Path planning method for particles in photoelectric tweezers micro-fluidic chip

Abstract

The invention discloses a path planning method for particles in a micro-fluidic chip of photoelectric tweezers, which comprises the following steps: acquiring image data in the micro-fluidic chip, identifying and classifying particles in an image in real time, and determining spatial position coordinates of each particle and position coordinates of a target area in real time; generating a corresponding virtual electrode based on the spatial position coordinate of each particle; setting a gravitational force source and a repulsive force source in the micro-fluidic chip, calculating a gravitational force vector of the virtual electrode, and calculating a total repulsive force vector of each repulsive force source borne by the spatial position of the current virtual electrode; synthesizing the gravitation vector and the total repulsive force vector into a total force vector, taking the direction of the total force vector as the movement direction of the target particle, and determining the movement rate of the target particle; and the next position of the target particle is predicted based on the motion direction and the movement speed of the target particle, the position of the target particle is updated through multiple iterations, and the virtual electrode drives the target particle to advance to the target area along the path with the minimum potential energy.

Description

Technical Field

[0001] This invention relates to the field of microfluidic chip technology, and in particular to a method for planning particle paths within a microfluidic chip using optoelectronic tweezers. Background Technology

[0002] Many breakthroughs in biomedical research have benefited from the ability to precisely manipulate tiny targets at the micro- and nanoscale. For example, related technologies have greatly advanced research progress in areas such as cell sorting, cell patterning, cell separation, and cell counting. To meet the clinical demand for antibody-free and label-free sorting technologies, label-free microfluidic sorting technologies have developed rapidly in recent years. In such systems, optoelectronic tweezers (OETs), which combine optical tweezers with dielectrophoresis, have attracted considerable attention. OETs utilize photoconductive materials and programmable optical patterns to generate tens of thousands of virtual electrodes, enabling high-throughput parallel manipulation of nanowires, microparticles, cells, and microstructures. The negative dielectrophoretic force generated by the ring-shaped optical electrodes can fix individual microscale targets at the center of the ring, forming a "trap" similar to traditional optical tweezers, with a minimum required light intensity of approximately 1 W / cm². -2 This significantly reduces the photothermal effect and compresses the cost of laser equipment. Leveraging this advantage, OET has been widely applied in areas such as cell viability assessment, cell lysis, cell fusion, embryo screening, and tumor cell isolation, demonstrating particularly outstanding practical value in cell sorting.

[0003] In recent years, to meet the need for precise manipulation of multiple particles in complex particle swarms using negative permittivity, OET systems have been gradually developing towards automation and parallelization. However, existing OET technologies still face several key challenges that need to be overcome. Currently, the virtual electrodes in this technology are usually preset offline, lacking online real-time updates and dynamic path planning capabilities, making it difficult to flexibly adjust cell movement trajectories and increasing the risk of particle collisions. Moreover, due to the random distribution of cell particles in the continuous phase, optical electrodes often overlap and collide, causing the target to escape the trap. To avoid cell runaway, real-time path planning is necessary for each optical electrode. Inspired by UAV navigation technology, the Artificial Potential Field (APF) method provides a new approach to solving these problems. Compared with traditional path planning algorithms, APF only requires defining the target point and obstacle model for rapid deployment, without the need to construct a topology graph, time safety interval, or heuristic function. It directly generates displacement-velocity vectors in real time with a small amount of simple calculations, significantly reducing computational complexity. Therefore, the flexibility of this method is highly compatible with the programmable characteristics of OET optical virtual electrodes, making it particularly suitable for optoelectronic tweezers microfluidic systems with high requirements for dynamics and real-time performance. However, to date, no research has applied APF to OET technology. Summary of the Invention

[0004] In view of the above-mentioned prior art, the present invention provides a microparticle path planning method in a microfluidic chip for optoelectronic tweezers, which mainly solves the technical problems existing in the background art.

[0005] To achieve the above objectives, the technical solution of this invention is implemented as follows: a microparticle path planning method within a microfluidic chip for optoelectronic tweezers, the method comprising the following steps:

[0006] Acquire image data from microfluidic chips and perform real-time identification and classification of particles in the images;

[0007] Each particle is assigned a unique identifier, and the spatial coordinates of each particle and the coordinates of the target area are determined in real time.

[0008] A virtual electrode is generated based on the spatial coordinates of each particle, and the spatial position of the virtual electrode is the current spatial position of the target particle.

[0009] Define the attraction and repulsion sources in the microfluidic chip, calculate the attraction vector from the spatial position of the virtual electrode to the attraction source, and calculate the total repulsion vector of each repulsion source at the current spatial position of the virtual electrode.

