Control method and device of neutral atomic quantum processing platform and storage medium

Abstract

The application discloses a control method and device of a neutral atom quantum processing platform and a storage medium, and relates to the field of quantum information technology. The control method comprises the following steps: encapsulating bottom-layer hardware channels, control registers and physical variables of the neutral atom quantum processing platform into physical semantic objects to obtain a physical semantic device registry; determining whether a received user instruction is classified as a command; if yes, generating an experiment planning result according to the user instruction and the physical semantic device registry; converting the experiment planning result into an experiment task graph; serializing the experiment task graph into an experiment task intermediate representation; converting the experiment task intermediate representation into executable timing and parameters of the neutral atom quantum processing platform; and generating a physical state index according to an experiment result obtained by executing the executable timing and parameters of the neutral atom quantum processing platform. The control method can greatly improve the working efficiency and stability of the neutral atom quantum processing platform, and avoid the risk of misoperation of bottom-layer hardware of a large model.

Description

Technical Field

[0001] This application relates to the field of quantum information technology, and more specifically, to a control method, apparatus, and storage medium for a neutral atom quantum processing platform. Background Technology

[0002] Neutral atom quantum information processing platforms, using electrically neutral atoms such as rubidium, cesium, strontium, and ytterbium as qubits or quantum simulation information carriers, are important candidate platforms for quantum computing, quantum simulation, quantum network nodes, and quantum precision measurement. A typical workflow for such platforms includes: preparing cold atom sources using techniques such as magneto-optical traps (MOTs), compressed magneto-optical traps (CMOTs), and polarization gradient cooling (PGC); loading atoms onto a programmable optical tweezers array generated by a spatial light modulator (SLM) or an acousto-optic deflector (AOD); obtaining single-atom occupancy maps through high-resolution fluorescence imaging and forming a defect-free target array using real-time feedback and atomic rearrangement; completing initial state preparation and single-qubit quantum gate manipulation through microwave, Raman, or optical pumping; introducing strong and controllable long-range dipole-dipole interactions through Rydberg excitation to realize a two-qubit quantum gate; and finally obtaining quantum state measurement results through imaging readout. The entire process requires long-term stability and multi-module coordination of a large number of hardware parameters, such as laser cooling optical detuning, laser power, magnetic field gradient, triaxial bias magnetic field, Rydberg excitation parameters, quantum gate pulse sequence, camera trigger timing, AOM / EOM / DDS / DAC / AWG, etc.

[0003] Existing technologies can control the aforementioned platforms using large model-assisted quantum experimental control systems. For example, systems like the HeuristicAutonomousLab (HAL) are representative. These systems utilize a large language model (LLM), a knowledge base, a planner, a code developer, and a runtime environment. The large model generates Python code to directly control experimental instruments. The knowledge base stores API documentation, standard protocols, and code examples; the runtime provides a state blackboard and code reuse mechanisms; and the signal path returns execution results to the planner as free text signals. This has been demonstrated in experiments such as superconducting resonant cavity characterization and quantum non-destructive measurements.

[0004] However, due to the lack of physical semantics in the underlying API, existing control interfaces mostly use device semantics to represent parameters such as DDS channel number, DAC voltage value, TTL number, and AOM RF power. Large models and new R&D personnel find it difficult to understand the physical correspondence between these hardware quantities and cooling optical detuning, laser light intensity, magnetic field gradient, optical tweezers trap depth, and Rydberg Rabi frequency. This leads to problems such as missing physical meaning, incorrect units, dimensional confusion, and safety overruns in the control commands generated by the model.

[0005] Furthermore, large models directly controlling the underlying hardware pose serious security risks. For example, large models may generate out-of-bounds parameters, excessively rapid rates of change, hardware interlock conflicts, incorrect camera exposure settings, coil overheating risks, misadjustment of slow variables, or the generation of unauditable scripts, thereby damaging instruments or putting the experimental platform into a dangerous state.

