Manipulation system and fluidic chip
By designing multiple first flow paths and second flow paths in the control system, using liquid movement to achieve precise position control of the object of operation in the operating area, the problems of unfixed liquid shape and uneven pressure are solved, and the operation object is improved The position stability and movement accuracy of objects.
Patent Information
- Application Number
- CN202380034689.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Priority Date
- 2022-04-26
- Filing Date
- 2023-04-14
- Publication Date
- 2024-11-26
AI Technical Summary
The prior art is difficult to effectively control the position of the operating object in the liquid, especially because the liquid is not fixed in shape and uneven pressure action lead to difficulty in controlling the position of the operating object.
An operation system is designed, including an operating area, a plurality of first flow paths and a second flow path, and precise position control of the object of operation is realized in the operating area by moving the liquid. The first flow path is connected to the operating area, the liquid moves in the first flow path and flows into the operating area, and the second flow path is connected to a plurality of first flow paths to uniformize the pressure and ensure the position of the operating object is stable.
The precise position control of the object of operation is realized, reducing local pressure changes when the liquid moves, and improving the stability and movement accuracy of the object of operation in the operation area.
Smart Images

Figure CN119032310A_ABST
Abstract
Description
Technical Field
[0001] The invention relates to a manipulation system and a fluid chip. Background Art
[0002] In recent years, the development of a technology called MicroTAS (Micro Total Analysis Systems) that uses tiny flow paths to perform biochemical processes has been actively carried out. MicroTAS is used, for example, for the analysis of cell-to-cell interactions and the construction of multi-cell array systems in the field of regenerative medicine. In the past, a technology was known that moves an object of operation in a liquid to a target position by moving the liquid in a flow path containing the liquid (for example, see Patent Document 1).
[0003] Prior art literature
[0004] Patent Literature
[0005] Patent Document 1: Japanese Patent Application Publication No. 2021-185782 Summary of the invention
[0006] Problems to be solved by the invention
[0007] Generally speaking, it is difficult to control the movement of an object in a liquid to a desired position by moving the liquid. One reason for the high difficulty of control is that the object is moved indirectly by using the liquid. Since the shape of the liquid is not fixed, the behavior of the liquid or the object around it is not easy to predict when simply repeating the action of applying a default pressure to the liquid, and the position control of the object becomes difficult.
[0008] The present invention has been made in view of the above-mentioned problems, and an object of the present invention is to provide a technology for easily controlling the position of an operation object.
[0009] Means for solving problems
[0010] In order to achieve the above-mentioned purpose, the manipulation system of the present invention comprises: an operation area, which contains a liquid and an operation object whose position is operated in the liquid; a first flow path, which is a plurality of flow paths connected to the operation area and containing the liquid; a liquid control unit, which moves the liquid in the first flow path; and a second flow path, which is a series of flow paths containing the liquid and is respectively connected to the plurality of the first flow paths.
[0011] That is, the operating area contains a liquid and an operating object in the liquid, and in the first flow path connected to the operating area, the liquid in the first flow path can be moved. A plurality of first flow paths are provided, and through the movement of the liquid in the first flow path, the liquid flows into the operating area from the first flow path, and the liquid flows out from the operating area to the first flow path. Therefore, in conjunction with the movement of the liquid in the first flow path, the liquid also moves in the operating area, and as a result, the operating object moves in the operating area. In this structure, when the liquid in the first flow path moves, the pressure changes locally and eventually returns to a stable state with uniform pressure. If such a pressure change occurs locally or the pressure change remains for a long time, control becomes difficult when the movement of the liquid is repeated in the first fluid.
[0012] Therefore, a second flow path is provided in the manipulation system. The second flow path is a series of flow paths containing liquid, and is connected to a plurality of first flow paths respectively. Therefore, the second flow path forms a flow path connecting a plurality of first flow paths to each other. As a result, the pressures in the plurality of first flow paths are uniform to each other, and it is difficult to generate a local pressure difference. In addition, even if the pressure changes due to the movement of the liquid in a certain first flow path, the pressure change is uniform as quickly as possible. Therefore, in the case of repeatedly moving the liquid to manipulate the position of the manipulated object, the position of the manipulated object can be easily controlled. BRIEF DESCRIPTION OF THE DRAWINGS
[0013] Figure 1A is a diagram showing a schematic structure of a control system, Figure 1B is a diagram showing key parts of the manipulation system.
[0014] Figure 2A is a diagram showing an example of an image captured by a microscope, Figure 2B is a diagram showing the shape of the operation area, FIG. 2C to FIG. 2E It is a diagram showing the structure of a switching unit.
[0015] Figure 3 is a block diagram showing the structure of the operating system.
[0016] Figure 4A is a diagram showing a simulation model, Figure 4B is a diagram illustrating a machine learning model.
[0017] Figure 5 is a diagram illustrating reinforcement learning.
[0018] Figure 6 is a flow chart of the reinforcement learning process.
[0019] Figure 7 It is a flowchart of the manipulation process.
[0020] FIG. 8A to FIG. 8C It is a diagram showing an example of moving an operation object to a target position.
[0021] Fig. 9 It is a diagram showing a configuration example for introducing an operation object into an operation area.
[0022] Fig.10 is a diagram showing a substrate implementing a manipulation system having multiple operation areas.
[0023] Fig.11 is an enlarged view showing a substrate implementing a manipulation system having multiple operation areas. DETAILED DESCRIPTION
[0024] Here, embodiments of the present invention will be described in the following order.
[0025] (1) Structure of the control system:
[0026] (2) Machine learning processing using AI (artificial intelligence):
[0027] (3) Manipulation and processing:
[0028] (4) Other implementation methods:
[0029] (1) Structure of the control system:
[0030] Figure 1A is a diagram showing a schematic structure of a control system, Figure 1B 1 is a diagram showing the main parts of the manipulation system 10. The manipulation system 10 of the present embodiment includes a computer 11, a microscope 12, a signal conversion interface 13, a solenoid valve 14, a pressure supply pipe 15, and a microfluidic chip 16.
[0031] In the present embodiment, the computer 11 is connected to the microscope 12 and the signal conversion interface 13. The computer 11 obtains image data output by the microscope 12. In addition, the computer 11 outputs a digital signal, which is input to the signal conversion interface 13. The details are described later. The microscope 12 has an image sensor capable of capturing an image contained in a field of view. In the present embodiment, the microscope 12 can capture an image with the operating area 60 of the microfluidic chip 16 described later as a field of view. Image data representing the captured image is transmitted to the computer 11.
[0032] The signal conversion interface 13 is a device that is connected to the computer 11 and the solenoid valve 14, and outputs a default current to the solenoid valve 14 according to the digital signal output by the computer 11. The solenoid valve 14 is connected to a compressor and a pressure supply pipe 15 (not shown), and the connection between the compressor and the pressure supply pipe 15 is switched by opening and closing the solenoid valve 14. Therefore, the solenoid valve 14 can switch the internal pressure of the pressure supply pipe 15 between a high state and a low state. That is, the solenoid valve 14 adjusts the internal pressure of the pressure supply pipe 15 according to the current supplied from the signal conversion interface 13. In this embodiment, the solenoid valve 14 makes the internal pressure of the pressure supply pipe 15 high during the period when the current output from the signal conversion interface 13 is at a high level, and makes the internal pressure of the pressure supply pipe 15 low during the period when the current output from the signal conversion interface 13 is at a low level.
[0033] The pressure supply pipe 15 is connected to the microfluidic chip 16. The microfluidic chip 16 is provided with an air pressure regulating unit 70 (described later). Therefore, the solenoid valve 14 can switch the internal pressure of the air pressure regulating unit 70 connected to the pressure supply pipe 15 between a high state and a low state. Figure 1A As shown, there are six solenoid valves 14 and pressure supply pipes 15. Figure 1A In the example, one of them is indicated by a number.
[0034] The microfluidic chip 16 is provided with a liquid-enclosing portion enclosing a liquid and a structure enclosing a gas. Figure 1B 2 shows the liquid inclusions and structures provided in the microfluidic chip 16. The microfluidic chip 16 is a rectangular parallelepiped having two sides larger than the other four sides, that is, a plate-like shape, and the liquid inclusions and structures are formed on the sides with larger areas than the other sides. Figure 1B The microfluidic chip 16 is shown as viewed from a direction perpendicular to a surface having a larger area than the others.
[0035] The microfluidic chip 16 is provided with a first flow path 40, a second flow path 50, an operation area 60, and a gas pressure adjustment unit 70. In this embodiment, the liquid is water and the gas is air. In addition, the operation object is introduced into the first flow path 40 or the operation area 60, but the structure for introducing the operation object is different. Figure 1B Omitted in .