[0010] The gravitational vector and the total repulsive vector are combined to form a total force vector, and the direction of the total force vector is taken as the direction of motion of the target particle. At the same time, the moving speed of the target particle is determined.

[0011] Based on the target particle's motion direction and speed, the next position of the target particle is predicted. The position of the target particle is updated through multiple iterations. The virtual electrode drives the target particle to move along the path of minimum potential energy to reach the target area.

[0012] Optionally, the particles in the image can be identified and classified in real time using the YOLOv5 neural network.

[0013] Optionally, a unique identifier can be assigned to each particle using the ByteTrack multi-target tracking algorithm, and the spatial coordinates of each particle and the position coordinates of the target area can be determined in real time.

[0014] Optionally, the gravitational source in the microfluidic chip is the target region, and the repulsive source is other particles besides the target particles, as well as preset obstacles and channel boundaries.

[0015] Optionally, when calculating the repulsion vector, the magnitude of the repulsion force of each repulsion source is adjusted according to the distance between it and the virtual electrode. The smaller the distance, the greater the repulsion force. When the distance exceeds a preset threshold, the repulsion force is zero.

[0016] Optionally, when calculating the gravitational vector, the magnitude of gravity is adjusted according to the distance from the virtual electrode to the gravitational source; the farther the distance, the greater the gravity, and when the distance is zero, the gravity is zero.

[0017] Optionally, during multiple iterations, the gravitational vector, repulsive vector, and total force vector are recalculated based on the latest spatial position coordinates of the target particle in each iteration, and the next position is updated accordingly.

[0018] Optionally, the total force vector always guides the target particle to move in the direction of decreasing potential energy, and each position update makes the target particle closer to the target area and farther away from the repulsive force source.

[0019] Optionally, the movement speed of the target particle is determined by combining the magnitude of the total force vector with the step size coefficient.

[0020] The beneficial effects of this invention are as follows: Based on visual feedback and artificial potential field planning, it stably achieves high-purity, collision-free, real-time sorting of multiple types of particles, breaking through the limitations of traditional photoelectric tweezers that rely on preset paths and offline control. It significantly improves the response speed, control accuracy, and versatility of the sorting system. Furthermore, this method can simultaneously track and manipulate multiple particles, achieving parallel processing and greatly improving the system's processing speed and experimental throughput. It is suitable for large-scale screening and high-throughput analysis, possessing good practical application value. It can also dynamically adjust the particle's trajectory and operating parameters according to experimental needs, ensuring high flexibility and adaptability in the experimental process. It provides an efficient solution for micro- and nano-scale particle manipulation in fields such as biomedicine, and is especially suitable for life science research scenarios such as cell sorting where high response speed and spatial accuracy are required. Attached Figure Description

[0021] Figure 1 This is a flowchart illustrating the microparticle path planning method within a microfluidic chip using optoelectronic tweezers in an embodiment of this application.

[0022] Figure 2 This is a schematic diagram illustrating the control principle of the photoelectric tweezers in the embodiments of this application;

[0023] Figure 3 This is a schematic diagram of the artificial potential field algorithm in the embodiments of this application;

[0024] Figure 4 This is a schematic diagram of the photoelectric tweezers intelligent control system in the embodiments of this application;

[0025] Figure 5 This is a schematic diagram showing the results of the particle collision-free path planning experiment in the embodiments of this application;

[0026] Figure 6 This is a schematic diagram of the results of the particle obstacle avoidance path planning experiment in the embodiments of this application;

[0027] Figure 7 This is a schematic diagram showing the results of the automatic particle sorting experiment in the embodiments of this application. Detailed Implementation

[0028] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the invention. In the following description, the expression "some embodiments" refers to a subset of all possible embodiments; however, it should be understood that "some embodiments" can be the same subset or different subsets of all possible embodiments and can be combined with each other without conflict.

[0029] In the following description, numerous specific details are set forth in order to provide a more thorough understanding of the invention. However, it will be apparent to those skilled in the art that the invention can be practiced without one or more of these details. In other instances, certain technical features well-known in the art have not been described in order to avoid obscuring the invention.

[0030] It should be understood that the present invention can be embodied in various forms and should not be construed as being limited to the embodiments set forth herein. Rather, providing these embodiments will make the disclosure thorough and complete, and will fully convey the scope of the invention to those skilled in the art. Furthermore, the terminology used herein is intended only to describe particular embodiments and is not intended to limit the invention. When used herein, the singular forms “a,” “an,” and “the” are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the terms “compose” and / or “comprising,” when used in this specification, identify the presence of the stated features, integers, steps, operations, elements, and / or components, but do not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups. When used herein, the term “and / or” includes any and all combinations of the associated listed items.