[0006] The content of the background section is merely technology known to the public and does not necessarily represent existing technology in the field. Summary of the Invention

[0007] This application aims to provide a control method, apparatus, and storage medium for a neutral atom quantum processing platform to solve at least one of the aforementioned technical problems.

[0008] According to one aspect of this application, a control method for a neutral atom quantum processing platform is provided. The control method includes: encapsulating the underlying hardware channels, control registers, and physical variables of the neutral atom quantum processing platform into physical semantic objects to obtain a physical semantic device registry; determining whether the received user instruction is classified as a command; if so, generating an experimental planning result based on the user instruction and the physical semantic device registry; converting the experimental planning result into an experimental task graph; serializing the experimental task graph into an intermediate representation of the experimental task; converting the intermediate representation of the experimental task into an executable timing sequence and parameters for the neutral atom quantum processing platform; and generating physical state indicators based on the experimental results obtained by executing the executable timing sequence and parameters of the neutral atom quantum processing platform.

[0009] According to some embodiments of this application, after encapsulating the underlying hardware channels, control registers, and physical variables of the neutral atom quantum processing platform into physical semantic objects to obtain a physical semantic device registry, the control method further includes: generating a physical parameter graph based on the underlying hardware channels, control registers, physical variables, and dependencies; generating experimental planning results based on user instructions and the physical semantic device registry includes: querying the physical parameter graph to obtain the adjustment state of the physical variables; and generating experimental planning results based on the adjustment state, user instructions, and the physical semantic device registry.

[0010] According to some embodiments of this application, after converting the intermediate representation of the experimental task into timing and parameters executable by the neutral atom quantum processing platform, the control method further includes: monitoring the execution of the neutral atom quantum processing platform according to the physical parameter diagram when the neutral atom quantum processing platform executes the executable timing and parameters, so as to obtain the monitoring results.

[0011] According to some embodiments of this application, when the neutral atom quantum processing platform executes executable timing and parameters, the execution status of the neutral atom quantum processing platform is monitored according to the physical parameter diagram to obtain monitoring results. This includes: when the neutral atom quantum processing platform executes executable timing and parameters, the execution status of the neutral atom quantum processing platform is monitored according to the physical parameter diagram based on a preset risk level to obtain monitoring results.

[0012] According to some embodiments of this application, after generating physical state indices based on experimental results obtained from executable timing and parameters executed by the neutral atom quantum processing platform, the control method further includes: generating structured physical state signals based on the experimental results and physical state indices; and determining the next set of experimental parameters based on the structured physical state signals, physical parameter diagrams, and experimental planning results.

[0013] According to some embodiments of this application, after generating a structured physical state signal based on experimental results and physical state indicators, the control method further includes: generating fault diagnosis and maintenance suggestions based on the physical state indicators, the structured physical state signal, and the physical parameter diagram.

[0014] According to some embodiments of this application, after generating fault diagnosis and maintenance suggestions based on physical state indicators, structured physical state signals, and physical parameter diagrams, the control method further includes: generating an experiment summary based on experimental results, physical state indicators, structured physical state signals, physical parameter diagrams, experimental planning results, fault diagnosis, and maintenance suggestions; saving the experiment summary, experiment task diagram, and executable timing and parameters as a reusable protocol if the user instruction is successfully executed; saving the experiment summary as a failure case if the user instruction fails to execute; and updating the knowledge base based on the reusable protocol and failure cases.

[0015] According to some embodiments of this application, after generating fault diagnosis and maintenance suggestions based on physical state indicators, structured physical state signals, and physical parameter diagrams, the control method further includes: generating experimental records and reports based on experimental results, physical state indicators, structured physical state signals, physical parameter diagrams, experimental planning results, fault diagnosis, and maintenance suggestions.

[0016] According to one aspect of this application, this application provides a control system for a neutral atom quantum processing platform. The control system includes: a physical semantic device registry module, a natural language target parsing module, a large-scale model experiment planning module, an experiment task graph generation module, an experiment task intermediate representation module, an experiment compiler module, and a result readout and analysis module.