[0036] The first flow path 40, the second flow path 50, and the operation area 60 are flow paths and areas containing liquid. The first flow path 40 is a straight hollow space, and one end is connected to the operation area 60. In addition, the other end of the first flow path 40 is connected to the second flow path 50. In this embodiment, three first flow paths 40 are provided. In this embodiment, the first flow path 40 is on the widest plane ( Figure 1B The two electrodes are evenly spaced in a plane parallel to the paper surface. Figure 1BIn the embodiment, the angle between adjacent first flow paths 40 is 120°.
[0037] The operation area 60 is an area containing liquid and used to configure the operation object at a predetermined position, and has an internal space that allows the operation object to move in a plane. In this embodiment, the operation area 60 is formed on the largest surface of the microfluidic chip 16, and the shape of the internal space of the operation area 60 enables the operation object to move in a plane along a direction parallel to the largest surface of the microfluidic chip 16. That is, the operation area 60 has an internal space that expands in a direction parallel to the surface at a certain depth (for example, a depth of less than 30 μm) on the largest surface of the microfluidic chip 16.
[0038] Furthermore, the shape of the internal space of the operation region 60 when viewed from a direction perpendicular to the largest surface of the microfluidic chip 16 is triangular. Figure 2A FIG. 1 shows an example of an image including the operation area 60 captured by the microscope 12. The operation area 60 is a space surrounded by three sides. Figure 2B The dotted lines shown in the figure indicate the extended and intersecting vertices V1 to V3, and the operation area 60 has a triangular internal space having the vertices V1 to V3. That is, in the present embodiment, if the inner wall of the internal space of the operation area 60 is extended, it becomes a triangle.
[0039] The first flow path 40 is connected to the vertex portion of the triangle formed by the operating area 60. In the present embodiment, three first flow paths 40 are formed in a direction substantially parallel to a straight line extending from the center of gravity of the triangle formed by the operating area 60 toward the vertex. In the present embodiment, the first flow path 40 has the same depth as the operating area 60 and a width smaller than one side of the triangle formed by the operating area 60. In the present embodiment, the first flow path 40 is a flow path extending in a straight line. In the present embodiment, the first flow path 40 is a hollow space having a rectangular cross-section. One end portion of each first flow path is as shown in FIG. Figure 2A As shown, it is connected to the operating area 60, and the other end is as shown Figure 1B It is shown connected to the second flow path 50. In the present specification, the depth is the length in the direction perpendicular to the largest surface of the microfluidic chip 16, and the width is the direction perpendicular to the depth direction and the short side direction.
[0040] like Figure 1BAs shown, the second flow path 50 includes a storage portion 51 and a connecting portion 52. In the present embodiment, the shape of the storage portion 51 when viewed from a direction perpendicular to the largest surface of the microfluidic chip 16 is circular. The storage portion 51 is connected to the first flow path 40, and its depth is deeper than the first flow path 40 (for example, 3 mm to 5 mm). In addition, the diameter of the circle formed by the storage portion 51 is larger than the width of the first flow path 40. In the present embodiment, three storage portions 51 are provided, all of which have the same depth and shape. In addition, in the present embodiment, the rectangular microfluidic chip 16 is arranged on a horizontal plane in such a way that the liquid levels of the three storage portions 51 are at the same height.
[0041] The connecting portion 52 is formed between two adjacent storage portions 51 in a manner connecting these storage portions 51. In the present embodiment, the connecting portion 52 is a hollow space having a rectangular cross-section. The depth of the connecting portion 52 is the same as that of the first flow path 40. The width is smaller than the diameter of the circle formed by the storage portions 51. Therefore, the storage portion 51 is larger than the connecting portion 52. In addition, in the present embodiment, the storage portions 51 and the connecting portion 52 exist around the triangle formed by the operating area 60, the storage portion 51 is equivalent to the vertex of the triangle larger than the operating area 60, and the connecting portion 52 is equivalent to the side of the triangle larger than the operating area 60. Therefore, the second flow path 50 can also be regarded as an annular flow path formed around the operating area 60 by connecting the storage portions 51.
[0042] The air pressure regulating part 70 is a structure including a passage containing gas. The air pressure regulating part 70 includes a switching part 71, a gas passage 72, and a connecting part 73. The connecting part 73 is a portion to which the pressure supply pipe 15 is connected. In the present embodiment, six connecting parts 73 are formed, and six pressure supply pipes 15 are connected to the connecting parts 73, respectively. The gas passage 72 is a hollow space connected to the connecting part 73, and connects the space directly below the switching part 71 formed directly under the first flow path 40 with the connecting part 73. In addition, as Figure 1B As shown, there are six air pressure adjustment parts 70, but in Figure 1B In the example, mark one of them with a number.
[0043] The switch unit 71 includes a mechanism capable of switching the first flow path 40 between a closed state and an unclosed state. In the present embodiment, two switch units 71 are provided for each of the three first flow paths 40 . FIG. 2C to FIG. 2E It is a diagram for explaining the structure of the switching unit 71 and the adjustment of the internal pressure of the first flow path 40 by the switching unit 71 . Figure 2C The switching unit 71 is shown as viewed from a direction perpendicular to the largest surface of the microfluidic chip 16 . Figure 2D , Figure 2E The diagram shows a state where the switching unit 71 is viewed from a direction parallel to the largest surface of the microfluidic chip 16 .
[0044] In this embodiment, the switch unit 71 includes a diaphragm 71a, and a space 71b is provided directly below the diaphragm 71a. The first flow path 40 and the space 71b are separated by a flexible film 40a, and the portion of the film 40a where the first flow path 40 and the space 71b overlap is the diaphragm 71a.
[0045] in addition, FIG. 2C to FIG. 2E The dark grey color shown is the diaphragm 71a. Figure 2D , Figure 2E The light grey color shown is the liquid contained in the first flow path 40. The space 71b is a rectangular parallelepiped cavity connected to the gas passage 72. The material of the diaphragm 71a is not limited, but in the present embodiment, it is made of resin.
[0046] As described above, the diaphragm 71 a is a portion where the first flow channel 40 and the space 71 b overlap, and has a rectangular shape when viewed from a direction perpendicular to the largest surface of the microfluidic chip 16 .
[0047] When the internal pressure of the pressure supply pipe 15 is high or low, the internal pressure of the space 71b is also high or low. The diaphragm 71a is a flexible film, and therefore its shape changes according to the internal pressure of the space 71b. Figure 2D The shape of the space 71b in a state where the internal pressure is low is shown. Figure 2E The shape of the space 71b is in a state where the internal pressure is high. Figure 2E As shown in FIG. 1 , when the internal pressure of the space 71b becomes high, the diaphragm 71a expands toward the first flow path 40 to close the first flow path 40. That is, the first flow path 40 becomes closed. Figure 2D As shown, when the internal pressure of the space 71b is low, the diaphragm 71a becomes flat, and the liquid can flow in the first flow path 40. That is, the first flow path 40 is opened.
[0048] Thus, when the first flow path 40 switches between the closed state and the unclosed state, the internal pressure of the first flow path 40 changes during the change process, and the liquid flows. Figure 2D The status shown switches to Figure 2E In the state shown, the liquid flows toward the left and right sides in the first flow path 40. In this embodiment, two switching parts 71 are provided for one first flow path 40. Therefore, by combining the actions of the two switching parts 71, the liquid can be moved from the first flow path 40 to the operation area 60, and then the liquid can be moved from the operation area 60 to the first flow path 40.
[0049] In the present embodiment, one switching unit 71 can make the first flow path 40 realize two states, a closed state and an unclosed state. In addition, in the present embodiment, two switching units 71 are respectively provided for the three first flow paths 40, so a total of six switching units 71 are provided. The state of the closed first flow path 40 corresponds to 1, and the unclosed state corresponds to 0, and the state realized by the total of six switching units 71 can be expressed by a 6-digit numerical value. Hereinafter, the state realized by the total of six switching units 71 and expressed by a 6-digit numerical value is referred to as the state of the switching unit 71.
[0050] In the present embodiment, by controlling the above devices using the computer 11 , it is possible to move the operation object to a predetermined target position relative to the operation area 60 . Figure 3 This is a diagram for explaining the structure of a computer 11. The computer 11 includes a control unit 20 and a storage medium 30. The control unit 20 includes a CPU, RAM, ROM, and GPU (not shown), and can execute various programs stored in the storage medium 30. The control unit 20 and the storage medium 30 may be composed of an integrated computer, or may be a structure in which at least a part of the device is a separate device connected by various cables. Of course, an output unit for outputting images, sounds, etc., an input unit for inputting user instructions, etc., may also be provided. The output unit or the input unit may also be connected via various interfaces.
[0051] In this embodiment, the control unit 20 can execute the manipulation program 21 and the machine learning program 22. The manipulation program 21 causes the control unit 20 to execute the function of moving the manipulation object to the target position based on the image of the manipulation area 60 captured by the microscope 12. When the manipulation program 21 is executed, the control unit 20 functions as the imaging unit 21a and the position control unit 21b.