[0031] It should also be noted that when an element is referred to as being "fixed to" another element, it can be directly attached to the other element or there may be an intervening element. When an element is referred to as being "connected to" another element, it can be directly connected to the other element or there may be an intervening element. The terms "vertical," "horizontal," "inner," "outer," "left," "right," and similar expressions used herein are for illustrative purposes only and do not represent the only possible implementation.

[0032] To fully understand this invention, a detailed structure will be presented in the following description to illustrate the technical solution proposed by this invention. Optional embodiments of the invention are described in detail below; however, in addition to these detailed descriptions, the invention may have other embodiments.

[0033] Please refer to the attached document. Figure 1 This application provides a method for microparticle path planning within a microfluidic chip using optoelectronic tweezers, the method comprising the following steps:

[0034] S1. Acquire image data from the microfluidic chip and perform real-time identification and classification of particles in the image;

[0035] S2. Assign a unique identifier to each particle and determine the spatial coordinates of each particle and the position coordinates of the target area in real time;

[0036] S3. Generate corresponding virtual electrodes based on the spatial coordinates of each particle. The spatial position of the virtual electrodes is the current spatial position of the target particle.

[0037] S4. Set the attraction and repulsion sources in the microfluidic chip, calculate the attraction vector from the spatial position of the virtual electrode to the attraction source, and calculate the total repulsion vector of each repulsion source at the current spatial position of the virtual electrode.

[0038] S5. Combine the gravitational vector and the total repulsive vector to form a total force vector, and use the direction of the total force vector as the direction of motion of the target particle, while determining the moving speed of the target particle.

[0039] S6. Based on the target particle's motion direction and speed, predict the target particle's next position. Update the target particle's position through multiple iterations. The virtual electrode drives the target particle to move along the path of minimum potential energy to reach the target area.

[0040] Specifically, Figure 2 This study demonstrates the basic structure and manipulation principle of an optoelectronic tweezers (OET) microfluidic chip. The chip consists of two transparent substrates (upper and lower plates), each with a conductive thin film covering its inner surface, and a liquid chamber containing suspended microparticles in between. A photoconductive layer (such as hydrogenated amorphous silicon a-Si:H) is also placed above the conductive film on the lower plate. When a digital micromirror device (DMD) projects a specific spot of light onto the photoconductive layer, a photoconductive effect occurs in a localized area, creating a non-uniform electric field distribution. Driven by an alternating electrical signal (provided by a signal generator), this non-uniform electric field induces negative permittivity in the microparticles within the liquid, causing them to move towards the region of lowest electric field strength—essentially becoming "trapped" at the center of the light ring—forming an optical trap similar to that of traditional optical tweezers. This process offers advantages such as non-contact operation, low power consumption, and programmability, representing a key physical basis for precise microparticle manipulation.

[0041] Figure 3 This diagram illustrates the path planning principle of the Artificial Potential Field (APF) method, used for obstacle avoidance and navigation control of particles in an optoelectronic tweezers system. In the diagram, the red particle represents the currently manipulated target, with the target endpoint denoted by "Goal." The straight line between the particle and the target generates an attractive force (F_att), driving the particle towards the target point. Simultaneously, two obstacles or other particles exist nearby, exerting repulsive forces (F_rep1 and F_rep2) on the target to prevent collisions. These two repulsive forces combine to form the total repulsive force F_rep. The system calculates the resultant force F (red arrow) of the attractive and repulsive forces in real time, which serves as the particle's direction of motion in the current frame. By continuously updating the current position and the direction of the resultant force, the particle will move along the path of minimum potential energy, ultimately avoiding obstacles and reaching the target point. This method does not require global path planning, relying only on local information about the current position and the surrounding environment. It boasts high computational efficiency and fast response speed, making it ideal for multi-target automatic manipulation tasks in microfluidic chips that demand high real-time performance and accuracy.

[0042] In one or more embodiments of this application, the YOLOv5 neural network is used to identify and classify particles in the image in real time, the ByteTrack multi-target tracking algorithm is used to assign a unique identifier to each particle, and the spatial coordinates of each particle and the position coordinates of the target area are determined in real time.