[0017] The Physical Semantic Device Registry module encapsulates the underlying hardware channels, control registers, and physical variables of the neutral atom quantum processing platform into physical semantic objects to obtain the Physical Semantic Device Registry. The Natural Language Target Parsing module determines whether the received user instructions are classified as command-type. If so, the Large Model Experiment Planning module generates experiment planning results based on the user instructions and the Physical Semantic Device Registry. The Experiment Task Graph Generation module converts the experiment planning results into an experiment task graph. The Experiment Task Intermediate Representation module serializes the experiment task graph into an experiment task intermediate representation. The Experiment Compiler module converts the experiment task intermediate representation into a timing sequence and parameters executable by the neutral atom quantum processing platform. The Result Readout and Analysis module generates physical state indicators based on the experimental results obtained from the executable timing sequence and parameters of the neutral atom quantum processing platform.

[0018] According to another aspect of this application, this application also provides a non-volatile computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, can implement the control method as described above.

[0019] The technical solution of this application introduces a physical semantic device registry as a structured intermediate layer between the large model and the underlying hardware, which can significantly reduce the risk of the large model misoperating the underlying hardware. Since the large model can only interact with the hardware through physical semantic objects with physical definitions, units, safety ranges, and risk levels, and cannot directly access raw hardware parameters such as DDS channel numbers and DAC voltage values, the possibility of the large model generating unit errors, dimensional confusion, and out-of-bounds parameters is fundamentally eliminated by architectural isolation. The mandatory constraints on the lower and upper limits of parameters by the safety range field in the physical semantic object ensure that even if the plan generated by the large model contains physically unreasonable parameter values, it will be intercepted during the legality verification stage of the physical semantic object and will not be passed to the underlying hardware for execution.

[0020] The technical solution of this application realizes a complete transformation link from user instructions to executable timing and parameter control flow, eliminating the need for experimental personnel to manually write hardware control scripts, and significantly reducing the operating threshold of the neutral atom experimental platform and the repetitive workload of R&D personnel.

[0021] The technical solution of this application generates physical state indicators by reading experimental results, providing quantitative input for subsequent closed-loop optimization and fault diagnosis. These physical state indicators are not raw images or free text, but rather structured numerical values ​​with physical meaning, quality indicators, and confidence levels. They can be directly used by the optimizer and large model planning module, avoiding misjudgment problems caused by unstructured feedback data. Attached Figure Description

[0022] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0023] Figure 1 A flowchart illustrating a control method 1000 according to an embodiment of this application is shown; Figure 2 A schematic diagram of the structure of a control device according to an embodiment of this application is shown.

[0024] Explanation of reference numerals in the attached figures: 1. Control device; 11. Physical semantic device registry module; 12. Natural language target parsing module; 13. Large model experiment planning module; 14. Experiment task graph generation module; 15. Experiment task intermediate representation module; 16. Experiment compiler module; 17. Result reading and analysis module; 18. Physical parameter graph module; 19. Safety supervision and transaction execution module; 20. Structured physical state signal module; 21. Parameter optimization module; 22. Fault diagnosis and maintenance suggestion module; 23. Knowledge base automatic update module; 24. Electronic experiment record and report generation module. Detailed Implementation

[0025] Exemplary embodiments will now be described more fully with reference to the accompanying drawings. However, these exemplary embodiments can be implemented in many forms and should not be construed as limited to the embodiments set forth herein; rather, they are provided so that this application will be thorough and complete, and will fully convey the concept of the exemplary embodiments to those skilled in the art. The same reference numerals in the drawings denote the same or similar parts, and therefore repeated descriptions of them will be omitted.

[0026] The described features, structures, or characteristics can be combined in any suitable manner in one or more embodiments. Numerous specific details are provided in the following description to give a full understanding of embodiments of this disclosure. However, those skilled in the art will recognize that the technical solutions of this disclosure can be practiced without one or more of these specific details, or other methods, components, materials, devices, etc. In these cases, well-known structures, methods, devices, implementations, materials, or operations will not be shown or described in detail.