[0052] In the present embodiment, the movement of the operation object is performed based on the result of the machine learning performed in advance. The machine learning program 22 causes the control unit 20 to perform a function of machine learning on a model for determining a control target for the switching unit 71 based on an image of the operation area 60 captured by the microscope 12, etc. If the machine learning program 22 is executed, the control unit 20 functions as a machine learning unit 22a.
[0053] (2) Machine learning processing using AI (artificial intelligence):
[0054] In this embodiment, the model for determining the control target for the switching unit 71 is machine-learned, but in order to support the machine learning, the model for simulating the movement of the operation object caused by the liquid is also machine-learned. Here, the former is simply referred to as the machine learning model 30b, and the latter is referred to as the simulation model 30c.
[0055] In the present embodiment, the simulation model 30c outputs information indicating the position of the operation object and the current state and the next state of the switching unit 71 as input data, and outputs information indicating the next position of the operation object as output data. Figure 4A 30c is a diagram schematically showing a simulation model 30c. Figure 4A In FIG. 3 , input data is displayed on the left side, and output data is displayed on the right side, and it is shown that if input data is input to the simulation model 30c, output data is output from the right side. Figure 4A In , (Xm, Ym) represents the position (coordinate) of the operation object. V1 to V6 represent the state (1 or 0) of each of the six switching units 71. Arrangement of V1 to V6 results in a 6-bit numerical value representing the state of the switching unit 71.
[0056] The reference numeral t is added to the reference numerals indicating the position of the operation object and the state of the switching unit 71. The reference numeral t is a reference numeral for identifying the number of times the switching unit 71 has been operated. Figure 4A As shown, in the simulation model 30c, the position of the operation object after the t-th action is input. In addition, in the simulation model 30c, the state of the switching unit 71 after the t-th action and the state of the switching unit 71 after the t+1-th action are input. Then, the position of the operation object after the t+1-th action is input. Therefore, according to the simulation model 30c, after the t-th action, the position of the operation object is a certain position, and the switching unit 71 is in a certain state. Then, by changing the switching unit 71 to the certain state, it is possible to simulate which position the operation object is displaced to.
[0057] Such a simulation model 30c can be machine-learned by various methods. For example, by optimizing a model composed of a neural network through machine learning, the simulation model 30c can be generated. Specifically, when the state of the switching unit 71 is a determined state, in the manipulation system 10, the position (coordinates) of the operation object is determined by photographing through the microscope 12. In this state, the computer 11 outputs a signal to the signal conversion interface 13 to change the state of the switching unit 71.
[0058] Then, the microscope 12 is used to take pictures again to determine the position of the operation object corresponding to the change in state. If relevant information of a series of actions is collected, the information representing the current position of the operation object and the current state and the next state of the switching unit 71 can correspond to the position of the operation object after the state change. This group of data is set as teacher data, and teacher data of the number of samples that meet the neural network learning is collected. If the teacher data is collected, the control unit 20 performs the known machine learning processing through the function of the machine learning unit 22a to optimize the neural network. The model obtained in this way becomes the simulation model 30c, which is recorded in the storage medium 30.
[0059] The machine learning of the machine learning model 30b in the environment is implemented in a state where the simulation model 30c has been created. The machine learning of the machine learning model 30b can also be implemented by various methods. Here, an example of performing machine learning on the machine learning model 30b by reinforcement learning is described. Figure 5 This is a diagram illustrating the learning of the machine learning model 30b using a reinforcement learning model consisting of an agent and an environment. Figure 5 The agent shown is equivalent to the function of selecting an action a that can be taken in the current state s. The environment is used to determine the next state s' according to the action a selected by the agent and the current state s, and to determine the reward r' according to the action a, state s, and state s'. The environment only needs to be defined so that the next state s' can be determined according to the action a and state s, and then the reward r' can be determined. Therefore, the state can be actually observed, or the state can be determined in a virtual environment. In this embodiment, the virtual environment is defined, for example, as follows: In the computer 11, the operation area 60 is virtually reproduced in a three-dimensional space set virtually. Moreover, the position of the operation object in the virtual three-dimensional space is input to the simulation model 30c, thereby simulating the state change in the virtual environment. In this way, if reinforcement learning is performed in a virtual environment using the simulation model 30c, the optimization of the machine learning model 30b can be achieved at a high speed compared to an environment that requires actual observation. In addition, the machine learning model 30b is optimized in a virtual environment, and after the operation of the operating system 10 using the machine learning model 30b is started, when the machine learning of the machine learning model 30b is further performed, the information observed in the real environment can also be used.
[0060] The algorithm of reinforcement learning can adopt various methods. Here, Q learning is explained by an example. In Q learning, action a is selected according to the current state s, and then the action value function Q(s, a) representing the expected benefit when action a is selected is assumed according to the default policy. Then, if the action value function Q(s, a) is optimized, the optimal policy is obtained. The policy can be defined in various ways. For example, if it is a greedy policy, it becomes a policy of selecting an action that maximizes the action value function Q(s, a). Under the assumption of such a policy, if the action value function Q(s, a) is optimized, the policy of selecting action a so that the action value function Q(s, a) becomes the maximum in state s is the optimal policy. The machine learning model 30b can be defined in various ways. Here, it is defined to include the action value function Q(s, a), and the data representing the action a that maximizes the action value function Q(s, a) becomes the output data.
[0061] In this embodiment, the state s is the position of the operation object, the state of the switching unit 71, and the target position of the operation object. The action a is the action of the switching unit 71 required to bring the operation object closer to the target position, that is, the next state of the switching unit 71. Therefore, the machine learning model 30b is defined so that information indicating the position of the operation object, the state of the switching unit 71, and the target position of the operation object is used as input data, and information indicating the action of the switching unit 71 required to bring the operation object closer to the target position is output as output data.
[0062] Figure 4B is a diagram schematically showing a machine learning model 30b. Figure 4A In addition, (Xg, Yg) is the position (coordinate) of the target position of the operation object. This position is the coordinate in the virtually constructed environment in a virtual environment, and is the position in the image captured by the microscope 12 in a real environment. Figure 4B As shown, the position of the operation object at the tth time, the state of the switching unit 71 after the tth action, and the target position of the operation object are input to the machine learning model 30b. These positions and states input to the machine learning model 30b are equivalent to Figure 5 The output of the machine learning model 30b is the state of the switching unit 71 after the t+1th action, which is equivalent to Figure 5 Therefore, if the tth time is taken as the current time, in the machine learning model 30b, the next action a (the state of the next switching unit 71) is output according to the current state s (the current state and target position of the operation object or switching unit 71).
[0063] Figure 6 is a flowchart showing the key parts of the reinforcement learning process used to generate the machine learning model 30b described above. Figure 6 In the embodiment, the processing of machine learning on the machine learning model 30b in the virtual environment is described. The information indicating the optimized simulation model 30c, the unoptimized machine learning model 30b, and the virtually constructed environment is recorded in the storage medium 30, and the implementation is performed. Figure 6The reinforcement learning process shown. If the reinforcement learning process starts, the control unit 20 initializes the virtual environment constructed for learning through the function of the machine learning unit 22a (step S100). That is, the control unit 20 sets the values of the variables used in the process of the reinforcement learning process to the initial values. Specifically, the control unit 20 sets the position of the operation object and the initial state of the switching unit 71, and then virtually sets the target position. The initial value of the position of the virtual operation object is the initial position during learning, and the virtually predetermined position becomes the initial position. The initial position can exist in multiple places according to the nature of the task, and each initial position is processed in each episode repeated in the learning loop process of steps S100 to S130.
[0064] The virtual target position can be set in the virtually reproduced operation area 60, and any position in the operation area 60 can be used as the virtual target position. In each scenario repeated in the loop processing of steps S100 to S130, the target position can be changed or fixed. In the operating system 10, if the operation object is moved toward a determined target position, the target position is preferably fixed to the determined target position in the same scenario (S100 to S125). Not limited to a determined target position, if multiple positions can be the target position, then in the reinforcement learning process, it is also preferred to use multiple positions as the initial values of the target position, and machine learning progresses. In the process of repeating the loop processing of steps S100 to S130, the initial value of the state of the switching unit 71 can be fixed to a determined state (for example, 000000, etc.) or can be changed.
[0065] Next, the control unit 20 inputs the state s to the machine learning model 30b through the function of the machine learning unit 22a (step S105). That is, the control unit 20 inputs the current state (the position of the operating object (Xmt, Ymt), the state V1t~V6t of the switching unit 71, and the target position (Xg, Yg)) to the machine learning model 30b being learned. As a result, the state V1t+1~V6t+1 of the switching unit 71 is output as the next action. In addition, in the process of machine learning, the state V1t+1~V6t+1 of the switching unit 71 as the next state may not be suitable, but if machine learning is carried out, it will be gradually optimized.
