Capacity-aware continuous microfluidic biochip cleaning optimization method

Abstract

The application provides a capacity-aware continuous microfluidic biochip cleaning optimization method, which comprises the following steps: step one, a fluid routing algorithm based on a coordination mechanism is used to calculate a flow path for a fluid transport task, and path conflict problems between grouped fluid transport tasks are considered through pollution-aware flow path planning; step two, a path-driven fluid scheduling algorithm is used to obtain an accurate execution interval for the fluid transport task and the cleaning task, and the storage position of intermediate fluid is obtained through a pre-calculated channel usage file; step three, a capacity-aware cleaning optimization algorithm is used to obtain a cleaning path set to cover all cleaning targets, and the actual buffer capacity constraint and the resource conflict constraint of the buffered liquid are considered; the application can strictly meet the given buffer capacity constraint condition, greatly avoid the conflict of various flow processing tasks, optimize the total length of the flow channel and the number of intersection points, and reduce the construction cost of the chip.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of continuous microfluidic biochip automation design technology, in particular to a capacity-aware continuous microfluidic biochip cleaning optimization method. BACKGROUND

[0002] As a new chip technology, continuous microfluidic biochip has attracted extensive attention due to its ability to perform various biochemical reactions in parallel and automatically on a coin-sized chip platform.

[0003] In order to complete the dynamic scheduling and distributed channel storage of fluid, the flow channel needs to be used by different flow paths, so that buffer solution is needed to clean the contaminants left by the previous fluid on the channel, and the cleaning demand is also increased.

[0004] However, the buffer solution per unit volume has limited capacity, so it cannot handle all cleaning targets. Therefore, this paper considers the cleaning optimization problem under the constraint of limited capacity, and proposes a capacity-aware continuous microfluidic biochip flow scheduling and cleaning optimization design process, which is committed to the optimization of flow path planning and cleaning path calculation. SUMMARY

[0005] The present application proposes a capacity-aware continuous microfluidic biochip cleaning optimization method, which can obtain a flow scheduling and cleaning optimization scheme that minimizes the total biochemical reaction execution time, cleaning time and buffer time, and a flow layer channel network solution that minimizes the total length of the flow channel.

[0006] The present application adopts the following technical solutions.

[0007] The capacity-aware continuous microfluidic biochip cleaning optimization method is used in the constraint scene of limited buffer solution capacity, and includes the following steps:

[0008] Step one, a fluid routing algorithm based on coordination mechanism is used to calculate the corresponding flow path for each fluid transportation task, and through pollution-aware flow path planning, the system considers the path conflict problem between the grouped fluid transportation tasks to effectively reduce the subsequent cleaning demand;

[0009] Step two, a path-driven fluid scheduling algorithm is used to obtain the accurate execution interval for fluid transportation tasks and cleaning tasks, and through the pre-calculated channel usage file, the buffer position of the intermediate fluid is obtained to improve the access efficiency and reduce the total biochemical reaction execution time;

[0010] Step three, the capacity-aware cleaning optimization algorithm is used to obtain the cleaning path set to cover all cleaning targets and consider the actual buffer capacity constraints and resource conflict constraints of the buffer liquid, so as to avoid cleaning failure and enable the biochemical reaction to be efficiently and correctly executed.

[0011] Step one is a flow channel wiring stage, specifically: a wiring algorithm based on a negotiation mechanism is used to calculate a complete flow path for each flow processing task; the flow processing task is first grouped according to its execution interval, tasks with similar intervals are grouped into the same group, then the tasks in the same group are negotiated and wired according to the constraint that the tasks in the same group can cross but do not share flow channels, and finally the tasks that fail to be wired are sequentially wired according to the constraint that the tasks in different groups can share flow channels;

[0012] Step two is a fluid scheduling stage, specifically: a path-driven fluid scheduling algorithm is used to sequentially schedule the flow processing tasks in each group; the tasks in the same group do not share flow channels and there is no cross-contamination problem; after a group of tasks is completed, a cleaning path set is calculated and the corresponding execution interval is determined according to the defined cleaning target; when intermediate fluid needs to be cached, the least conflicting flow channel with other paths is selected on the pre-planned flow path to execute the caching task;

[0013] Step three is a cleaning optimization stage, specifically: a capacity-aware cleaning optimization method is used to calculate a group of cleaning path sets and the corresponding execution intervals; first, the cleaning target set is determined based on the pre-calculated channel usage protocol, then a group of cleaning path sets is obtained by using a depth-first search algorithm under the given buffer capacity constraints and channel resource constraints of the buffer liquid, and finally the execution interval of each cleaning path is obtained under the constraint of channel resource conflict.

[0014] The continuous microfluidic biochip includes a double-layer architecture composed of two elastic layers, one of which is a flow layer and the other is a control layer, and both layers are provided with channels to form a distributed channel storage architecture that can execute flow processing tasks.

[0015] The channels in the flow layer are flow channels for transporting reaction samples / reagents; the channels in the control layer are control channels for conducting air pressure; the channels are made of elastic material, and the intersection position of the channels of the two elastic layers acts as a valve, which can be regarded as a valve; the channels of the two elastic layers are connected to an external pressure source.

[0016] In the control layer, external pressure is introduced through the control port, which presses the membrane at the valve to squeeze downward, so that the channel segment is blocked and cannot transport liquid, and when the pressure is removed, the membrane restores to the initial position by its own elastic force; in the flow layer, the liquid is transported by the pressure introduced through the flow port.

[0017] In the distributed channel storage architecture, the intermediate fluid can be cached in any idle flow channel, so that the flow channel has the dual role of transportation and storage, and the flow processing tasks performed by the flow channel include the following:

[0018] Task A1, fluid transportation one: transport intermediate fluid from one component to another, which corresponds to the transportation path in the chip: input port -> source component -> target component -> waste port;

[0019] Task A2, fluid transportation two: transport input samples / reagents to target components, which corresponds to the transportation path in the chip: input port -> target component -> waste port;

[0020] Task A3, fluid storage: transport intermediate fluid to flow channel cache, which corresponds to the transportation path in the chip: input port -> source component -> cache channel -> waste port;

[0021] Task A4, fluid extraction: transport intermediate fluid from the cache channel to the target component, which corresponds to the transportation path in the chip: input port -> cache channel -> target component -> waste port.