[0027] Furthermore, the terms “comprising” and “having”, and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or apparatus that includes a series of steps or units is not limited to the steps or units listed, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to such process, method, product, or apparatus.

[0028] The terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish different objects, not to describe a specific order.

[0029] The technical solutions of this application will be clearly and completely described below with reference to the accompanying drawings of the embodiments. Obviously, the described embodiments are only some, not all, of the embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0030] The English terms used in this application, their full English names, and their corresponding Chinese definitions are as follows: MOT, Magneto-Optical Trap, is a device that uses laser cooling and magnetic field gradients to trap atoms. 2DMOT, two-dimensional magneto-optical trap; 3DMOT, or three-dimensional magneto-optical trap, typically uses three pairs of counter-propagating cooling beams and anti-Helmholtz coils to form a three-dimensional trap. CMOT, Compressed Magneto-Optical Trap, is used to increase density or optimize subsequent loading. PGC, Polarization Gradient Cooling, is commonly used to reduce atomic temperature. AOM, Acousto-Optic Modulator, is used to control optical frequency, light intensity, or switching. EOM, Electro-Optic Modulator, is used for phase, frequency, or intensity modulation. DDS, Direct Digital Synthesizer, is used to generate high-precision RF signals; DAC, Digital-to-Analog Converter; ADC, Analog-to-Digital Converter; AWG, Arbitrary Waveform Generator; TTL, Transistor-Transistor Logic; SLM, SpatialLightModulator, is used to generate programmable optical tweezers arrays; AOD, Acousto-Optic Deflector, is used for rapidly moving or generating optical tweezers; DMD, Digital Micromirror Device, is an array of a large number of micromirrors that reflects laser light by rapidly flipping each micromirror to generate an array of optical tweezers of arbitrary shapes. Rydberg atom is an atom in a high principal quantum number excited state that has strong long-range interactions. Rabifrequency is the frequency of coherent oscillation in a quantum two-level system under a driving field. DSL, Domain-Specific Language; YAML, YAMLAin'tMarkupLanguage, is an abbreviation for recursion. JSON, JavaScriptObjectNotation, is a JavaScript object notation. FPGA stands for Field-Programmable Gate Array. API stands for Application Programming Interface. The molasses temperature (cloning temperature) refers to the thermodynamic temperature reached by atoms during the cooling phase of optical molasses. LLM, Large Language Model; VLM, Vision-Language Model, is a visual language model or multimodal model.

[0031] See Figure 2The control device 1 of the neutral atom quantum processing platform provided in this application includes: a physical semantic device registry module 11, a natural language target parsing module 12, a large model experiment planning module 13, an experiment task graph generation module 14, an experiment task intermediate representation module 15, an experiment compiler module 16, a result reading and analysis module 17, a physical parameter graph module 18, a security monitoring and transaction execution module 19, a structured physical state signal module 20, a parameter optimization module 21, a fault diagnosis and maintenance suggestion module 22, a knowledge base automatic update module 23, and an electronic experiment record and report generation module 24.

[0032] According to one aspect of this application, a control method 1000 for a neutral atom quantum processing platform is provided. See also... Figure 1 The control method 1000 may include steps S1 to S7.

[0033] In step S1, the physical semantic device registry module 11 encapsulates the underlying hardware channels, control registers, and physical variables of the neutral atom quantum processing platform into physical semantic objects to obtain the physical semantic device registry.

[0034] According to example embodiments, a neutral atom quantum processing platform can be a quantum information processing platform using neutral atoms such as rubidium, cesium, strontium, and ytterbium as carriers. Examples include neutral atom quantum computing platforms based on magneto-optical traps, optical tweezers arrays, single-atom fluorescence imaging, absorption imaging, optical pumping, microwave or Raman manipulation, Rydberg excitation, quantum gates, and quantum readout; and quantum simulation, quantum network nodes, and quantum precision measurement platforms.

[0035] The neutral atom quantum processing platform may include electronic control units (such as FPGA, MCU, real-time controller, DDS, DAC, ADC, TTL, AWG, AOM / EOM driver, camera trigger, driver layer and application software layer API, etc.) and physical execution units.