[0066] Next, the control unit 20 simulates the position of the operation object corresponding to the action through the function of the machine learning unit 22a (step S110). That is, the control unit 20 inputs the position of the operation object (Xmt, Ymt), the current state V1t to V6t of the switching unit 71, and the next state V1t+1 to V6t+1 of the switching unit 71 obtained in step S105 to the simulation model 30c recorded in the storage medium 30. As a result, the position of the next operation object (Xmt+1, Ymt+1) is output. In this way, the movement phenomenon of the operation object equivalent to it can be reproduced in the virtual environment without using the real environment.
[0067] Next, the control unit 20 performs empirical observation (step S115) through the function of the machine learning unit 22a. That is, the control unit 20 determines the state or reward corresponding to the action in steps S105 and S110. Specifically, the control unit 20 determines the reward based on the target position (Xg, Yg) and the position of the next operation object (Xmt+1, Ymt+1). The reward can also be determined by various methods. Specifically, when the distance between the target position (Xg, Yg) and the position of the next operation object (Xmt+1, Ymt+1) is within a specified distance, a reward is defined to obtain a positive reward. In addition, for example, the smaller the distance between the target position (Xg, Yg) and the position of the next operation object (Xmt+1, Ymt+1), the greater the reward, or the greater the distance, the greater the negative reward. In any case, if a reward is defined, for example, by using a known definition, the action value function Q(s, a) can be defined.
[0068] Next, the control unit 20 updates the machine learning model 30b through the function of the machine learning unit 22a (step S120). The machine learning model 30b includes an action value function Q(s, a). In reinforcement learning, the purpose is to optimize the action value function Q(s, a) that is not optimized initially. Moreover, in the present embodiment, the action value function Q(s, a) is defined by a multi-layer neural network, and the multi-layer neural network is optimized during the reinforcement learning process. The optimization of the multi-layer neural network can be implemented by various well-known algorithms, for example, a structure that can optimize the multi-layer neural network by an objective function that minimizes the TD error (Temporal Difference) can be adopted.
[0069] If the action value function Q(s, a) is updated according to a known algorithm, that is, the machine learning model 30b is updated, the control unit 20 determines whether the current scenario processing is completed through the function of the machine learning unit 22a (step S125). In the present embodiment, a scenario is from the initial observation of the position of the operation object to the operation object reaching the target position (Xg, Yg), or until the arrival at the target position (Xg, Yg) fails. The failure condition only needs to be determined in advance. For example, a structure can be adopted in which the failure condition is deemed to be satisfied when the state s changes more than the default number of times without the operation object reaching the target position (Xg, Yg). In addition, a structure can also be adopted in which the failure condition is deemed to be satisfied when the operation object reaches the outside of the operation area 60.
[0070] In step S125, if it is not determined that the current scenario processing is completed, the control unit 20 repeats the processing after step S105. In step S125, if it is determined that the current scenario processing is completed, the control unit 20 determines whether the machine learning is completed (step S130). The completion condition of machine learning can also be determined in various ways. For example, for all multiple environments prepared in advance, a structure can be adopted in which the completion condition of machine learning is deemed to be satisfied when the processing of steps S105 to S125 is completed. In addition, the satisfaction of the completion condition can be determined by an indicator for evaluating whether the machine learning model 30b is fully optimized, such as the task achievement rate.
[0071] In step S130, if it is not determined that the machine learning is completed, the control unit 20 repeats the processing after step S100 through the function of the machine learning unit 22a. In the present embodiment, at this time, the environment is initialized so that it becomes an initial state different from the previous scenario. In step S130, if it is determined that the machine learning is completed, the control unit 20 ends the reinforcement learning process. As a result, an optimized machine learning model 30b is obtained.
[0072] (3) Manipulation and processing:
[0073] Next, the manipulation process for placing the manipulation object at the target position will be described. Figure 7is a flowchart showing the manipulation process. The manipulation process is implemented by the control unit 20 executing the manipulation program 21 when the optimized machine learning model 30b is recorded in the storage medium 30. If the manipulation program 21 is executed, the control unit 20 functions as a shooting unit 21a and a position control unit 21b. The function of the shooting unit 21a is to output a control signal to the microscope 12 and obtain image data representing an image within the field of view of the microscope 12. The function of the position control unit 21b is to output a control signal to the signal conversion interface 13 so that the signal conversion interface 13 outputs a current.
[0074] The user issues an instruction to start the manipulation process through an input unit (not shown). When the manipulation process starts, the control unit 20 receives the setting of the target position (step S200). That is, the control unit 20 obtains the coordinates input by the user through an input unit (not shown) and regards them as the target position in the operation area 60.
[0075] Next, the control unit 20 introduces the operation object through the function of the position control unit 21b (step S205). That is, the user operates the structure not shown in the figure to introduce the operation object into the first flow path 40, the second flow path 50 or the operation area 60. The control unit 20 controls the signal conversion interface 13 through the function of the position control unit 21b to operate the switching unit 71, thereby performing a process for moving the operation object into the operation area 60. The process for moving the operation object into the operation area 60 is, for example, a process for moving the operation object in the first flow path 40 to the side of the operation area 60, and the action required in the switching unit 71 is a predetermined simple action.
[0076] Next, the control unit 20 determines whether the operation object exists in the operation area through the function of the imaging unit 21a (step S210). That is, the control unit 20 controls the microscope 12 and obtains image data, and extracts the operation object based on the image data. In addition, the control unit 20 determines whether the position of the operation object is within the operation area 60. In the case where it is not determined that the operation object exists in the operation area, the control unit 20 repeats the processing after step S205. However, in the case where step S205 is repeated, the operation object is not newly introduced, and processing for moving the operation object relative to the operation area 60 is performed.
[0077] In step S210, if it is determined that the operation object exists in the operation area, the control unit 20 performs the processing after step S210. In addition, the state in which the operation object is determined to exist in the operation area is, for example, Figure 2A The status shown. Figure 2A In FIG. 6A , a rectangular frame is added to the image of the operation object, and a rectangular frame is also shown at the target position, but these rectangular frames overlap with the image and are not objects that actually exist in the operation area 60 .
[0078] In step S215, the control unit 20 performs state observation. That is, the control unit 20 controls the microscope 12 and obtains image data through the function of the imaging unit 21a, and obtains the current position (Xmt, Ymt) of the operation object based on the image data. In addition, the control unit 20 outputs a control signal to the signal conversion interface 13 to obtain the current state (V1t~V6t) of the switching unit 71. Furthermore, the control unit 20 obtains the target position (Xg, Yg) set in step S200.
[0079] Next, the control unit 20 determines whether the operation object has reached the target position (step S220). That is, if the current position (Xmt, Ymt) of the operation object obtained in step S215 is within a predetermined distance from the target position (Xg, Yg) set in step S200, the control unit 20 deems that the operation object has reached the target position. In step S220, if it is determined that the operation object has reached the target position, the control unit 20 ends the manipulation process.
[0080] In step S220, if it is determined that the operation object has not reached the target position, the control unit 20 inputs the position (Xmt, Ymt) of the operation object, the current state (V1t to V6t) of the switching unit 71, and the target position (Xg, Yg) of the operation object to the machine learning model 30b through the function of the position control unit 21b (step S225). That is, the control unit 20 inputs the current state to the machine learning model 30b. As a result, the machine learning model 30b inputs the next action, that is, the state (V1t+1 to V6t+1) of the switching unit 71 to be set next.
[0081] Based on the output, the control unit 20 controls the switching unit 71 through the function of the position control unit 21b (step S230). That is, the output of the machine learning model 30b is a 6-bit value indicating the state of the switching unit 71 that should be set next. Therefore, the control unit 20 controls the signal conversion interface 13 so that the current is supplied to the solenoid valve 14 corresponding to the value 1, and the current supply to the solenoid valve 14 is stopped corresponding to the value 0. As a result, the six solenoid valves 14 adjust the air pressure, the diaphragm 71a of the switching unit 71 is actuated through the pressure supply pipe 15, and the state of the first flow path 40 changes. As a result, the liquid in the first flow path 40 moves, and the operation object in the operation area 60 moves through the movement of the liquid. As described above, in this embodiment, the air pressure adjustment unit 70, the imaging unit 21a, the position control unit 21b, the signal conversion interface 13, the solenoid valve 14, and the pressure supply pipe 15 constitute the liquid control unit.
[0082] If step S230 is executed, the control unit 20 repeats the processing after step S215. According to the above processing, the control unit 20 can move the operation object to the target position by repeatedly observing the state and inputting the state to the machine learning model 30b. The machine learning model 30b outputs the state of the next switching unit 71 required to move to the target position by inputting the current state. By repeating such position control of the operation object, the operation object reaches the target position. If the machine learning model 30b performs machine learning appropriately, the processing can be achieved through simple processing such as observation and input of the state, and the operation object can be correctly moved to the target position. Therefore, the usually very difficult task of moving the operation object caused by the state switching of the switching unit 71 can be achieved very simply and correctly.