[0022] In step two, the distributed channel storage architecture cleans the residual liquid slag on the flow channel through a cleaning operation to avoid cross contamination between fluids; in the cleaning operation, the buffer liquid injected from the external port flows through the contaminated flow channel / component for cleaning, and the upper limit of the contaminant that can be cleaned by the buffer liquid is related to the volume of the injected buffer liquid; in the cleaning operation, when the buffer liquid flows through the contaminated flow channel / component that is not the cleaning target due to the limitation of the flow channel, the available capacity of the buffer liquid will be consumed, resulting in an increase in the number of cleaning operations;

[0023] When the maximum allowable consumption value of the contaminant that can be cleaned by a unit volume of buffer liquid needs to be set, let ψ represent the cleaning capacity value of a unit volume of buffer liquid; assume that μ is the consumption value of the processing cleaning target, and v is the consumption value of the contaminated flow channel / component flowing through the non-cleaning target; thus, the actual cleaning capacity constraint of the buffer liquid is defined as:

[0024] μ + v ≤ ψ Formula 1;

[0025] In step two, the shared flow channel needs buffer liquid to perform the cleaning operation, and the cleaning needs between different flow paths need to be systematically considered to improve the execution efficiency of the transportation task and thus speed up the completion of the biochemical reaction; the method for grouping fluid transportation tasks is as follows:

[0026] Let each transportation task tk i ∈T be first associated with an execution interval (st i ,et i), to indicate the ith transport task from time point st i to et i is scheduled to be executed; then, these transport tasks are sorted in non-decreasing order according to their start execution times, and are traversed in turn; and a dynamic task group G t is used to record those transport tasks with similar execution intervals. When a task tk i is traversed, it can be inserted into G t if and only if G t is empty or the following inequality holds:

[0027] st i -et i-1 <<δ(i>1) Equation 2;

[0028] where st i is the start time of tk i , et i-1 is the end time of tk i-1 , and δ is a user-defined threshold; otherwise, G t is saved, and the traversal is continued with a new dynamic task group, tk i being the first element of the new dynamic task group. Finally, when the traversal is completed, a plurality of task groups are formed, each of which contains a plurality of transport tasks that need to be scheduled to be executed without pollution, and the transport tasks in the same group should avoid transport conflicts to reduce cleaning tasks; while the cleaning tasks between different groups can be executed in parallel with the biochemical reaction operations without conflicts, thereby improving the execution efficiency of the biochemical reaction.

[0029] In step three, a fluid routing method based on a negotiation mechanism is used to obtain the cleaning path set, in which a routing graph is divided into a connection graph generated by a connection grid; each grid point is represented as a flow / waste liquid port, an inlet / outlet port of a component, or a cross point; each edge connecting two grid points is represented as a channel segment that can perform fluid buffering and transport; the connection graph formed constructs the required flow paths for all transport tasks in the scheduling; each routing grid is associated with a historical cost H at the rth iteration, and the update function of the historical cost is calculated as:

[0030] F(n i )=G(n i )+H(n i )+C(n i ) Equation 4;

[0031]

[0032] where G(n i) represents the path length from the source point to n i , H(n i ) represents the estimated path length from n i to the target point, C(n i ) represents the additional wiring cost, represents the historical cost of grid point n i at the rth iteration, U r (n i ) is a binary (0 / 1) variable value to represent whether n i has been constructed by the flow path of the same group of transportation tasks, and C c is a parameter value defined by the user to represent the crossing cost of a used grid point. Finally, for the tasks that fail to wire, the wiring is performed in turn according to the constraint that different groups of tasks can share the flow channel. At this time, the modified A* routing algorithm is also used to plan the flow path, and the wiring cost F(n i ) of the current search grid point n i can be calculated by formula five, and the calculation method of C(n i ) is different, which is expressed in formula as:

[0033]

[0034] Wherein, C w is the cleaning cost, C c is the crossing cost, C S is a positive integer defined by the user; task(n i ) is the transportation task currently occupying the grid point n i , and tk j is the transportation task currently planning the flow path;

[0035] After completing the fluid wiring, a set of flow paths P corresponding to all transportation tasks and a flow channel network in a single chip architecture are obtained. According to the two results, a fluid scheduling scheme is executed, so that all flow processing tasks can be effectively executed without conflict on the generated chip architecture. All transportation tasks will be scheduled in turn according to the grouping, and there is no cross contamination between the flow path scheduling of the same group, effectively reducing the number of time-consuming cleaning operations. After the fluid transportation tasks of the same group are completed, the cleaning target is determined according to the channel use protocol, a set of cleaning paths is obtained to cover all cleaning targets, so that the transportation tasks of different groups can quickly reuse the flow channels / components previously passed by the flow, thereby speeding up the execution process of biochemical reactions.

[0036] The cleaning optimization method further includes a path preparation and intermediate fluid caching strategy, specifically:

[0037] For the fluid transportation task group g iThe flow path corresponding to the j-th task in It is necessary to determine at what time it can be scheduled. First, if Target component c b Being flushed by other liquids tran Occupied, and the liquid is not If the input liquid is bound to the operation, then the liquid needs to be transported into the flow channel for storage first;

[0038] Secondly, assuming flu tran The bound flow path is t_path. An idle flow channel segment is found along this flow path to buffer the liquid, while minimizing conflicts with other transport tasks. Therefore, each flow channel d... k，l They are all associated with an evaluation value ev k，l This is to determine the superiority of selecting a stream channel as a cache location; then, a stream channel can be selected as a cache channel based on the following evaluation criteria:

[0039] 1) To avoid altering the pre-calculated flow paths for other fluid transport tasks, the liquid buffer location can be selected within its corresponding flow path.

[0040] 2) To avoid potential conflicts between fluid transport and buffering, for fluid... tran Each flow channel d on the corresponding flow path k，l If d is used next k，l The earlier the fluid transport task is scheduled, the better the ev k，l The larger the value of ev, the higher the flow path. k，l The stream channel with the smallest value will be selected as the buffer location;

[0041] Once the cache location and corresponding cache path are determined, the intermediate liquid is temporarily cached within the flow channel. Component c b It can then be used to receive The liquid being transported;

[0042] Finally, it is necessary to ensure Each flow channel segment is available. If a flow channel segment is occupied by a cached liquid, the corresponding subpath needs to be reconstructed to bypass the cached channel. If the subpath cannot be reconstructed, a new flow channel needs to be introduced to form a valid flow path.