[0043] Real-time image data from within the microfluidic chip is acquired using a high-resolution microscopic imaging module and a color camera. The YOLOv5 neural network then rapidly identifies and accurately classifies particles within the images, effectively distinguishing between different types of particles (such as polystyrene microspheres with different color markings). Even in scenarios with dense particle density or partial overlap, high recognition accuracy is maintained. The ByteTrack multi-target tracking algorithm, based on the YOLOv5 recognition results, assigns a unique identifier (ID) to each detected particle, ensuring continuous tracking during particle movement and avoiding confusion due to positional changes or temporary occlusion. Simultaneously, the algorithm calculates and updates the spatial coordinates of each particle in real time, accurately capturing its dynamic motion. Furthermore, based on the particle's category information, it determines the target area coordinates for each particle (such as a specific exit channel in a sorting experiment), providing clear target guidance for subsequent path planning.

[0044] It should be noted that the specific process of using the YOLOv5 neural network to quickly identify and accurately classify particles in the image, and using the ByteTrack multi-target tracking algorithm to assign a unique identifier to each particle and determine the spatial coordinates of each particle and the position coordinates of the target area in real time, has been disclosed in the applicant's prior invention patents "CN119259133B, A method for intelligent tracking and parallel control of droplet array in a photoelectric wetting microfluidic chip", "CN119259132A, A photoelectric microfluidic method for automatically generating droplets on demand", and "CN119619027A, A photoelectric microfluidic automatic control method for screening microalgae growth conditions". It will not be described in detail here, but relevant details can be referred to the above-mentioned prior patents.

[0045] In one or more embodiments of this application, a ring-shaped virtual electrode matching the particle size is constructed in real time.

[0046] Specifically, once a particle enters the detection area of ​​the microfluidic chip, the ByteTrack multi-target tracking algorithm locks its spatial position in real time. It automatically adjusts the inner and outer diameters of the annular electrodes based on the particle's diameter, ensuring that the optical trap formed by the electrodes stably encapsulates the particle while preventing particle escape or decreased manipulation precision due to size mismatch. These annular virtual electrodes are formed by projecting optical patterns onto the photoconductor layer using a digital micromirror device (DMD). The photoconductivity of the photoconductor layer generates a non-uniform electric field, which then confines the particle to the center of the annular ring through negative permittivity, achieving non-contact capture and manipulation of the particle.

[0047] It should be noted that the specific process of generating the corresponding virtual electrode based on the spatial coordinates of each particle has been disclosed in the applicant's prior invention patents "CN119259133B, a method for intelligent tracking and parallel control of droplet array in a photoelectric wetting microfluidic chip", "CN119259132A, a photoelectric microfluidic method for automatically generating droplets on demand", and "CN119619027A, a photoelectric microfluidic automatic control method for screening microalgae growth conditions". It will not be described in detail here, but relevant details can be referred to the above-mentioned prior patents.

[0048] In one or more embodiments of this application, the attraction source in the microfluidic chip is the target region, and the repulsion source is other particles besides the target particles, as well as preset obstacles and channel boundaries.

[0049] Specifically, based on particle category information or experimental pre-sets, a specific region within the microfluidic chip is designated as an attractive force source, i.e., the target region that the target particle must reach. For example, in a sorting experiment, corresponding to the exit channels of different particle categories, the repulsive force source is configured as follows: in addition to the currently manipulated target particle, all other particles act as independent repulsive force sources. Furthermore, it includes pre-set obstacles in the system (such as fixed obstacles simulating a confined channel environment) and the chip's channel boundaries. The attractive force sources consistently exert an attractive force on the target particle, driving it towards the target region, while each repulsive force source exerts a repulsive force on the target particle. The system can dynamically generate a safe and efficient motion path for each target particle. The directional effect of the attractive force sources ensures that the particle always moves towards the target region, while the real-time feedback from the repulsive force sources effectively avoids various interferences and obstacles during the motion process. Especially in complex scenarios where multiple particles move simultaneously, the mutual repulsion between particles naturally resolves path intersection conflicts without the need for additional scheduling mechanisms.

[0050] In one or more embodiments of this application, when calculating the gravitational vector, the magnitude of the gravitational force is adjusted according to the distance from the virtual electrode to the gravitational source. The farther the distance, the greater the gravitational force. When the distance is zero, the gravitational force is zero. When calculating the repulsive vector, the magnitude of the repulsive force of each repulsive source is adjusted according to the distance between it and the virtual electrode. The smaller the distance, the greater the repulsive force. When the distance exceeds a preset threshold, the repulsive force is zero.

[0051] In the specific calculation, the gravitational vector between the virtual electrode corresponding to the target particle and the target region is first calculated. Since the gravitational force of any particle is equal to the negative of the gravitational potential energy gradient, the specific calculation formula is as follows:

[0052]

[0053] Where, x∈R 2 k represents the current two-dimensional position. att x is the attraction coefficient. goal For the target area location, U att (x) represents the attractive potential energy.