[0083] A total of 6 switching units 71 are provided in the first flow path 40, so the state that can be realized by the switching unit 71 is any one of 64 states that can be expressed in 6 bits. In the machine learning model 30b, according to the current state, the next state is selected from the 64 states, and the switching process of the switching unit 71 is repeated to make it the next state, thereby the operation object moves to the target position in the operation area 60. In each of the 64 states, the liquid can move from the first flow path 40 to the operation area 60 side, the liquid moves in the opposite direction of the movement, the liquid moves from the first flow path 40 to the second flow path 50 side, and the liquid moves in the opposite direction of the movement. Immediately after the state change of the switching unit 71, the movement of the liquid occurs, and the pressure changes at this time. However, if the pressure change remains until the next state change, the movement of the liquid caused by the next state change is different from the originally assumed movement. Therefore, in the process of repeating the state change of the switching unit 71, the movement direction and movement amount of the liquid are different from the original assumption, and as a result, the position control of the operation object becomes difficult.
[0084] However, in this embodiment, if Figure 1BAs shown, a plurality of first flow paths 40 are connected to the operation area 60, and a second flow path 50 is connected to each of the first flow paths 40. Moreover, in the present embodiment, the second flow path 50 is a series of flow paths surrounding the first flow path 40. That is, the plurality of first flow paths 40 are connected via the second flow path 50. With such a structure, in the present embodiment, the pressure in the first flow path 40 is easily equalized via the second flow path 50. Therefore, it is difficult to generate a local pressure difference in the first flow path 40 or the operation area 60. In addition, even if the pressure changes by the movement of the liquid in the first flow path 40, the pressure change is uniform as soon as possible. Therefore, in the case of operating the position of the operation object by repeating the movement of the liquid, the movement of the liquid used to operate the operation object is easy to control. Furthermore, in the present embodiment, the switching unit 71 is provided in the first flow path 40, and is not provided in the second flow path 50. Therefore, the operation of the switching unit 71 necessary for operating the position of the operation object and the equalization of the pressure in the first flow path 40 can be performed in different places. Therefore, the operation of the switching unit 71 does not hinder the pressure uniformity.
[0085] Furthermore, in the present embodiment, the first flow path 40 is thinner than the second flow path 50. Specifically, the diameter of the circle formed by the reservoir 51 constituting the second flow path 50 is at least 10 times larger than the width of the first flow path 40 and at least several times larger than the width of the connecting portion 52. Furthermore, the width of the connecting portion 52 is several times larger than the width of the first flow path 40.
[0086] Therefore, the storage section 51 is much larger than the first flow path 40 and the connecting section 52. Therefore, the storage section 51 can effectively absorb the pressure changes generated in the first flow path 40 and make the pressure in the first flow path 40 uniform as soon as possible. In addition, the storage sections 51 are connected by the connecting section 52, so that even if the multiple first flow paths 40 respectively generate different pressure changes, they can be effectively absorbed as soon as possible to make the pressure in the first flow path 40 uniform. In this way, the second flow path 50 functions as a very effective buffer for the pressure changes or liquid movement generated in the first flow path 40. Therefore, the second flow path 50 eliminates the pressure changes in the first flow path 40 or the operating area 60 as quickly as possible and makes them uniform, and reaches a balanced state as soon as possible. Therefore, it is possible to enter the next state output by the machine learning model 30b as soon as possible. In addition, the movement of the liquid when entering the next state can be easily controlled.
[0087] Moreover, the second flow path 50 is composed of a connecting portion 52 and a storage portion 51 having a diameter larger than the width of the connecting portion 52. In addition, the first flow path 40 is connected to the storage portion 51. Therefore, the storage portion 51, which is relatively large compared to the first flow path 40 or the connecting portion 52, functions effectively as a buffer for making the pressure changes generated in the first flow path 40 or the connecting portion 52 uniform as soon as possible. Moreover, the storage portions 51 are connected through the connecting portion 52, thereby the pressure difference generated between the multiple first flow paths 40 is also uniform as soon as possible. In addition, the connecting portion 52 is a linear structure connecting the storage portions 51. If the connecting portion 52 is a straight line, the storage portions 51 can be connected with the shortest possible distance, so the pressure difference between the storage portions 51 can be efficiently made uniform.
[0088] Moreover, the operation area 60 has an internal space that allows the operation object to move in a plane along a direction parallel to the largest surface of the microfluidic chip 16. Therefore, in this embodiment, as the moving direction of the operation object, almost a two-dimensional direction can be assumed, and the position of the operation object can be easily determined and easily controlled. Moreover, the shape of the internal space of the operation area 60 observed from the direction perpendicular to the largest surface of the microfluidic chip 16 is a triangle. Therefore, the internal space is an extremely simple shape, and the operation area 60 can be easily formed.
[0089] Moreover, the first flow path 40 is connected to the vertex of the triangle formed by the operation area 60. Therefore, when the liquid flows from the first flow path 40 into the operation area 60, or when the liquid flows out from the operation area 60 to the first flow path 40, it is difficult for the liquid to stagnate or swirl in the operation area 60. On the other hand, if the first flow path 40 is connected to the side rather than the vertex of the triangle, it is easy for the liquid to stagnate or swirl in the operation area 60. Therefore, compared with the structure in which the first flow path 40 is connected to the side of the triangle, in this embodiment, the position control of the operation object is easy.
[0090] Furthermore, in the above-mentioned embodiment, two switching parts 71 are provided for one first flow path 40. Therefore, the liquid can flow in and out between the first flow path 40 and the operation area 60 independently in each of the three first flow paths 40 or by combining the three first flow paths 40. Therefore, the position of the operation object can be easily controlled in the operation area 60.
[0091] (4) Other implementation methods:
[0092] The above embodiment is an example for implementing the present invention, and various embodiments can be adopted in addition to this. For example, the shape and structure of the second flow path 50 are not limited to the above-mentioned embodiment. Therefore, it can be the following method: around the operating area 60 and the first flow path 40, a flow path of a certain width similar to the shape of the operating area 60 (for example, a polygon) and larger in size is formed, thereby forming the second flow path 50. In addition, a large storage portion can be formed around the operating area 60 and the first flow path 40 to form the second flow path, and various methods can also be adopted. Moreover, in the case where the second flow path 50 is composed of a storage portion 51 and a connecting portion 52, the storage portion 51 can also be a shape other than a circle.
[0093] Furthermore, the position simulation of the operation object according to the action in step S110 may be realized by a method other than using the simulation model 30c. For example, simulation based on other methods, finite element analysis or various fluid simulations, simulation using molecular dynamics, etc. may be used.
[0094] Moreover, the input and output methods of the machine learning model 30b are not limited to the above examples. For example, the state of the switching unit 71 is not limited to the 6-bit expression, but can also be expressed as a 1-digit decimal number (any value from 0 to 5, etc.). In addition, the position of the t+1th operation object output by the simulation model 30c can also be expressed not by coordinates, but by a displacement vector from the current position, etc. Moreover, the target position can also be defined by a wide area instead of a point.
[0095] Moreover, the state of a switching unit 71 may not be a dual state, but a multi-state, and may also change continuously. Moreover, the machine learning model 30b may also be updated during operation. For example, it may also be configured to update the simulation model 30c based on the data accumulated during operation, and to perform reinforcement learning using the updated simulation model 30c, and the more the operation is performed, the higher the accuracy becomes. At this time, the environment used to determine the state, etc. may be a real environment, in which case the coordinates of the operating object under the real environment are observed. Moreover, the moving direction of the operating object is not limited to the two-dimensional direction, and the position may also be controlled in the three-dimensional direction. In this case, the shooting of the microscope 12 is implemented from multiple different directions.
[0096] Moreover, the reinforcement learning method is not limited to the Q learning as described above, and methods such as SARSA can also be used. In addition, a method of modeling the policy model and the action value function model separately can also be used, such as the Actor-Critic algorithm. Moreover, the machine learning model of the next state of the output switching unit 71 is not limited to the structure generated by reinforcement learning, but can also be a structure generated by a neural network or the like. The machine learning method is not limited to the above-mentioned implementation mode, and the number of layers or nodes constituting the model, the type of activation function, the type of loss function, the type of gradient descent method, the type of optimization algorithm of the gradient descent method, the existence and batch number of small batch learning, the learning rate, the discount rate, the initial value, the type and existence of the overfitting suppression method, the existence of the convolution layer, the size of the filter in the convolution operation, the type of filter, the type of padding and stride, the type and existence of the pooling layer, the existence of the full combination layer, the existence of the recursive structure, and various elements can be appropriately selected to perform machine learning. Of course, learning can also be performed through other machine learning, such as support vector machines or clustering. Moreover, machine learning that automatically optimizes the structure of the model (for example, the number of layers or the number of nodes per layer, etc.) can also be performed.