[0043] In the cleaning optimization method, the scheduling method for fluid extraction and transportation is as follows:

[0044] For a fluid transport task group g i The j-th fluid transport task If the liquid it is bound to The liquid is currently cached within the streaming channel; it can be marked as `flu_cache` and needs to be retrieved from the cache location `cs`. c Extract to target component c b The corresponding extraction path and execution range also need to be calculated, consisting of three sub-paths: f_port→cs. c CS c →c b , and c b →w_port is obtained using Dijkstra's pathfinding algorithm;

[0045] Once the extraction path f_path is calculated, the corresponding start time t start (f_path) depends on the following three conditions: 1) the target component c b It's all ready, 2) Bound operations All is ready, and 3) the extraction path f_path is also ready; therefore, t start (f_path) is calculated as follows:

[0046]

[0047] Where max(.) represents a maximum value function, t ready (c b () is the readiness time of the target component. It is a task The ready time of the bound operation, and t ready (f_path) is the ready time for path extraction;

[0048] if The bound liquid It's still on the source component, and the flow path it's bound to. If everything is ready, the liquid can be transported directly from the source component to the target component; depending on Preparation time, and depending on The shipping delay.

[0049] A cleaning operation is triggered to perform cleanup when the target component is contaminated or when all legitimate flow paths constructed for fluid transport are unavailable due to contaminated flow channels. The cleaning optimization method also includes a method for detecting cleaning targets, used to identify a set of contaminated areas that actually need cleaning, thereby reducing the time-consuming cleaning process. Specifically:

[0050] Let each flow channel d k，l Associated with a parameter u k，lto represent the number of flow paths using the channel, when the pre-constructed flow path uses the flow channel, the value of u k，l needs to be reduced by one; and, each flow channel d k，l is also associated with a Boolean value s k，l to represent the current state of the flow channel, where "0" represents a clean state, and "1" represents a contaminated state; similarly, each component c j is also associated with a count parameter u j and a Boolean value s j ; s k，l and s j are initially set to "0" respectively, and are updated according to the following rules:

[0051] Rule 1, after completing each flow processing task, the s k，l and s j values of each flow channel d k，l and component c j that the flow path passes through are set to "1";

[0052] Rule 2, after completing each cleaning operation, the s k，l and s j values of each flow channel d k，l and component c j that the cleaning path passes through are set to "0";

[0053] Rule 3, after completing each flow processing task, the u k，l and u j values of each flow channel d k，l and component c j that the flow path passes through are reduced by one; when the value is 0, the corresponding flow channel / component will not need to be used again.

[0054] In the cleaning operation, when the buffer solution per unit volume has a limited cleaning capacity, cannot be moved on the chip without considering its capacity demand on the contaminated flow channel / component, and a single fixed volume of buffer solution cannot handle all cleaning targets, the cleaning capacity demand analysis of each contaminated flow channel / component uses the following method:

[0055] Case 1, when cleaning the flow channel d k，l , its corresponding capacity demand cd k，l is calculated as:

[0056] cd k，l = len k，l x T w Equation Eight;

[0057] Case 2, when cleaning the component c bAt that time, its corresponding capacity requirement is cd b The calculation is as follows:

[0058] cd b =vol b ×T w Formula Nine;

[0059] Scenario 3: If a cleaning target is connected to another contaminated flow channel / component, the two will form a new cleaning target, and the capacity requirement of this cleaning target will be the sum of the values.

[0060] The cleaning optimization method also includes a capacity-aware cleaning path calculation method, specifically:

[0061] Let Tg be a set of cleaning targets w ={tgw1, tgw2, ..., tgw n The corresponding cleaning capacity requirement is Cd = {cd1, cd2, ..., cd}. n A set of optimized paths is determined through cleaning path calculation to cover all cleaning targets and minimize the total cleaning time.

[0062] First, construct a weighted undirected graph G. t (Q, R) represents the connection between the external port and the cleaning target, where Q is the connection between the external port Pt and the cleaning target Tg. w The components are composed of edges, and the edges represent the connectivity between them. For each edge r... i，j ∈R, its weight value w i，j This represents the cost of buffer flow between associated nodes, and can be calculated as:

[0063] w i，j =dis i，j +cd i，j Formula 10;

[0064] Among them, dis i，j For point q i With q j The length of the flow path, and cd i，j The cleaning capacity requirement for the contaminated flow channels / components included in this flow path;

[0065] The flow path between the two points is obtained using Dijkstra's algorithm, and flow channels / components with existing fluid will not be selected. Secondly, on the generated undirected graph, a set of cleaning paths needs to be found, where each path starts at the input port, flows through the cleaning target, and finally ends at the output port, ensuring that the total cleaning capacity requirement is less than the maximum cleaning capacity of the given buffer solution. Therefore, a depth-first search-based capacity-aware cleaning calculation method is adopted, specifically:

[0066] The search process begins with a cleaning target that is closest to the input port, and then traverses the entire graph until an output port is found; a node table is created to record all available nodes and is updated after each search: an adjacent node is selected if its corresponding edge has the smallest weight value, and previously selected nodes are removed from the table to prevent cycles; backtracking is performed when there are no child nodes or the cleaning capacity of the buffer is insufficient.

[0067] When there are still cleaning targets, a new round of path search will start with the cleaning target closest to the input port. Unless there are no available nodes, nodes covered by the previously determined cleaning path will not be selected by the current path. The search ends when all cleaning targets are covered by the set of paths.

[0068] This invention considers the cleaning optimization problem under finite capacity constraints and proposes a capacity-aware continuous microfluidic biochip flow scheduling and cleaning optimization design process to optimize flow path planning and cleaning path calculation. It can obtain a flow scheduling and cleaning optimization scheme that minimizes the total execution time of biochemical reaction, cleaning time and buffering time, as well as a flow layer channel network solution that minimizes the total length of flow channels.

[0069] This invention, while strictly meeting the given buffer capacity constraints, greatly avoids conflicts between various stream processing tasks, and also optimizes the total length of the stream channels and the number of intersections, thereby reducing the chip construction cost. Attached Figure Description

[0070] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments:

[0071] Appendix Figure 1 This is a schematic diagram of a continuous microfluidic biochip architecture;

[0072] Appendix Figure 2 This is a schematic diagram of a stream processing task in a microfluidic biochip with a distributed channel storage architecture.

[0073] Appendix Figure 3 This is a schematic diagram illustrating the effectiveness of distributed channel storage;

[0074] AppendixFigure 4 is a schematic diagram of a grouping of fluid transport tasks. DETAILED DESCRIPTION

[0075] As shown, the capacity-aware continuous microfluidic biochip cleaning optimization method for a buffer capacity limited constraint scenario includes the following steps:

[0076] Step one, a fluid routing algorithm based on a coordination mechanism is used to calculate the corresponding flow path for each fluid transport task. Through pollution-aware flow path planning, the system considers the path conflict problem between the grouped fluid transport tasks to effectively reduce the subsequent cleaning requirements.