[0036] The physical execution unit may include an atom preparation module, a programmable optical tweezers array module, a quantum state manipulation module, an interaction control module, a readout imaging module, and a telemetry and interlocking module.

[0037] The atom preparation module can generate cold atom sources, such as 2DMOT, 3DMOT, CMOT, optical cloning, polarization gradient cooling, rubidium / cesium sources, Zeeman reducers, or atom beam sources.

[0038] The programmable optical tweezers array module can generate static / dynamic optical tweezers arrays through AOD, SLM, DMD, AWG or scanning mirror, and perform loading, rearrangement, movement, merging and partitioning operations.

[0039] The quantum state manipulation module can perform optical pumping, microwave, radio frequency, Raman transitions, single-qubit gates, Ramsey interferometry, spin echo, state selection, and local addressing.

[0040] The interaction control module can perform RydbergAtom two-photon excitation, Rabifrequency calibration, Rydberg beam overlap, blocking radius / interaction intensity calibration, controlled Z-gate, and controlled phase gate.

[0041] The readout imaging module may include absorption imaging, fluorescence imaging, single-atom occupancy mapping, quantum readout, camera triggering, image quality control, and threshold classification.

[0042] The telemetry and interlocking module can monitor frequency locking error, optical power, RF readback value, coil current / temperature, vacuum, camera status, cooling water, interlocking signal and environmental parameters.

[0043] The underlying hardware channel can be the physical signal path connecting the electronic control unit and the physical execution unit. The control register can be a storage unit located in the underlying hardware (such as DDS chip, DAC chip, FPGA internal logic).

[0044] Physical variables can be parameters that have clear physical meaning in the experiment and can be measured or set, such as laser detuning (MHz), magnetic field gradient (G / cm), optical tweezers power (mW), number of atoms, Rabi frequency (kHz), etc.

[0045] Physical semantic objects (PSEs) are data structures with physical meaning formed by abstracting and encapsulating the underlying hardware channels, control registers, and the physical quantities they control. Each PSE contains the physical variable name, unit, physical definition, hardware path, safety range, recommended range, minimum configurable resolution, latency, allowable rate of change, risk level, dependent variables, calibration model from hardware quantity to physical quantity, actual readback source, known failure modes, most recent calibration time (calibration data version), and responsible module (or maintainer).

[0046] The physical semantic device registry can be a queryable and verifiable mapping table of physical semantic objects.

[0047] For example, the physical semantic device registry module 11 can traverse all hardware resources of the neutral atom quantum processing platform and generate a physical semantic object instance for each controllable physical quantity according to a predefined configuration template, thereby obtaining the physical semantic device registry.

[0048] In step S2, the natural language target parsing module 12 determines whether the received user instruction is classified as a command.

[0049] According to the example embodiment, the user inputs experimental goals or queries in natural language. The natural language goal parsing module 12 can classify user instructions into query, command, or approval categories. Query-type instructions (such as "What is the current number of MOT atoms?") only return the answer and do not change the experimental state. Approval-type instructions are used for manual confirmation of high-risk tasks, slow-variable tasks, or tasks that have been rolled back and are being recovered. Command-type instructions (such as "Optimize today's neutral atom quantum processor to stabilize the 3DMOT state before 100 ms optical tweezer loading, prioritizing stability and then increasing the loading rate.") can trigger a series of experimental tasks.

[0050] For example, the natural language target parsing module 12 receives user input and categorizes it by analyzing syntactic structure and semantics. For example, user commands containing verbs such as "optimize," "run," and "calibrate" are usually categorized as "commands."

[0051] If the user instruction is determined to be classified as a command, proceed to step S3.

[0052] In step S3, the large model experiment planning module 13 generates the experiment planning results based on the user instructions and the physical semantic device registry.