[0054] The total repulsive force vector is obtained through vector synthesis. This involves iterating through all repulsive sources near the current virtual electrode that meet the interaction conditions, calculating the repulsive force vector generated by each source, and then superimposing these vectors to obtain the total repulsive force vector. The formula for calculating the total repulsive force vector is as follows:

[0055]

[0056] Similarly, the repulsive force generated by each repulsive source is equal to the negative of the repulsive potential energy gradient. Based on the repulsive potential energy, the repulsive force generated by each repulsive source can be obtained. The formula for calculating the repulsive potential energy is as follows:

[0057]

[0058] Where, k rep Here, ρ(x) is the repulsion coefficient, ρ0 is the distance from the current position to the nearest obstacle boundary, and U is the threshold value. rep (x) represents the repulsive potential energy.

[0059] The gravitational vector and the total repulsive vector are combined into a total force vector using the following formula:

[0060]

[0061] The total force vector always guides the target particle to move in the direction of decreasing potential energy, and each position update makes the target particle closer to the target area and further away from the repulsive force source.

[0062] Furthermore, during the multiple iterations, each iteration recalculates the gravitational vector, repulsive vector, and total force vector based on the latest spatial position coordinates of the target particle, and updates the next position accordingly.

[0063] At the start of each iteration, the high-frame-rate microscopic imaging module captures the latest spatial coordinates of the target particle in real time. Based on this latest position, the system recalculates the gravitational vector, using the target region as the gravitational source, and dynamically adjusts the magnitude of the gravitational force according to the distance between the current virtual electrode and the target region; the farther the distance, the stronger the gravitational force. Simultaneously, it iterates through all repulsive sources near the target particle, recalculating the repulsive vector of each source based on its distance from the current position; the smaller the distance, the stronger the repulsive force, and the repulsive force becomes zero when the distance exceeds a preset threshold. These repulsive vectors are then combined to form a total repulsive vector. The recalculated gravitational vector and the total repulsive vector are combined to form a total force vector. The direction of this total force vector is used as the direction of movement of the target particle, and the movement speed is determined by combining it with a step size coefficient. This allows the system to predict and update the next position of the target particle, and the virtual electrode is also updated to this position synchronously, completing one iteration.

[0064] In one or more embodiments of this application, the moving speed of the target particle is determined by calculating the magnitude of the total force vector combined with the step size coefficient, and the specific calculation formula is as follows:

[0065]

[0066] Among them, v t Where v is the actual velocity at the previous time step. t * The candidate instantaneous velocity is calculated from the resultant force at the current position. The updated velocity is controlled by the damping factor η and obtained through exponential smoothing between the two.

[0067] See Figure 4The second aspect of this invention discloses an intelligent tracking and control system for droplet arrays within a photoelectric wetting microfluidic chip. The system includes an image acquisition and target recognition module, an artificial potential field path planning module, and an intelligent control and drive execution module.

[0068] The image acquisition and target recognition module includes a microscopic optical system consisting of a light source, objective lens, and CCD camera. The CCD camera is connected to a target detection and tracking module composed of a YOLOv5 neural network and the ByteTrack multi-target tracking algorithm to transmit the acquired microfluidic chip images to this module in real time, enabling particle identification, classification, location, and ID assignment. The artificial potential field (APF) path planning module is connected to the image acquisition and target recognition module, receiving particle identification, classification, and location information output by the target detection and tracking module. It then performs Agent initialization, target assignment, APF path planning, and state updates. The path planning results are synchronously transmitted to the intelligent control and drive execution module. The OET intelligent control system in the intelligent control and drive execution module is connected to the artificial potential field path planning module, receives the path planning results and generates virtual electrode projection instructions. The OET intelligent control system is connected to the DMD projection device and the signal generator respectively to send the virtual electrode projection instructions to the DMD projection device. At the same time, the control signal generator provides an electric field drive signal. The DMD projection device projects the optical pattern onto the surface of the OET chip to form a virtual electrode. The electric field drive signal generated by the signal generator forms a negative permittivity, realizing non-contact manipulation of the target particles.

[0069] In use, the photoelectric tweezers chip is placed under an upright microscope, and a color CCD camera acquires images of particles within the chip in real time, transmitting the data to the artificial intelligence module. The YOLOv5-ByteTrack algorithm sequentially detects and tracks particles in the field of view, assigning each particle a unique ID and determining the target endpoint based on its category. Subsequently, the APF path planner generates the next motion strategy. The system sends instructions to the optical projection module via control software, and the DMD projects an optical pattern onto the OET chip to drive particle movement, while a signal generator simultaneously provides an electrical drive signal. The coordinated operation of these modules enables high-speed detection, accurate classification, and robust manipulation of particles, providing strong technical support for label-free cell sorting.