[0077] Step two, a path-driven fluid scheduling algorithm is used to obtain accurate execution intervals for fluid transport tasks and cleaning tasks. Through the pre-calculated channel usage file, the cache location of the intermediate fluid is obtained to improve access efficiency and reduce the total time of biochemical reaction execution.

[0078] Step three, a capacity-aware cleaning optimization algorithm is used to obtain a cleaning path set to cover all cleaning targets and consider the actual buffer capacity constraints and resource conflict constraints with cached liquids, thereby avoiding cleaning failures and enabling biochemical reactions to be efficiently and correctly executed.

[0079] Step one is the flow channel routing phase, specifically: a routing algorithm based on a negotiation mechanism is used to calculate the complete flow path for each flow processing task. The flow processing task is first grouped according to its execution interval, and tasks with similar intervals are grouped together. Then, according to the constraints that tasks in the same group can be crossed but not shared, and tasks in different groups cannot share flow channels, the routing is negotiated. Finally, for tasks that fail to route, according to the constraint that tasks in different groups can share flow channels, the routing is performed in sequence.

[0080] Step two is the fluid scheduling phase, specifically: a path-driven fluid scheduling algorithm is used to schedule each grouped flow processing task in sequence. There is no problem of sharing flow channels and cross contamination between tasks in the same group. After a group of tasks is completed, according to the defined cleaning target, a cleaning path set is calculated and the corresponding execution interval is determined. When intermediate fluids need to be cached, the flow channel with the least conflict with other paths is selected on the pre-planned flow path to perform the caching task.

[0081] The third step is a cleaning optimization stage, specifically: a capacity-aware cleaning optimization method is used to calculate a set of cleaning path sets and corresponding execution intervals; first, a cleaning target set is determined based on a pre-calculated channel usage protocol, then a set of cleaning path sets is obtained by using a depth-first search algorithm under the constraints of given buffer capacity constraints and channel resource constraints of the buffer solution, and finally the execution interval of each cleaning path is obtained under the constraint of channel resource conflict.

[0082] As shown in Figure 1 The continuous microfluidic biochip includes a double-layer architecture composed of two elastic layers, one of which is a flow layer and the other is a control layer, and channels are arranged in both elastic layers to form a distributed channel storage architecture that can perform flow processing tasks.

[0083] The channels in the flow layer are flow channels for transporting reaction samples / reagents; the channels in the control layer are control channels for conducting air pressure; the channels are made of elastic material, and the intersection position of the channels of the two elastic layers acts as a valve, which can be regarded as a valve; the channels of the two elastic layers are connected to an external pressure source.

[0084] In the control layer, external pressure is introduced through the control port, which presses the membrane at the valve to squeeze downward, so that the channel segment is blocked and cannot transport liquid, and when the pressure is removed, the membrane restores to the initial position by its own elastic force; in the flow layer, the liquid is transported by the pressure introduced through the flow port.

[0085] As shown in Figure 2 In the distributed channel storage architecture, intermediate fluid can be buffered in any idle flow channel, so that the flow channel has the dual functions of transportation and storage, and the executed flow processing tasks include the following contents:

[0086] Task A1, fluid transportation I: transport intermediate fluid from one component to another, and the corresponding transportation path in the chip is: input port->source component->target component->waste port;

[0087] Task A2, fluid transportation II: transport the input sample / reagent to the target component, and the corresponding transportation path in the chip is: input port->target component->waste port;

[0088] Task A3, fluid storage: transport intermediate fluid to flow channel buffer, and the corresponding transportation path in the chip is: input port->source component->buffer channel->waste port;

[0089] Task A4, fluid extraction: transport intermediate fluid from the buffer channel to the target component, and the corresponding transportation path in the chip is: input port->buffer channel->target component->waste port.

[0090] In this example, under the distributed channel storage architecture, intermediate fluids can be cached in any idle channel, allowing the channels to perform a dual function of transport and storage. While this can improve the efficiency of biochemical reactions and reduce chip fabrication costs, the resource conflicts between transport and caching still need to be carefully considered. Figure 3 As shown, when component C3 needs to receive the intermediate liquid produced by C2 and C5, the target component C4, where the intermediate fluid produced by C3 is, is performing other biochemical reactions. Therefore, the intermediate fluid produced by C3 needs to be transported to a certain flow channel for buffering. At this time, the flow channel d... 6，7 d 6，11 d 7，12 and d 11，12 All can be used as alternative buffer channels. However, flow paths p1, p2, and p3 will be scheduled sequentially; therefore, when flow channel d... 6，7 d 6，11 and d 7，12 When used as a buffer channel, it will cause the three flow paths to be transported at longer distances. Therefore, d 11，12 That is the most effective cache location.

[0091] In step two, the distributed channel storage architecture uses a cleaning operation to remove residual liquid residue on the flow channels to avoid cross-contamination between fluids. During the cleaning operation, buffer solution injected from an external port flows through the contaminated flow channels / components for cleaning. The upper limit of the contaminant capacity that the buffer solution can clean is related to the volume of the injected buffer solution. During the cleaning operation, when the buffer solution flows through contaminated flow channels / components that are not the cleaning target due to the limitations of the flow channels, the available capacity of the buffer solution will be consumed, resulting in an increase in the number of cleaning operations.

[0092] When setting the maximum allowable consumption of contaminants that a unit volume of buffer solution can clean, let ψ represent the cleaning capacity per unit volume of buffer solution; assuming μ is the consumption value for the target being cleaned, and υ is the consumption value for contaminant flow channels / components flowing through non-target areas; therefore, the actual cleaning capacity constraint of the buffer solution is defined as:

[0093] μ+υ≤ψ Formula 1;

[0094] In step two, the shared flow channels require buffer solution for cleaning. The cleaning requirements between different flow paths need to be systematically considered to improve the efficiency of the transport task and thus accelerate the completion of the biochemical reaction. The specific method for grouping fluid transport tasks is as follows:

[0095] Let each transportation task be tk. i ∈T is first associated with an execution interval (st i et iLet ), to represent the i-th transportation task starting from time point st. i to et i The tasks are scheduled for execution; then, these transportation tasks are sorted in non-decreasing order according to their start times, and traversal is performed sequentially; and a dynamic task group G is used. t This is used to record transportation tasks with similar execution ranges. Whenever a task tk is encountered during iteration... i At that time, it can be inserted into G t If and only if G t When the set is empty or the following inequality holds:

[0096] st i -et i-1 <<δ(i>1) Formula 2;

[0097] Among them, st i It's TK i The start time, et i-1 It's TK i-1 The end time and δ are user-defined thresholds; otherwise, G t This will be saved and the traversal will continue with a new dynamic task group, tk i This is the first element of this new dynamic task group; finally, when the traversal is completed, multiple task groups are formed, each containing multiple transport tasks that need to be scheduled and executed without contamination. Transport tasks within the same group should avoid transport conflicts to reduce cleaning tasks; while cleaning tasks between different groups can be executed in parallel with biochemical reaction operations without conflict, thereby improving the execution efficiency of biochemical reactions.