[0097] In addition, the control system may be implemented by multiple devices, or a system in which machine learning is implemented on a server and position control is implemented on a client. Moreover, at least a part of the imaging unit 21a, the position control unit 21b, and the machine learning unit 22a may be divided into multiple devices. Moreover, part of the structure of the above-mentioned embodiment may be omitted, and the order of processing may be changed or omitted.
[0098] The operation area can be any area that can contain liquid and the operation object whose position is operated in the liquid. That is, the operation area is a storage area of liquid, and within the operation area, the position of the operation object can be variable. Since the operation object is operated by being transported in the liquid, at least the target position of the operation object or the part of the path that becomes close to the target position is set to a state where liquid exists. In the operation area, since there is a liquid that can move, there is a three-dimensional structure that can accumulate liquid, but the moving direction of the substantial operation object can be three-dimensional or not. That is, the operation object can move along a specified plane in a planar manner, or it can move one-dimensionally, as in the above-mentioned embodiment.
[0099] The liquid is not particularly limited, and may be, for example, water, physiological saline, blood, serum, phosphate buffer, HEPES and other good buffers, oil, magnetic fluid, polymer solution, photocurable resin, etc. In addition, the liquid may be a combination thereof. The water is not particularly limited and may be appropriately selected according to the purpose, for example, industrial water, tap water, ion exchange water, filtered water, reverse osmosis water, distilled water (pure water), ultrapure water, etc.
[0100] The object of operation can be selected appropriately as long as it is an object that can move in a liquid. For example, any particle such as latex particles, silica particles, magnetic particles, cells, liposomes, etc. can be the object of operation. In addition, it can be a receptor of a bead, an antibody, etc. or a nanocolloid-modified microparticle, etc. In addition, it can also be an active object such as a sperm, a microorganism, etc.
[0101] The operation object only needs to be operated so as to be arranged at a predetermined position in the liquid present in the operation area, and the predetermined position may be a predetermined target position, a predetermined track position, etc. In any case, the position where the operation object is to be arranged before movement may be determined as the predetermined position.
[0102] When the manipulation system performs a position manipulation of the manipulation object, the liquid and the manipulation object may be contained in the manipulation area, and the liquid or the manipulation object may be present outside the manipulation area before the manipulation begins. In addition, the state of containing the liquid or the manipulation object may be a state of being present therein. For example, if it is a liquid, the manipulation area or each flow path may be filled with the liquid, or may contain a portion of air, etc.
[0103] The first flow path is connected to the operation area and is a plurality of flow paths containing liquid. That is, the operation area is connected to at least two first flow paths containing liquid. As a result, the inflow path of the liquid to the operation area and the outflow path of the liquid from the operation area are ensured, and the operation of the operation object can be ensured to be performed with the movement of the liquid. The shape and cross-sectional shape of the first flow path are arbitrary as long as the liquid can move inside. In addition, in order to prevent the pressure distribution from becoming complicated and stagnation from occurring, a simple shape, such as a flow path extending in a straight line, is preferred. In addition, the cross-sectional shape is also preferably a simple shape, for example, a cross-section is circular or rectangular, and a straight-line flow path with a constant line can be used. The number of first flow paths is not limited to 3 as in the above-mentioned embodiment, and can also be various numbers.
[0104] The liquid control unit only needs to be able to move the liquid in the first flow path. Various structures can be used to move the liquid in the first flow path. The structure for moving the liquid by directly applying pressure to the liquid is as described above, and is not limited to a structure that switches between a closed state and an unclosed state of the first flow path. For example, various structures that move the liquid in the first flow path can be used. The structure for applying force to the liquid is also not limited to a structure that changes the diaphragm by air pressure, and can be configured so that force acts on the liquid through various actuators such as piezoelectric elements. In addition, when applying force to the liquid, a structure in which the force acts indirectly can be used, for example, a structure in which the force acts on the deformable first flow path to move the liquid in the first flow path, etc.
[0105] The second flow path is a series of flow paths containing liquid, and it only needs to be connected to multiple first flow paths respectively. That is, multiple first flow paths can be connected to each other through the second flow path. It can also be configured so that the second flow path is connected to multiple first flow paths respectively, and the liquid in the second flow path is connected to the liquid in the first flow path, as a result, the pressure difference between the first flow paths can be made uniform. Therefore, the shape of the second flow path is not limited. For example, in the above-mentioned embodiment, three connecting parts 52 are provided, and the second flow path 50 forms a triangle, but it is not limited to such a structure, and the connecting part 52 can also be a shape with only two sides of the triangle. In addition, it is also possible that the shape is not clearly distinguished between the storage part and the connecting part 52. For example, it is also possible to form a connecting part 52 with the same thickness as the diameter of the storage part 51 to connect the storage part 51.
[0106] The second flow path is a series of flow paths, and the second flow path itself can be a connected flow path. In order to make the pressure uniform, the second flow path preferably forms a closed loop as in the above embodiment, but the second flow path can also be a series of flow paths with ends. In any case, the plurality of first flow paths respectively connected to the operating areas are connected to the second flow path formed as a series of flow paths, thereby making the pressure of the first flow path uniform.
[0107] The shape of the internal space of the operating area when viewed from a direction perpendicular to the plane in which the operating object moves can be a polygon. Therefore, the shape of the operating area is not limited to a triangle as in the above-mentioned embodiment, and can be a quadrilateral or the like, or a polygon with more vertices. In addition, the shape of the polygon can be the internal space, that is, the space in which the operating object can move. Therefore, in the case where the outer shape of the operating area and the shape of the internal space are different, the shape of the internal space can be a polygon. Moreover, the internal space can be any space in which the operating object can exist within the operating area. Therefore, a fence or a lattice structure through which liquid passes but the operating object does not pass can also be used to separate the interior of the operating area.
[0108] Moreover, in the above-mentioned embodiment, two switching parts 71 are provided for one first flow path 40, but the number of the first flow paths 40 and the switching parts 71 is not limited to such an example. For example, one first flow path 40 may be provided with three or more switching parts 71, and the number of the switching parts 71 provided in the plurality of first flow paths 40 may be different for each first flow path 40. If one or more switching parts 71 are provided for each first flow path 40, and a total of (the number of first flow paths 40 + 1) or more switching parts 71 are provided, the inflow of liquid into the operation area 60 and the outflow of liquid from the operation area 60 can be realized by the combination of the states of the switching parts 71. If a total of (the number of first flow paths 40 + 1) switching parts 71 are formed, the position control of the operation object in the operation area 60 can be realized with a small number. In addition, if a total of (the number of first flow paths 40 + 1) or more switching parts 71 are formed, more precise position control can be realized as the number of switching parts 71 increases.
[0109] Furthermore, there may be multiple target positions and multiple operation objects. Fig. 8A The image captured by the microscope 12 in an example in which there are two operation objects and one target position is shown. Even in such a method, if two operation objects are introduced into the operation area 60, one target position is set, and the machine learning model 30b is generated by machine learning, control can be performed so that the positions of the operation objects are manipulated to reach the target position. Fig. 8A In the figure, the image on the left shows the initial state immediately after the operation object is introduced, and the image in the center shows the state halfway to the target position. In addition, the image on the right shows the state where the operation object has reached the target position. In this way, it was confirmed that two operation objects can be moved to one target position through machine learning.
[0110] Figure 8B , Figure 8C The image captured by the microscope 12 in an example in which there are two operation objects and two target positions is shown. Even in such a method, if two operation objects are introduced into the operation area 60, two target positions are set, and the machine learning model 30b is generated by machine learning, control can be performed so that the positions of the operation objects are manipulated to reach the target positions. Figure 8B , 8C In the figure, the left image also shows the initial state immediately after the operation object is introduced, and the central image shows the state halfway to the target position. In addition, the right image shows the state where the operation object has reached the target position. As shown in these, it has been confirmed that two operation objects can be moved to two target positions through machine learning. In addition, Figure 8A to Figure 8C As shown, various combinations can be adopted as the number of operation objects, the number of target positions, and the initial position.
[0111] Fig. 9 4 is a diagram showing an example of a structure for introducing an operation object into the first flow path 40 or the operation area 60. Fig. 9 In the example shown, the structure for introducing the operation object includes a third flow path 80, a second storage unit 81, a second switching unit 82, a second gas passage 83, and a second connection unit 84. Fig. 9 In the example shown, the third flow path 80, the second storage unit 81, the second switching unit 82, the second gas passage 83, and the second connection unit 84 form three groups, but for simplicity of the drawing, one group is assigned a reference numeral.
