (51) and converts it into first control signal data; and an output step (S5) in which the first server (4) generates and applies a control command for the robot based on the first control signal data.

[0091] An input step (S1) may be performed in which a first server (4) placed inside the robot (1) obtains state information (41) from at least one sensor module inside the robot (1). The sensor module may include first to third sensors (21, 22, 23), and may also include other sensors in addition to the first to third sensors (21, 22, 23). The state information (41) may include dynamic information, visual information (220), and distance information (230). The description of the first to third sensors (21, 22, 23), the state information (41), and the first to third sensors (21, 22, 23) may be the same as that described with reference to FIG. 1 and FIG. 2.

[0092] Next, a modeling step (S2) may be performed in which the first server (4) generates quantum feature data (42) based on state information (41). The description of the quantum feature data (42) may be the same as that described with reference to FIGS. 1 and FIGS. 2.

[0093] Next, an interpretation step (S3) may be performed in which the second server (5) receives quantum feature data (42), performs a QNN operation, and generates quantum interpretation data (51). The description of the QNN operation and the quantum interpretation data (51) may be the same as that described with reference to FIGS. 1 and FIGS. 2.

[0094] Next, a conversion step (S4) may be performed in which the first server (4) receives the quantum interpretation data (51) and converts it into the first control signal data. The description of the quantum interpretation data (51) may be the same as the description with reference to FIGS. 1 and FIGS. 2.

[0095] Next, an output step (S5) may be performed in which the first server (4) generates and applies a control command for the robot (1) based on the first control signal data. The description of the control command for the robot (1) may be the same as the description with reference to FIGS. 1 and FIGS. 2.

[0096] A robot control method (200) combining on-device-based classical computing and quantum computing according to the technical concept of the present invention can generate quantum feature data (42) based on state information (41) received from a sensor module within the robot (1) by a first server (4) placed within the robot (1), and can perform QNN computation by receiving the quantum feature data (42). As a result, in the process of generating a signal to control the robot (1), the computation process proceeds by combining the first server (4) related to classical computing and the second server (5) related to quantum computing, so the computation speed can be dramatically increased compared to when computation is performed using only the second server (5). Therefore, real-time path optimization and obstacle avoidance performance can be improved, and data processing efficiency can be increased.

[0097] In addition, since the first server (4) is positioned inside the robot (1), the status information (41) obtained from the first to third sensors (21, 22, 23) of the robot (1) can be processed directly at the first server (4) inside the robot (1) without passing through a network. And the first server (4) can generate first control signal data and immediately transmit it to the drive unit (12) of the robot (1). As a result, the data processing speed can be increased.

[0098] In addition, the reliability of quantum computing can be enhanced by applying noise-resistant quantum algorithms (QAOA).

[0099] For the reasons mentioned above, a robot control method (200) combining classical computing and quantum computing can be utilized in various fields such as autonomous driving, industrial automation, disaster relief, and welfare robots, and in particular, the stability and efficiency of the robot can be maximized in complex and unpredictable environments.

[0100] FIG. 7 is a diagram illustrating a robot control method combining on-device-based classical computing and quantum computing according to an exemplary embodiment. Except for the details described below, there may be overlap with the descriptions made with reference to FIG. 1, FIG. 2, and FIG. 6. Therefore, redundant descriptions will be omitted, and only the differences will be described.

[0101] Referring to FIGS. 1, FIGS. 3, and FIGS. 7, a robot control method (201) combining on-device-based classical computing and quantum computing according to an exemplary embodiment may be provided to control a robot (1).

[0102] A robot control method (200) combining on-device-based classical computing and quantum computing may include an input step (S11) in which a first server (4) placed inside the robot (1) obtains state information (41) from at least one sensor module inside the robot (1); a modeling step (S22) in which the first server (4) generates quantum feature data (42) based on the state information (41); an interpretation step (S33) in which a second server (5) placed outside the robot (1) receives the quantum feature data (42) and performs a quantum neural network (QNN) operation to generate quantum interpretation data (51); a conversion step (S44) in which a third server (6) placed outside the robot (1) receives the quantum interpretation data (51) and converts it into second control signal data (43); and an output step (S55) in which the first server (4) receives the second control signal data (43) and generates and applies a control command for the robot (1).