[0070] To verify the robustness and collision avoidance capability of the multi-objective cooperative manipulation method based on the Artificial Potential Field (APF) method described in this invention in complex environments, multiple sets of experiments were designed to verify its effectiveness.

[0071] Experiment 1

[0072] See Figure 5To verify the robustness and collision avoidance capability of the multi-objective cooperative control method based on artificial potential field method described in this invention under complex environments, a set of extreme working condition experiments were designed to test whether individual intelligent agents get stuck in local minima or "get stuck" during path planning.

[0073] In this embodiment, the system simultaneously manipulates three different types of particles (labeled green_1, blue_2, and red_3, respectively) and constructs a cross-channel environment with an "equilateral triangle-inverted triangle" structure, so that each particle is geometrically constrained in two orthogonal directions (e.g., Figure 5 (As shown in A). The system detected the three types of particles at coordinates (667, 182), (1035, 976), and (329, 964) respectively, and assigned them target endpoints: Goal_1 = (650, 1000), Goal_2 = (300, 200) and Goal_3 = (1000, 200).

[0074] Each endpoint generates an attractive potential field for particles of the same type, while repulsive potential fields exist between particles of different types. After the system activates the visual detection and path planning module, the three particles converge at the path intersection point (650, 466) (e.g., Figure 5 As shown in A.ii), the deviation between the particles and the target halo is significantly reduced (see...). Figure 5 B). At this point, the residual path lengths of green_1, blue_2, and red_3 from their respective target points are 459.3 μm, 482.7 μm, and 369.7 μm, respectively.

[0075] Because the target position of the halo is refreshed in real time, the direction of its resultant force is constantly updated. Even if the three forces are in a state of equilibrium at a certain moment, it will be broken in the next frame. Figure 5 (A.iii) This avoids falling into a static minimum. Under this mechanism, the three particles circumvent the obstacle and continue to advance towards the target.

[0076] Specifically, during the motion, red_3 mainly moves along the y-axis, but at approximately 1.8 seconds, it experiences a slight x-axis shift due to the repulsive force of the green_1 and blue_2 halos, resulting in a slight increase in its path length (see...). Figure 5 C). As the distance between particles increases, the repulsive effect weakens until it disappears, and all three particles eventually reach their respective endpoints accurately. Figure 5 A.iv)

[0077] Throughout the process, the statistical data are as follows: the maximum particle-halo displacement differences for green_1, blue_2, and red_3 are 6.5 μm, 6.4 μm, and 3.0 μm in the x-direction, respectively, and 6.1 μm, 5.6 μm, and 9.2 μm in the y-direction, respectively. The starting and ending points of red_3 are basically aligned on the y-axis, with minimal x-direction displacement except for a slight offset caused by repulsive compression. Figure 5 D). The average moving velocities of the three particles were 64.3 μm / s, 63.7 μm / s, and 50.5 μm / s, respectively, corresponding to aura velocities of 70.2 μm / s, 69.5 μm / s, and 55.2 μm / s, respectively. The maximum particle-aura velocity difference was 5.9 μm / s. Figure 5 E).

[0078] The above results verify that the system described in this invention can still stably achieve real-time path planning and effective collision avoidance under harsh working conditions such as multiple targets and complex path intersections, ensuring that each particle reaches the predetermined target point along an independent, collision-free trajectory, and has good scalability and stability.

[0079] Experiment 2

[0080] See Figure 6 To further verify the effectiveness of the path planning method based on Artificial Potential Field (APF) described in this invention under conditions of spatial obstacles and path conflicts, this embodiment is designed to examine the system's obstacle avoidance capability, real-time performance, and path stability in multi-object dense scenes.

[0081] In this embodiment, the system simultaneously manipulates four particles (blue_1, blue_2, green_3, and red_4), and a row of parallel obstacles is placed in the experimental area to simulate a typical confined passage environment. Specifically, four obstacles (such as...) are placed at coordinates (50, 550), (450, 550), (850, 550), and (1250, 550). Figure 6 (As shown in Ai). The system identifies two blue particles of the same category in the field of view, so target point allocation can be implemented by category or by ID. This embodiment uses a number binding method, setting the target points as: goal_1 = (50, 900), goal_2 = (450, 900), goal_3 = (850, 900), goal_4 = (1250, 900).