[0112] The third flow path 80 is connected to the side of the triangle formed by the operation area 60, and is a flow path extending in the extension direction of the first flow path 40. The third flow path 80 contains the same liquid as the liquid in the operation area 60, and the depth and width can be the same as the first flow path 40. The second reservoir 81 is connected to the end of the third flow path 80 that is not connected to the operation area 60. The diameter of the second reservoir 81 is several times larger than the width of the third flow path 80, and contains the same liquid as the liquid in the operation area 60, and accumulates the operation object in the liquid.
[0113] The second switching part 82, the second gas passage 83, and the second connecting part 84 have the same structure as the switching part 71, the gas passage 72, and the connecting part 73. However, the setting positions are different. That is, the second switching part 82 is formed directly below the third flow path 80, and the second gas passage 83 becomes a passage connecting the second connecting part 84 directly below the second switching part 82. The second connecting part 84 is connected to the solenoid valve 14. That is, in Fig. 9 In the example shown, there are a total of nine solenoid valves 14. By selecting the nine solenoid valves 14, the operation object is introduced.
[0114] That is, through the solenoid valve 14, the internal pressure of the second gas passage 83 or the gas passage 72 is switched to either a high state or a low state. As a result, the third flow path 80 or the first flow path 40 is switched to either a closed state or an unclosed state. If such a state switching is used, the operation object accumulated in one of the three second storage parts 81 can be moved to the operation area 60. For example, only one of the three third flow paths 80 is set to an unclosed state, and the other two third flow paths 80 are set to a closed state. In this state, any one of the switching parts 71 is operated to form a flow of liquid from the operation area 60 to the first flow path 40, thereby, the operation object accumulated in the second storage part 81 can be introduced into the operation area 60. According to the above structure, the operation object can be easily introduced into the operation area 60.
[0115] In addition, according to Fig. 9In the structure shown in FIG. 1 , the third flow path 80 and the second storage unit 81 are arranged inside the second flow path 50, so that the operation object accumulated near the operation area 60 can be introduced into the operation area 60. Therefore, the distance that the operation object should move for introduction is short, and the operation object can be easily introduced into the operation area 60. Fig. 9 In the structure shown, a plurality of second accumulation sections 81 are provided. Therefore, different operation objects can be accumulated in different second accumulation sections 81, and the operation object of the imported object can be switched. Specifically, the following structure can be adopted: an operation object for machine learning, such as beads, is accumulated in one second accumulation section 81, and an operation object of research object, such as cells, is accumulated in another second accumulation section 81, and a second accumulation section 81 is selected according to the respective scenes of machine learning and research, and the desired operation object is imported into the operation area 60.
[0116] Furthermore, the operation area is not limited to one location, and a configuration may be adopted in which a placement operation of the operation object can be performed using a plurality of operation areas. Fig.10 , Fig.11 It is a diagram for explaining a manipulation system having a plurality of operation areas. Fig.10 1 is a diagram showing a substrate 100 for implementing a manipulation system. Fig.11 This is an enlarged view of the periphery of the operation area provided on the substrate 100 .
[0117] The manipulation system is realized by a base plate 100 having a plurality of units Ut. Fig.10 In the figure, the boundary of the unit Ut is represented by a dotted line. A plurality of electromagnetic valves 140 are connected to the substrate 100. The electromagnetic valves 140 are connected to a signal conversion interface not shown in the figure. The signal conversion interface can be implemented in a circuit having the same function as the signal conversion interface shown in FIG1 . In addition, similarly to the embodiment shown in FIG1 , a computer not shown in the figure is connected to the signal conversion interface, and a microscope not shown in the figure is connected to the computer. These computers and microscopes may have the same functions as the computer 11 and the microscope 12.
[0118] That is, a machine learning model is recorded in a storage medium (not shown) of the computer, and the machine learning model is preliminarily machine-learned based on the operation area set on the substrate 100. In addition, the computer obtains an image of the operation area described later of each unit Ut through a microscope, outputs a signal to the signal conversion interface based on the image, and performs the configuration operation of the operation object in the operation area by controlling the solenoid valve 140.
[0119] The electromagnetic valve 140 is connected to a compressor (not shown) and a pressure supply pipe 150, and the connection between the compressor and the pressure supply pipe 150 is switched by opening and closing the electromagnetic valve 140. Therefore, the electromagnetic valve 140 can switch the internal pressure of the pressure supply pipe 150 between a high state and a low state.
[0120] The pressure supply pipe 150 is connected to a gas pressure regulating unit 700 (described later) formed on the substrate 100. Therefore, the solenoid valve 140 can switch the internal pressure of the gas pressure regulating unit 700 connected to the pressure supply pipe 150 between a high state and a low state.
[0121] The plurality of units Ut provided on the substrate 100 are provided with a liquid-enclosing structure containing a liquid and a structure containing a gas. Specifically, the units Ut include a first flow path 400, a second flow path 500, an operation area 600, and a gas pressure adjustment unit 700. In this embodiment, the liquid is water and the gas is air. In addition, the operation object is introduced into the first flow path 400 or the operation area 600, but the structure for introducing the operation object is Fig.10 , Fig.11 Omitted in .
[0122] The first flow path 400, the second flow path 500, and the operation area 600 are flow paths and areas containing liquid. The operation area 600 is an area containing liquid and performing configuration operations on the operation object inside, and has an internal space that allows the operation object to move in a plane. In the present embodiment, the operation area 600 and the like are formed on the largest surface of the substrate 100, and the shape of the internal space of the operation area 600 enables the operation object to move in a plane along a direction parallel to the largest surface of the substrate 100. That is, the operation area 600 has an internal space that expands in a direction parallel to the surface at a certain depth (for example, a depth of less than 30 μm) on the largest surface of the substrate 100. Moreover, the shape of the internal space of the operation area 600 when observed from a direction perpendicular to the largest surface of the substrate 100 is a quadrilateral.
[0123] The first flow path 400 is a linear hollow space, and one end thereof is connected to the vertex of the quadrilateral formed by the operation area 600. In the present embodiment, the first flow path 400 is a flow path extending in a direction parallel to the diagonal line of the quadrilateral formed by the operation area 600. The other end of the first flow path 400 is connected to a part of the second flow path 500, that is, the flow path 510. In the present embodiment, the depth of the first flow path 400 is the same as the depth of the operation area 600, and the width of the first flow path 400 is smaller than the width of the second flow path 500.
[0124] The second flow path 500 includes a flow path 510, a flow path 520, a flow path 530, and a storage unit 540. The flow path 510 is a square frame-shaped flow path formed around the operation area 600. In the flow path 510, the first flow path 400 is connected to the inner peripheral side of the vertex part of the square. The flow path 530 is a flow path extending in the direction in which the units Ut are arranged, that is, a flow path formed at two locations sandwiching the flow path 510.
[0125] Flow path 510 is a series of flow paths connecting four first flow paths 400, and flow path 520 is a flow path connecting flow path 510 and flow path 530. The depths of flow paths 510, 520, and 530 are the same as those of first flow path 400. The widths of flow paths 510, 520, and 530 are greater than those of first flow path 400.
[0126] The reservoir 540 is a circular portion connected to the end of the flow path 530. The reservoir 540 is deeper than the first flow path 400 (for example, 3 mm to 5 mm). The diameter of the circle formed by the reservoir 540 is larger than the width of the flow path 530.
[0127] The gas pressure regulating unit 700 is a structure including a passage for enclosing gas. In the present embodiment, the gas pressure regulating unit 700 includes switching units 710 and 711 (see Fig.11 ) and gas passages 720, 721. Of course, the above-mentioned Figure 1B A portion such as the connecting portion 73 shown.
[0128] The switch unit 710 includes a mechanism capable of switching the first flow path 400 between a closed state and an unclosed state. In this embodiment, each of the four first flow paths 400 is provided with one or more switch units 710. In addition, the total number of switch units 710 is preferably equal to the number of first flow paths 400 + 1. Fig.10 , Fig.11 In the example shown, the first flow path 400 located at the upper left of the operation area 600 is provided with two switching parts 710, and the other first flow paths 400 are provided with one switching part 710. Therefore, the total number of switching parts 710 is 5. The switching parts 710 and the first flow paths 400 also have the above-mentioned Figure 2C to Figure 2E With the structure shown, the internal pressure of the first flow path 400 can be adjusted by the switching unit 710 .
[0129] The gas passage 720 is a passage for gas extending from each switching section 710. In each unit Ut, the gas passage 720 extends from the switching section 710 side to the electromagnetic valve 140 side, and further extends along the direction in which each unit Ut is arranged. The gas passages 720 of each unit Ut are connected. In the present embodiment, in the switching section 710 existing in each unit Ut, the gas passages 720 extending from the switching section 710 having the same relative positional relationship with respect to the operation area 600 are connected. For example, Fig.10 , Fig.11 In the embodiment, gas passages 720 extending from a switching portion 710 located at the upper right of the operation area 600 are connected to each other.