[0098] like Figure 4 The diagram illustrates a scheduling scheme for a biochemical reaction, where four components (c1-c4) are used to execute nine operations (o1-o9), and ten transport tasks (indicated by blue boxes). After the grouping method described above, these transport tasks are divided into four groups (g1-g4). Transport tasks within the same group should avoid transport conflicts as much as possible to reduce cleaning tasks. Cleaning tasks between different groups can be executed in parallel with the biochemical reaction operations without conflict, thereby improving the execution efficiency of the biochemical reaction.

[0099] In step three, a negotiation-based fluid routing method is used to obtain the cleaning path set. In this method, the routing graph is divided into a connection graph generated by a connecting grid. Each grid point represents a flow / waste port, component inlet / outlet port, or intersection. Each edge connecting two grid points represents a channel segment capable of performing fluid buffering and transportation. The resulting connection graph constructs the required flow paths for all transportation tasks in the scheduling process. Each routing grid is associated with a historical cost in the r-th iteration. Furthermore, the update function for historical costs is calculated as follows:

[0100] F(n i )=G(n i )+H(n i )+C(n i Formula 4;

[0101]

[0102] Where G(n) i ) represents the distance from the source to n i The path length, H(n) i ) indicates from n i The estimated path length to the target point, C(n) i ) indicates the additional wiring cost. This represents the grid point n at the r-th iteration. i The historical cost, U r (n i ) is a binary (0 / 1) variable value to indicate whether n i There are flow paths constructed by the same group of transportation tasks, and C c User-defined parameter values ​​are used to represent the cross cost of a used grid point. Finally, for tasks that fail to route, routing is performed sequentially based on the constraint that different groups of tasks can share flow paths. At this point, a modified A* pathfinding algorithm is also used to plan the flow path, with the current searched grid point n... i The wiring cost F(n) i The same formula five can be used to calculate C(n). i The calculation methods for ) are different, and are expressed by the following formula:

[0103]

[0104] Among them, C w In exchange for cleaning, C c For the crossover cost, C S For a user-defined positive integer; task(n) i ) represents the currently occupied grid point n i The transportation mission, and tk j For the current transportation task involving flow path planning;

[0105] After completing the fluid routing, a set of flow paths P corresponding to all transport tasks and a flow channel network in a single chip architecture are obtained. Based on these two results, a fluid scheduling scheme is executed so that all flow processing tasks can be executed efficiently and without conflict on the generated chip architecture. All transport tasks are scheduled sequentially according to the grouping situation, and there is no cross-contamination between flow paths in the same group, which effectively reduces the number of time-consuming cleaning operations. After all the fluid transport tasks in the same group are completed, the cleaning target is determined according to the channel usage protocol, and a set of cleaning paths is obtained to cover all cleaning targets. Thus, transport tasks in different groups can quickly reuse the flow channels / components that have been previously flowed through, thereby accelerating the execution process of the biochemical reaction.

[0106] The cleaning optimization method also includes path preparation and intermediate fluid caching strategies, specifically:

[0107] For fluid transport task group g i The flow path corresponding to the j-th task in It is necessary to determine at what time it can be scheduled.

[0108] First, if Target component c b Being flushed by other liquids tran Occupied, and the liquid is not If the input liquid is bound to the operation, then the liquid needs to be transported into the flow channel for storage first;

[0109] Secondly, assuming flu tran The bound flow path is t_path. An idle flow channel segment is found along this flow path to buffer the liquid, while minimizing conflicts with other transport tasks. Therefore, each flow channel d... k，l They are all associated with an evaluation value ev k，l This is to determine the superiority of selecting a stream channel as a cache location; then, a stream channel can be selected as a cache channel based on the following evaluation criteria:

[0110] 1) To avoid altering the pre-calculated flow paths of other fluid transport tasks, the liquid buffer location can be selected within its corresponding flow path.

[0111] 2) To avoid potential conflicts between fluid transport and buffering, for fluid... tran Each flow channel d on the corresponding flow path k，l If d is used next k，l The earlier the fluid transport task is scheduled, the better the ev k，l The larger the value of ev, the higher the flow path. k，l The stream channel with the smallest value will be selected as the buffer location;

[0112] Once the cache location and corresponding cache path are determined, the intermediate liquid is temporarily cached within the flow channel. Component c b It can then be used to receive The liquid being transported;

[0113] Finally, it is necessary to ensure Each flow channel segment is available. If a flow channel segment is occupied by a cached liquid, the corresponding subpath needs to be reconstructed to bypass the cached channel. If the subpath cannot be reconstructed, a new flow channel needs to be introduced to form a valid flow path.

[0114] In the cleaning optimization method, the scheduling method for fluid extraction and transportation is as follows:

[0115] For a fluid transport task group g i The j-th fluid transport task If the liquid it is bound to The liquid is currently cached within the streaming channel; it can be marked as `flu_cache` and needs to be retrieved from the cache location `cs`. c Extract to target component c b The corresponding extraction path and execution range also need to be calculated, consisting of three sub-paths: f_port→cs. c CS c →c b , and c b →w_port is obtained using Dijkstra's pathfinding algorithm;

[0116] Once the extraction path f_path is calculated, the corresponding start time t start (f_patth) depends on the following three conditions: 1) the target component c b It's all ready, 2) Bound operations All is ready, and 3) the extraction path f_path is also ready; therefore, t start (f_path) is calculated as follows:

[0117]

[0118] Where max(.) represents a maximum value function, t ready (c b () is the readiness time of the target component. It is a task The ready time of the bound operation, and t ready (f_path) is the ready time for path extraction;

[0119] if bound liquid Also on the source component, and its bound flow path If the source component has been prepared, the liquid can be directly transported from the source component to the target component; Depending on the preparation time, and Depending on the transport delay.