[0082] After initiating path planning, the four particles move simultaneously towards their respective endpoints, gradually approaching the obstacle zone. Upon reaching the edge of the obstacle, the distances of blue1, blue_2, green_3, and red_4 from their respective endpoints are 188.3 μm, 307.6 μm, 231.7 μm, and 215.5 μm, respectively. Figure 6 B). At this point, the system detects that there are path conflicts between red_4 and green_3, and between blue_2 and blue1 (e.g., ...). Figure 6 (as shown in A.ii).

[0083] Without any explicit scheduling or time interpolation, the system automatically adjusts the repulsive force between particles in the artificial potential field, allowing red_4 and blue_2 to preferentially enter the obstacle zone, while blue1 and green_3 are temporarily guided to a waiting area above the obstacle. During this process, the maximum deviations between red_4 and blue_2 and their corresponding auras are maintained within 25.4 μm and 23.9 μm, respectively. Figure 6 C) indicates that its controlled trajectory is highly consistent with the planned path.

[0084] In contrast, blue1 and green_3, which were in a waiting state, significantly decelerated and stabilized almost at the center of their respective halos, with particle-halo deviations decreasing to 4.6 μm and 2.3 μm, respectively. Subsequently, the system continuously updated the potential field force directions of green_3 and blue1, guiding them to bypass the obstacle area and complete path adjustments. Ultimately, all four particles safely and without collision reached their respective target points. Figure 6 A.iii) Verify the correctness of the path calculation and the cooperative control capability of the system under obstacle constraints.

[0085] Compared with traditional multi-objective path planning methods that often resolve conflicts by constructing time hierarchy graphs, adding waiting mechanisms, or globally replanning individual paths, the artificial potential field method proposed in this invention does not require explicit construction of a time axis or mandatory addition of delay logic. Instead, it naturally guides conflicting individuals to avoid conflicts by adjusting the direction of the resultant force in real time, significantly improving the computational efficiency and real-time response capability of the system.

[0086] Statistical data shows that during the entire manipulation process, the average speeds of blue1, blue_2, green_3, and red_4 were 87.2 μm / s, 60.6 μm / s, 38.0 μm / s, and 67.4 μm / s, respectively, while the corresponding average speeds of the auras were 94.6 μm / s, 66.1 μm / s, 39.4 μm / s, and 71.8 μm / s, with a maximum speed difference of only 7.4 μm / s. Figure 6 D) verified that the system maintains stable control and high consistency of operation even under high-density obstacle conditions.

[0087] In summary, this embodiment fully demonstrates that the artificial potential field method of the present invention is not only applicable to unobstructed free manipulation scenarios, but also capable of real-time and stable path planning and obstacle avoidance control under conditions of complex obstacle constraints and path conflicts, and has good versatility and scalability.

[0088] Experiment 3

[0089] See Figure 7 In Experiments 1 and 2, the multi-particle manipulation capabilities of the system described in this invention were verified in collision-free navigation and restricted obstacle avoidance environments, respectively. To further verify the feasibility of the system in biomedical applications, this embodiment designed and implemented real-time sorting experiments of various types of particles to evaluate the system's practical application potential in single-cell sorting tasks.

[0090] The model sample used was commercially available polystyrene microspheres. These microspheres are characterized by uniform particle size, stable composition, and good biocompatibility, and are widely used in simulation tests for cell and biological particle sorting. This experiment selected three types of color-labeled polystyrene microspheres (red, blue, and green) and used a flow rate of 1 μL·min⁻¹. -1 The flow rate is used to mix and inject the mixture into the inlet of the microfluidic chip.

[0091] The chip outlet is designed with three branch channels, corresponding from top to bottom to the destination areas of green, blue, and red microspheres. After system startup, by setting the spatial coordinates of the outlet area and activating the artificial potential field path planning module, different types of microparticles are navigated and sorted to their corresponding channels (e.g., ...). Figure 7 (As shown in A).

[0092] Once a particle enters the visual detection area, the system immediately completes the following process:

[0093] Identification and labeling: Identifying particle categories and determining their precise spatial locations;

[0094] ID assignment: Assign a unique ID to each particle and add it to the set D to be sorted;

[0095] Target generation: Create a halo navigation target corresponding to the particle type and add it to the halo set T;

[0096] Real-time navigation uses an artificial potential field method, with the aura driving the corresponding particles to move toward a preset exit area;

[0097] Status Update: The system continuously monitors the deviation between the particles and the halo. When the deviation exceeds a set threshold, a self-feedback correction process is triggered. If the particles successfully reach the target area and maintain a low deviation, they are removed from set D.