[0103] The input step (S11) in which a first server (4) placed inside the robot (1) obtains state information (41) from at least one sensor module inside the robot (1), the modeling step (S22) in which the first server (4) generates quantum feature data (42) based on the state information (41), and the interpretation step (S33) in which a second server (5) placed outside the robot (1) receives the quantum feature data (42) and performs a quantum neural network (QNN) operation to generate quantum interpretation data (51) may each be identical to the corresponding steps S1, S2, or S3 in FIG. 6.

[0104] Following step S33, a conversion step (S44) may be performed in which a third server (6) located outside the robot (1) receives quantum interpretation data (51) and converts it into second control signal data (43). The description of the quantum interpretation data (51) may be the same as the description described with reference to FIGS. 1 and FIGS. 3. The description of the second control signal data (43) may be the same as the description described with reference to FIGS. 1 and FIGS. 3.

[0105] Next, an output step (S55) may be performed in which the first server (4) receives the second control signal data (43) and generates and applies a control command for the robot (1). The description of the control command for the robot (1) may be the same as the description with reference to FIGS. 1 and FIGS. 3.

[0106] The above-described embodiments are merely exemplary, and various modifications and equivalent alternative embodiments are possible therefrom for those skilled in the art. Accordingly, the true scope of technical protection according to the exemplary embodiments must be determined by the technical concept described in the following claims.

Claims

Claim 1 A robot control method for controlling a robot using on-device-based quantum computing, comprising: an input step in which a first server disposed within the robot obtains state information from at least one sensor module within the robot; a modeling step in which the first server generates quantum feature data based on the state information; an interpretation step in which a second server disposed outside the robot receives the quantum feature data and performs a quantum neural network (QNN) operation to generate quantum interpretation data; a conversion step in which the first server receives the quantum interpretation data and converts it into control signal data; and an output step in which the first server generates a control command for the robot based on the control signal data and applies it to the robot to control the robot. Claim 2 A robot control method according to claim 1, wherein the sensor module includes an IMU (Inertial Measurement Unit) configured to acquire dynamic information regarding the movement of the robot, and the state information includes the dynamic information. Claim 3 A robot control method according to claim 2, wherein the dynamic information is represented as a vector, the quantum feature data includes IMU qubit data, and in the modeling step, the first server generates the IMU qubit data based on the dynamic information, and the IMU qubit data includes at least one qubit to which the dynamic information is mapped to a quantum rotation state. Claim 4 A robot control method according to claim 1, wherein the sensor module includes a camera sensor configured to acquire distance information for a specific target, and the state information includes the distance information. Claim 5 A robot control method according to claim 4, wherein the distance information is represented as a vector, the quantum feature data includes camera qubit data, and in the modeling step, the first server generates the camera qubit data based on the distance information, and the camera qubit data includes at least one qubit to which each sample point of the distance information is mapped. Claim 6 In paragraph 5, the robot comprises: a main body part including the sensor module; and a driving part responsible for the movement of the main body part, and The first server above controls the robot by controlling the speed and direction of the drive unit in the output step, Robot control method. Claim 7 A robot control method according to claim 1, wherein the first server comprises a Field-Programmable Gate Array (FPGA). Claim 8 A robot control method for controlling a robot using on-device-based quantum computing, comprising: an input step in which a first server disposed within the robot obtains state information from at least one sensor module within the robot; a modeling step in which the first server generates quantum feature data based on the state information; an interpretation step in which a second server disposed outside the robot receives the quantum feature data and performs a quantum neural network (QNN) operation to generate quantum interpretation data; a conversion step in which a third server disposed outside the robot receives the quantum interpretation data and converts it into control signal data; and an output step in which the first server receives the control signal data, generates a control command for the robot, and applies it to the robot to control the robot. Claim 9 A robot control method according to claim 8, wherein, in the modeling step, the first server generates data converted into rotation angles of quantum rotation gates (RY, RX, RZ) based on the state information and provides the converted data to the second server. Claim 10 A robot control method according to claim 9, wherein in the interpretation step, the second server performs the QNN operation by applying a Deep Quantum Neural Network (DQNN) structure.

Images