[0120] When the target component is contaminated or all legal flow paths built for fluid transport are not available due to contaminated flow channels, a cleaning operation is triggered to perform cleaning; the cleaning optimization method also includes a cleaning target detection method to find a set of contaminated places that actually need to be cleaned, thereby reducing the time-consuming cleaning process, in particular:

[0121] Let each flow channel d k，l be associated with a parameter u k，l to represent the number of flow paths using the channel, when the pre-built flow path uses the flow channel, the value of u k，l needs to be reduced by one; and each flow channel d k，l is also associated with a Boolean value s k，l to represent the current state of the flow channel, where "0" represents a clean state, and "1" represents a contaminated state; Similarly, each component c j is also associated with a count parameter u j and a Boolean value s j ; s k，l and s j are initially set to "0" respectively, and are updated according to the following rules:

[0122] Rule 1, after completing each flow processing task, the s k，l and s j values of each flow channel d k，l and component c j passed through by the flow path are set to "1";

[0123] Rule 2, after completing each cleaning operation, the s k，l and s j values of each flow channel d k，l and component c j passed through by the cleaning path are set to "0";

[0124] Rule 3, after completing each flow processing task, the u k，l and u j values of each flow channel d k，l and component c jall values are reduced by one; when the value is 0, the corresponding flow channel / component will not need to be used any more.

[0125] When the buffer solution in unit volume has limited washing capacity, cannot be moved on the chip arbitrarily without considering its capacity demand on the contaminated flow channel / component, and a single fixed volume of buffer solution cannot handle all washing targets, the washing capacity demand analysis of each contaminated flow channel / component uses the following method in the washing operation:

[0126] Case one, when the washing flow channel d k，l , its corresponding capacity demand cd k，l is calculated as:

[0127] cd k，l = len k，l × T w Equation Eight;

[0128] Case two, when the washing component c b , its corresponding capacity demand cd b is calculated as:

[0129] cd b = vol b × T w Equation Nine;

[0130] Case three, for a washing target, if the contaminated flow channel / component connected thereto is also another washing target, the two form a new washing target, and the capacity demand of the washing target is the cumulative value.

[0131] The washing optimization method further includes a capacity-aware washing path calculation method, specifically:

[0132] Suppose a set of washing targets Tg w = {tgw1, tgw2, …, tgw n}, and its corresponding washing capacity demand is Cd = {cd1, cd2, …, cd n}; determine a set of optimized paths to cover all washing targets and minimize the total washing time through washing path calculation;

[0133] First, construct a weighted undirected graph G t (Q, R) to represent the connection between the external port and the washing target, wherein Q is composed of external ports Pt and washing targets Tg w , and the edge represents the connection between them. For each edge r i，j ∈ R, its weight value w i，j represents the flow cost of the buffer solution between the associated nodes, and can be calculated as:

[0134] w i，j = dis i，j + cd i，j Equation ten;

[0135] where dis i，j is the flow path length from point q i to q j , and cd i，j is the washing capacity requirement of the flow path containing the contamination flow channel / component;

[0136] The flow path between two points is calculated according to Dijkstra algorithm, and the flow channel / component currently containing fluid will not be selected; secondly, on the generated undirected graph, a set of washing paths needs to be found, each of which starts from the input port, flows through the washing target, and finally ends at the output port, and makes the total washing capacity requirement less than the maximum washing capacity of the given buffer; therefore, a capacity-aware washing calculation method based on depth-first search is adopted, which is as follows:

[0137] The search process starts from a washing target closest to the input port, and then traverses the entire graph until an output port is found; a node table is created to record all available nodes, and is updated after each search: an adjacent point can be selected if its corresponding edge has the smallest weight value, and the previously selected point is removed from the table to prevent loops; when there are no child nodes or the washing capacity of the buffer is insufficient, backtracking is performed;

[0138] When there are still washing targets, a new round of path search is started from the washing target closest to the input port, and unless there are no available nodes, the nodes covered by the previously determined washing paths will not be preferentially selected by the current path, and when all washing targets are covered by the path set, the search ends.

Claims

1. A method for capacity-aware continuous microfluidic biochip cleaning optimization, characterized by: The method is used for a buffer capacity limited constraint scene, The method comprises the following steps: Step one, a fluid routing algorithm based on coordination mechanism is used to calculate the corresponding flow path for each fluid transport task, and path conflict between grouped fluid transport tasks is considered through pollution-aware flow path planning, so that subsequent cleaning requirements are effectively reduced; Step two, a path-driven fluid scheduling algorithm is used to obtain accurate execution intervals for fluid transport tasks and cleaning tasks, and the storage position of intermediate fluid is obtained through a pre-calculated channel usage file, so as to improve access efficiency and reduce the total time of biochemical reaction execution; Step three, a capacity-aware cleaning optimization algorithm is used to obtain a cleaning path set to cover all cleaning targets, and the actual buffer capacity constraint and the resource conflict constraint of the cached liquid are considered, so that cleaning failure is avoided, and biochemical reactions can be efficiently and correctly executed.

2. The method of claim 1, wherein: Step one is a flow channel routing stage, specifically: a routing algorithm based on a negotiation mechanism is used to calculate a complete flow path for each fluid transport task; the fluid transport task is first grouped according to its execution interval, tasks with similar intervals are grouped into the same group, then the tasks in the same group are routed according to the constraint that the tasks in the same group can cross but not share the flow channel, and finally the tasks that fail to be routed are routed in turn according to the constraint that the tasks in different groups can share the flow channel; Step two is a fluid scheduling stage, specifically: a path-driven fluid scheduling algorithm is used to schedule each grouped fluid transport task in turn; there is no problem of cross contamination between tasks in the same group and no sharing of flow channels; after a group of tasks is completed, a cleaning path set is calculated and the corresponding execution interval is determined according to the defined cleaning target; when intermediate fluid needs to be cached, the flow channel with the least conflict with other paths is selected on the pre-planned flow path to perform the caching task; Step three is a cleaning optimization stage, specifically: a capacity-aware cleaning optimization method is used to calculate a group of cleaning path sets and corresponding execution intervals; first, the cleaning target set is determined based on the pre-calculated channel usage protocol, then a group of cleaning path sets is obtained by using a depth-first search algorithm under the given buffer capacity constraint and the channel resource constraint of the cached liquid, and finally the execution interval of each cleaning path is obtained under the constraint of channel resource conflict.