[0098] Clear the object in T;

[0099] New Particle Access: When a new particle enters the field of view, the system assigns it a target according to its ID order and adds it to the navigation sequence to achieve continuous and stable multi-target navigation.

[0100] During this process, the system does not set any path buffer or waiting mechanism, and relies entirely on the real-time resultant force direction update based on repulsion to achieve collision-free, high-purity real-time sorting of the three types of particles in a mixed flow state.

[0101] To evaluate the performance of the sorting system, multiple rounds of sample collection and particle counting analysis were performed at the chip inlet and three outlets (see [link to experiment]). Figure 7 BE). The results showed that the proportions of the three types of particles in the inlet of the mixed liquid were approximately 37.67% (red), 29.86% (blue), and 32.47% (green); the outlet purity after system navigation reached: 94.23% for red particles, 95.11% for blue particles, and 93.52% for green particles (see BE). Figure 7 F).

[0102] The results fully demonstrate that, without the need for pre-programmed patterns or path intersection markings, the system described in this invention can stably achieve high-purity, collision-free, real-time sorting of multiple types of particles based on visual feedback and artificial potential field planning. It has good practical application value and is especially suitable for life science research scenarios such as cell sorting, which have high requirements for response speed and spatial accuracy.

[0103] The above are merely specific embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. The scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for microparticle path planning within a microfluidic chip using optoelectronic tweezers, characterized in that, The method includes the following steps: Acquire image data from microfluidic chips and perform real-time identification and classification of particles in the images; Each particle is assigned a unique identifier, and the spatial coordinates of each particle and the coordinates of the target area are determined in real time. A virtual electrode is generated based on the spatial coordinates of each particle, and the spatial position of the virtual electrode is the current spatial position of the target particle. Define the attraction and repulsion sources in the microfluidic chip, calculate the attraction vector from the spatial position of the virtual electrode to the attraction source, and calculate the total repulsion vector of each repulsion source at the current spatial position of the virtual electrode. The gravitational vector and the total repulsive vector are combined to form a total force vector, and the direction of the total force vector is taken as the direction of motion of the target particle. At the same time, the moving speed of the target particle is determined. Based on the target particle's motion direction and speed, the next position of the target particle is predicted. The position of the target particle is updated through multiple iterations. The virtual electrode drives the target particle to move along the path of minimum potential energy to reach the target area.

2. The method for particle path planning within a microfluidic chip using optoelectronic tweezers according to claim 1, characterized in that, The YOLOv5 neural network is used to identify and classify particles in images in real time.

3. The method for particle path planning within a microfluidic chip using optoelectronic tweezers according to claim 1, characterized in that, The ByteTrack multi-target tracking algorithm assigns a unique identifier to each particle and determines the spatial coordinates of each particle and the coordinates of the target area in real time.

4. The method for particle path planning within a microfluidic chip using optoelectronic tweezers according to claim 1, characterized in that, In a microfluidic chip, the gravitational source is the target region, and the repulsive source is other particles besides the target particle, as well as preset obstacles and channel boundaries.

5. The method for particle path planning within a microfluidic chip using optoelectronic tweezers according to claim 1, characterized in that, The total repulsive force vector is obtained by superimposing the repulsive force vectors of each repulsive force source. When calculating the repulsive force vector, the magnitude of the repulsive force of each repulsive force source is adjusted according to the distance between it and the virtual electrode. The smaller the distance, the greater the repulsive force. When the distance exceeds the preset threshold, the repulsive force is zero.

6. The method for particle path planning within a microfluidic chip using optoelectronic tweezers according to claim 1, characterized in that, When calculating the gravitational vector, the magnitude of gravity is adjusted according to the distance from the virtual electrode to the gravitational source. The farther the distance, the greater the gravity, and the gravity is zero when the distance is zero.

7. The method for particle path planning within a microfluidic chip using optoelectronic tweezers according to claim 1, characterized in that, During the multiple iterations, each iteration recalculates the gravitational vector, repulsive vector, and total force vector based on the latest spatial coordinates of the target particle, and updates the next position accordingly.

8. The method for particle path planning within a microfluidic chip using optoelectronic tweezers according to claim 1, characterized in that, The total force vector always guides the target particle to move in the direction of decreasing potential energy, and each position update makes the target particle closer to the target area and further away from the repulsive force source.

9. The method for particle path planning within a microfluidic chip using optoelectronic tweezers according to claim 1, characterized in that, The moving speed of the target particle is determined by combining the magnitude of the total force vector with the step size coefficient.

Images