[0130] In the substrate 100, a pressure supply pipe 150 extending from one electromagnetic valve 140 is connected to one gas passage 720. Therefore, when the internal pressure of the pressure supply pipe 150 becomes high or low by the control of the electromagnetic valve 140, the internal pressure of the gas passage 720 also becomes high or low. As a result, it is possible to switch between a state in which the liquid cannot flow in the first flow path 400 and a state in which the liquid can flow.
[0131] As described above, in this embodiment, the gas pressure regulator 700 constitutes a structure that applies pressure for moving the liquid in the first flow path to the liquid. In addition, the electromagnetic valve 140 and the pressure supply pipe 150 constitute a pressure supply unit that supplies pressure to the structure provided in the plurality of units Ut.
[0132] In the present embodiment, a switching portion 711 is further formed in the flow path 520, and a gas passage 721 is connected to the switching portion 711. The structures of the switching portion 711 and the gas passage 721 are the same as the structures of the switching portion 710 and the gas passage 720. Therefore, by the control of the solenoid valve 140, if the internal pressure of the pressure supply pipe 150 becomes a high state or a low state, the internal pressure of the gas passage 721 also becomes a high state or a low state. As a result, it is possible to switch between a state in which the liquid cannot flow in the flow path 520 and a state in which the liquid can flow.
[0133] If the liquid cannot flow through the flow path 520, the flow of the liquid between the flow path 510 and the flow path 530 is cut off, so the pressure of the flow path 510 of each unit Ut is not affected by the pressure fluctuation caused by the operation in other units Ut. If the liquid can flow through the flow path 520, the liquid can flow between the flow path 510 and the flow path 530, so the pressure in each flow path 510 of the plurality of units Ut can be made uniform.
[0134] In this embodiment, the substrate 100 further includes supply control units 910a to 910d (see FIG. 1 ) for switching between a state in which pressure can be supplied to the gas pressure regulating unit 700 and a state in which pressure cannot be supplied, for each of the plurality of units Ut. Fig.10 ) That is, the supply control units 910a to 910d can switch between a state in which the gas in the intersecting gas passages 720 and 721 can flow and a state in which the gas cannot flow.
[0135] The supply control units 910a to 910d can be implemented by various structures. For example, the pressure can be adjusted by using a solenoid valve and a pressure supply pipe. Figure 2C to Figure 2E The diaphragm or the like acts to cut off or open the gas passages 720, 721. In addition, the structure of directly switching or opening the gas passages 720, 721 by the electromagnetic valve may also be adopted.
[0136] Supply control units 910a to 910d simultaneously switch whether or not the pressure in the intersecting gas passages 720 and 721 can be supplied. Fig.10 In the example shown, the supply control units 910a to 910d can simultaneously switch whether the pressure in the 14 intersecting gas passages can be supplied. Fig.10 The pressure in the five gas passages 720 and two gas passages 721 in the unit Ut shown in the uppermost section and the five gas passages 720 and two gas passages 721 in the unit Ut shown in the third section from the top can be supplied. On the other hand, the supply control unit 910c can simultaneously switch Fig.10 Whether the pressure in the five gas passages 720 and two gas passages 721 in the unit Ut shown in the uppermost row and the five gas passages 720 and two gas passages 721 in the unit Ut shown in the second row from the top can be supplied.
[0137] The supply control unit 910 is connected to a signal conversion interface (not shown), and the computer controls the operation of the supply control units 910a to 910d through the signal conversion interface. That is, the computer can select an arbitrary unit Ut by switching the combination of cutting off or opening the gas passages of the supply control units 910a to 910d, thereby enabling or disabling the gas pressure adjustment by the electromagnetic valve 140.
[0138] For example, it is assumed that the supply control units 910a and 910c are set to a state in which pressure cannot be supplied, and the supply control units 910b and 910d are set to a state in which pressure can be supplied. In this case, Fig.10 The uppermost and third units Ut shown are set to be unable to supply pressure by the supply control unit 910a, and thus are invalid. In addition, the second unit Ut from the upper is set to be unable to supply pressure by the supply control unit 910c, and thus is invalid. Fig.10The unit Ut at the bottom shown is enabled to supply pressure by the supply control units 910b and 910d, and is thus effective.
[0139] For the effective unit Ut, pressure can be supplied by the electromagnetic valve 140, and the switching can be performed through the gas passages 720, 721 and the switching parts 710, 711. That is, the pressure of the liquid in the first flow path 400 is adjusted or the flow path 520 is opened and closed. According to the above structure, in this embodiment, an arbitrary unit Ut can be selected by the control of the computer, and the placement operation of the operation object can be performed in the operation area 600 of the selected unit Ut.
[0140] Moreover, in a structure in which the position of an operation object in an operation area is controlled by the movement of liquid from a plurality of first flow paths, a method in which each first flow path is connected to a second flow path can also be implemented as a method invention. In addition, the above-mentioned system and method can be assumed to be implemented as a single device or implemented by a plurality of devices, including various methods. For example, a microscope having the above-mentioned components or an accessory for being mounted on an existing microscope can be provided. Moreover, the invention is also established as a recording medium for a program executed in a control unit. Of course, the recording medium of the software can be a magnetic recording medium or a semiconductor memory, and the same consideration can be fully realized in any recording medium developed in the future.
[0141] Description of symbols
[0142] 10: operating system; 11: computer; 12: microscope; 13: signal conversion interface; 14: solenoid valve; 15: pressure supply pipe; 16: microfluidic chip; 20: control unit; 21: operating program; 21a: shooting unit; 21b: position control unit; 22: machine learning program; 22a: machine learning unit; 30: storage medium; 30b: machine learning model; 30c: simulation model; 40: first flow path; 50: second flow path; 51: storage unit; 52: connecting unit; 60: operating area; 70: air pressure adjustment unit; 71: switching unit; 71a: diaphragm; 71b: space; 72: gas passage; 73: connecting unit; 80: third flow path; 81: second storage unit; 82: second switching unit; 83: second gas passage; 84: second connecting unit.
Claims
1. A manipulation system for placing an object to be manipulated at a predetermined position in a manipulation area containing a liquid, comprising: the operating area; A first flow path, which is a plurality of flow paths connected to the operation area and containing the liquid; a liquid control unit that moves the liquid in the first flow path; as well as The second flow path is a series of flow paths containing the liquid, and is connected to the plurality of first flow paths respectively.
2. The operating system according to claim 1, wherein: The operation area has an internal space for moving the operation object in a two-dimensional manner, and a shape of the internal space viewed from a direction perpendicular to the plane is a polygon.
3. The operating system according to claim 2, wherein: The first flow path is connected to the vertices of the polygon.
4. The manipulation system according to claim 1 or 2, wherein: The first flow path is thinner than the second flow path.
5. The operating system according to claim 1, wherein: The liquid control unit includes a switching unit configured to move the liquid by switching between a state in which the first flow path is closed and a state in which the first flow path is not closed. One or more of the switching units are provided for each of the first flow paths, and a total of (the number of the first flow paths + 1) or more switching units are provided.
6. The operating system according to claim 1, wherein: The liquid control unit includes a switching unit configured to move the liquid by switching between a state in which the first flow path is closed and a state in which the first flow path is not closed. Two of the switching units are provided for one of the first flow channels.
7. The manipulation system according to claim 5 or 6, wherein: The liquid control unit performs machine learning on a machine learning model in a manner that takes as input the position of the operation object, the state of the switching unit, and the target position of the operation object, and outputs an action in the switching unit that is necessary to bring the operation object close to the target position, The position of the operation object, the state of the switching unit, and the target position of the operation object are input, and the switching unit is operated based on the obtained output.
8. The manipulation system according to claim 1 or 2, wherein: The second flow path includes a connecting portion and a storage portion of the liquid that is larger than the connecting portion. The storage unit is connected to the first flow path, The connecting portion is formed to connect the plurality of storage portions.
9. A fluid chip comprising: An operation area, which contains a liquid and performs the configuration operation of the operation object inside; A first flow path, which is a plurality of flow paths connected to the operation area and containing the liquid; a structure that applies pressure for moving the liquid in the first flow path to the liquid; as well as The second flow path is a series of flow paths containing the liquid, and is connected to the plurality of first flow paths respectively.
10. A control system comprising a plurality of units, The unit has: An operation area, which contains a liquid and performs the configuration operation of the operation object inside; A first flow path, which is a plurality of flow paths connected to the operation area and containing the liquid; a structure that applies pressure for moving the liquid in the first flow path to the liquid; as well as The second flow path is a series of flow paths containing the liquid, and is connected to the plurality of the first flow paths respectively. The second flow paths included in the plurality of the units are connected to each other. The control system comprises: a pressure supply unit that supplies the pressure to the structure included in the plurality of units; and A supply control unit switches, for each of the plurality of units, a state in which the pressure supply unit can supply the pressure to the structure and a state in which the pressure cannot be supplied.