3. The method of claim 2, wherein: The continuous microfluidic biochip comprises a double-layer architecture composed of two elastic layers, one elastic layer is a flow layer, and the other elastic layer is a control layer, channels are arranged in both elastic layers to form a distributed channel storage architecture that can execute fluid transport tasks; The channels in the flow layer are flow channels for transporting reaction samples / reagents; the channels in the control layer are control channels for conducting air pressure; the channels are made of elastic material, and the intersection positions of the channels of the two elastic layers act as valves, which can be regarded as valves; the channels of the two elastic layers are connected to an external pressure source; In the control layer, external pressure is introduced by the control port, which presses the membrane down to block the channel and prevent the liquid from transporting. When the pressure is removed, the membrane restores to the original position by its elasticity. In the flow layer, the liquid is pushed by the pressure introduced by the flow port.

4. The method of claim 3, wherein: In the distributed channel storage architecture, the intermediate fluid can be cached in any idle flow channel, so that the flow channel has the dual role of transportation and storage. The fluid transportation tasks performed by the flow channel include the following: Task A1, fluid transportation one: transport the intermediate fluid from one component to another. The corresponding transportation path in the chip is: input port-> source component-> target component-> waste port; Task A2, fluid transportation two: transport the input sample / reagent to the target component. The corresponding transportation path in the chip is: input port-> target component-> waste port; Task A3, fluid storage: transport the intermediate fluid to the flow channel cache. The corresponding transportation path in the chip is: input port-> source component-> cache channel-> waste port; Task A4, fluid extraction: transport the intermediate fluid from the cache channel to the target component. The corresponding transportation path in the chip is: input port-> cache channel-> target component-> waste port.

5. The method of claim 4, wherein: In step two, the distributed channel storage architecture cleans the residual liquid slag on the flow channel through a cleaning operation to avoid cross contamination between fluids. In the cleaning operation, the buffer liquid injected from the external port flows through the contaminated flow channel / component for cleaning. The upper limit of the contaminant that can be cleaned by the buffer liquid is related to the volume of the injected buffer liquid; In the cleaning operation, when the buffer liquid flows through the contaminated flow channel / component that is not the cleaning target due to the limitation of the flow channel, the available capacity of the buffer liquid will be consumed, which will increase the number of cleaning times; When the maximum allowable consumption value of the contaminant that can be cleaned by a unit volume of buffer liquid is set, let ψ represent the cleaning capacity value of a unit volume of buffer liquid; Assuming that μ is the consumption value of the cleaning target, and v is the consumption value of the contaminated flow channel / component that flows through the non-cleaning target. Therefore, the actual cleaning capacity constraint of the buffer liquid is defined as: μ + v ≤ ψ Formula One; The shared flow channel in step two needs buffer liquid to perform the cleaning operation. The cleaning needs between different flow paths need to be systematically considered to improve the execution efficiency of the transportation task and accelerate the completion of the biochemical reaction. The grouping method of the fluid transportation task is as follows: Let each transportation task tk i ∈Tbe first associated with an execution interval (st i ,et i ) to indicate that the ith transportation task is scheduled to be executed from time point st i to et i ; then, these transportation tasks are sorted in non-decreasing order according to their starting execution time and are traversed in turn; and a dynamic task group G t is used to record those transportation tasks whose execution intervals are close to each other; when a task tk i is traversed, it can be inserted into G t if and only if G t is empty or the following inequality holds: st i -et i-1 <<δ(i>1) Equation Two; where st i is the start time of tk i , et i-1 is the end time of tk i-1 and δ is a user-defined threshold; otherwise, G t will be saved and the iteration will continue with a new dynamic task group, tk i being the first element of the new dynamic task group; finally, when the iteration is completed, a plurality of task groups are formed, each of which contains a plurality of transport tasks that need to be executed without pollution, and the transport tasks in the same group should avoid transport conflicts to reduce cleaning tasks; while the cleaning tasks between different groups can be executed in parallel with the biochemical reaction operation without conflict, thereby improving the execution efficiency of the biochemical reaction.

6. The method of claim 5, wherein: In step three, the capacity-aware cleaning optimization algorithm adopts a fluid routing method based on negotiation mechanism to obtain the cleaning path set. In this method, the routing graph is divided into a connection graph generated by a connection grid; each grid point is represented as a flow / waste liquid port, an inlet / outlet port of a component or a cross point; each edge connecting two grid points is represented as a channel segment that can perform fluid buffering and transportation; the connection graph formed constructs the required flow path for all transportation tasks in the schedule; each routing grid is associated with a historical cost at the rth iteration And the calculation method of the update function of the historical cost is: F(n i ) = G(n i ) + H(n i ) + C(n i ) Equation Four; Where G(n) i ) represents the distance from the source to n i The path length, H(n) i ) indicates from n i The estimated path length to the target point, C(n) i ) indicates the additional wiring cost. This represents the grid point n at the r-th iteration. i The historical cost, U r (n i ) is a binary (0 / 1) variable value to indicate whether n i There are flow paths constructed by the same group of transportation tasks, and C c Use user-defined parameter values ​​to represent the cross cost of a used grid point; Finally, the tasks of wiring failure are wired according to the constraint that different groups of tasks can share the flow channel in turn; at this time, the modified A* routing algorithm is also used to plan the flow path, and the wiring value F(n i ) of the current search grid point n i ) can be calculated by formula four, and the calculation method of C(n i ) is different, which is expressed in formula: wherein C w is a cleaning cost, C c is a crossing cost, C s is a user-defined positive integer; task(n i ) is a transport task currently occupying grid point n i , and tk j is the transport task currently being flow path planned. After the fluid routing is completed, a set of flow path collection P corresponding to all transport tasks and a flow channel network in a single chip architecture are obtained; a fluid scheduling scheme is executed according to the two results, so that all fluid transport tasks can be effectively executed without conflict on the generated chip architecture; all transport tasks will be scheduled in turn according to the grouping, and there is no cross contamination between the flow path scheduling of the same group, effectively reducing the number of time-consuming cleaning operations; after the fluid transport tasks of the same group are completed, the cleaning target is determined according to the channel usage protocol, and a set of cleaning paths is obtained to cover all cleaning targets, so that the transport tasks of different groups can quickly reuse the flow channels / components previously passed by the flow, thereby speeding up the biochemical reaction execution process.

7. The method of claim 6, wherein: The cleaning optimization method further comprises a path preparation and intermediate fluid caching strategy, specifically: for the fluid transportation task group g i , the flow path corresponding to the jth task , it is necessary to determine at which time it can be scheduled, First, if the target component c b is being occupied by other liquid flu tran and the liquid is not the input liquid of the binding operation, the liquid needs to be transported into the flow channel for storage first; Second, assume that flu tran The bound flow path is t_path, along which a free flow channel segment is searched to buffer the liquid, and the searching is performed without conflict with other transport tasks as much as possible; therefore, each flow channel d k，l is associated with an evaluation value ev k，l to determine the superiority of a flow channel selected as a buffer location; then, a flow channel can be selected as a buffer channel according to the following evaluation criteria: 1) In order to avoid changing the flow path previously calculated by other fluid transport tasks, the buffer position of the liquid can be selected on the corresponding flow path, 2) To avoid potential conflicts of fluid transportation and buffering, for flu tran each flow channel d on the corresponding flow path k，l If the time of the next fluid transportation task using d k，l is earlier, the value of ev k，l is larger; therefore, the flow channel with the minimum value of ev k，l on this flow path will be selected as the buffering location; When the cache location and corresponding cache path are determined, the intermediate liquid is temporarily cached within the flow channel, component c b may be used to receive the transported liquid; Finally, it is necessary to ensure that each flow channel segment is available, if a flow channel segment is occupied by the buffer liquid, the corresponding sub-path needs to be reconstructed to bypass the buffer channel; if the sub-path cannot be reconstructed, a new flow channel needs to be introduced to form a legal flow path; In the cleaning optimization method, the scheduling method for fluid extraction and transport is: For a fluid transport task group g i The jth fluid transport task in g If its bound liquid Is being cached in the flow channel, the liquid can be marked as flu_cache, and needs to be extracted from the cache location cs c To the target component c b ; The corresponding extraction path and execution interval also need to be calculated, and three sub-paths, f_port→cs c , cs c →c b , and c b →w_port are obtained by Dijkstra routing algorithm respectively; When the extraction path f_path has been calculated, the corresponding start time tdtart(f_parth) depends on the following three conditions: 1) the target component c b has been prepared, 2) the operation to be bound has been prepared, and 3) the extraction path f_path has also been prepared; thus, t start (f_parth) is calculated as follows: where max( ) denotes a maximum function, t ready (c b ) is the ready time of the target component, is the ready time of the task bound operation, and t ready (f_path) is the ready time of the extraction path; If the bound liquid is still on the source component, and its bound flow path is already prepared, then the liquid can be directly shipped from the source component to the target component; depends on the preparation time, and depends on the shipping delay.

8. The method of claim 7, wherein: When the target component is contaminated or all legal flow paths constructed for fluid transport are unavailable due to contaminated flow channels, a cleaning operation is triggered to perform cleaning; the cleaning optimization method also includes a cleaning target detection method for finding a set of contaminated places that actually need to be cleaned, thereby reducing the time-consuming cleaning process, specifically: Let each flow channel d k,l be associated with a parameter u k,l representing the number of flow paths using the channel, and let u k,l be decremented by one when a pre-constructed flow path uses the flow channel; and let each flow channel d k,l be further associated with a Boolean value s k,l representing the current state of the flow channel, where "0" represents a clean state and "1" represents a contaminated state; and let each component c j be associated with a counter parameter u j and a Boolean value s j ; and let s k,l and s j be initially set to "0" and updated according to the following rules: Rule 1. After each fluid transport task is completed, each flow channel d through which the flow path has passed k,l and component c j s k,l and s j values are set to "1"; Rule 2, after each washing operation, each flow channel d through which the washing path passes k,l and the components c j of s k,l and s j are set to "0"; Rule 3: After each fluid transport task is completed, the flow path passes through each flow channel d. k,l and component c j u k，l with u j The value is decremented by one; when the value is 0, the corresponding stream channel / component will no longer be needed.

9. The method of claim 8, wherein: In the cleaning operation, when the buffer solution has a limited cleaning capacity per unit volume and cannot be moved arbitrarily on the chip without considering its capacity demand on the contaminated flow channel / component, and a single fixed volume of buffer solution cannot handle all cleaning targets, the cleaning capacity demand analysis of each contaminated flow channel / component uses the following method: Case 1, when the cleaning flow path d k，l its corresponding capacity requirement cd k，l is calculated as: cd k，l = len k，l x T w Equation Eight; Case two, when the cleaning assembly C b its corresponding capacity demand cd b is calculated as: cd b = vol b x T w Equation Nine; Case three, for a cleaning target, if the contaminated flow channel / component connected thereto is also another cleaning target, the two form a new cleaning target, and the capacity demand of the cleaning target is the cumulative value.

10. The method of claim 9, wherein: The cleaning optimization method also includes a capacity-aware cleaning path calculation method, specifically: Let Tg be a set of cleaning targets w ={tgw1, tgw2, ..., tgw n The corresponding cleaning capacity requirement is Cd = {cd1, cd2, ..., cd}. n A set of optimized paths is determined through cleaning path calculation to cover all cleaning targets and minimize the total cleaning time. First, a weighted undirected graph G t (Q, R) is constructed to represent the connection between the external ports and the cleaning targets, where Q is composed of external ports Pt and cleaning targets Tg w and the edge represents the connectivity between them; for each edge r i，j ∈ R, its weight value w i，j represents the flow cost of the buffer between the associated nodes, and can be calculated as: w i,j = dis i,j + cd i，j Equation Ten; where dis i，j is the flow path length from point q i to point q j , and cd i，j is the cleaning demand of the contamination flow channel / assembly contained in the flow path. The flow path between two points is obtained according to the Dijkstra algorithm, and the flow channel / component currently containing fluid will not be selected; secondly, on the generated undirected graph, a set of cleaning path collections needs to be found, each of which starts from the input port, flows through the cleaning target, and finally ends at the output port, and the total cleaning capacity demand is less than the maximum cleaning capacity of the buffer solution; therefore, a capacity-aware cleaning calculation method based on depth-first search is adopted, specifically: The search process starts from a cleaning target closest to the input port, and then traverses the entire graph until an output port is found; a node table is created to record all available nodes, and is updated after each search: an adjacent point can be selected if its corresponding edge has the smallest weight value, and the previously selected point is removed from the table to prevent loops; when there are no child nodes or the cleaning capacity of the buffer solution is insufficient, backtracking is performed; When there are still cleaning targets, a new round of path search is started with the cleaning target closest to the input port, and unless there is no available node, the nodes covered by the previously determined cleaning path will not be preferred by the current path selection, and when all cleaning targets are covered by the path set, the search ends.

Images