[0053] According to the example implementation, the experiment planning result can be a top-level design scheme for achieving the user's command objective. The experiment planning result defines which high-level tasks need to be executed, rather than specific hardware operations. The experiment planning result may include information such as task nodes, task dependencies, adjustable variables, objective functions, measurement metrics, preconditions, postconditions, stopping conditions, failure conditions, rollback strategies, and required permissions.

[0054] Task nodes may include health check tasks, calibration tasks, optimization tasks, data acquisition tasks, diagnostic tasks, maintenance tasks, and reporting tasks.

[0055] For example, the large model experiment planning module 13 can call a large model (such as an LLM or VLM), read user instructions and the physical semantic device registry, and output the experiment planning results. The large model's prompt is injected with the contents of the physical semantic device registry, guiding the large model to use only the physical variables defined in the physical semantic device registry during planning and to follow their unit and range constraints.

[0056] The large-scale model experiment planning module 13 interacts with the hardware only through physical semantic objects and does not directly access the underlying hardware channels. For example, the large model cannot directly call "DDS3.set_freq(118.2e6) (writes the frequency setting value of 118.2MHz to the 3rd DDS chip)" or "DAC7.set_voltage(1.31) (writes the voltage value of 1.31V to the 7th DAC channel)", but should call physical variables with units and safety ranges, such as "set the detuning of the MOT cooling light relative to atomic transitions to -16 MHz".

[0057] In step S4, the experimental task graph generation module 14 converts the experimental planning results into an experimental task graph.

[0058] According to the example embodiment, the experimental task graph can be a concretization of the experimental planning results into an executable task flow. For example, the experimental task graph can be represented as a directed acyclic graph or a finite state machine. At a higher level, nodes in the experimental task graph can represent task nodes, and edges in the experimental task graph can represent dependencies between task nodes.

[0059] Each node includes inputs, outputs, preconditions, postconditions, risk level, required hardware resources, estimated duration, and rollback strategy.

[0060] For example, the experimental task graph generation module 14 configures the input, output, preconditions, postconditions, risk level and other attributes of each task node, and creates directed edges according to the dependencies between task nodes, thereby obtaining the experimental task graph.

[0061] In step S5, the experimental task intermediate representation module 15 serializes the experimental task graph into an experimental task intermediate representation.

[0062] According to the example embodiment, the experimental task intermediate representation is a hardware-independent intermediate language that describes the experimental process. The experimental task intermediate can be a restricted, declarative, or imperative language, such as a DSL, YAML, JSON, or database task table format.

[0063] Experimental task intermediate representations can only contain whitelisted physical operation primitives, experimental sequence primitives, and analysis primitives. For example, experimental task intermediate representations may include: setting physical parameters, running experimental sequences, acquiring image triplets, analyzing absorption imaging, updating the optimizer, and rolling back to the baseline.

[0064] For example, the experimental task intermediate representation module 15 traverses the experimental task graph through a serialization module and converts it into an experimental task intermediate representation in a target format (such as DSL, YAML, JSON, or a database task table).

[0065] In step S6, the experimental compiler module 16 converts the intermediate representation of the experimental task into timing and parameters executable by the neutral atom quantum processing platform.

[0066] According to the example embodiment, the executable timing and parameters are the sequence of low-level control instructions ultimately issued to the hardware (such as DDS, DAC, FPGA) of the neutral atom quantum processing platform. The executable timing and parameters may include a detailed timetable of when and what actions each hardware channel performs.

[0067] For example, experimental compiler module 16 performs at least the following transformations: 1. Convert physical variables into hardware settings. For example, convert MOT.cooling.detuning_MHz into laser frequency locking settings, AOM center frequency, and DDS frequency; convert MOT.gradient_G_per_cm and biasfields into multi-channel coil currents.

[0068] 2. Convert experimental sequence primitives into hardware timing sequences. For example, hardware timing sequences include: MOT loading, turning off MOT light and magnetic field, turning on probe light, camera exposure, atomic image acquisition, probe image acquisition, dark field image acquisition, and restoring baseline configuration.

[0069] 3. Generate a synchronization schedule for TTL, DDS, DAC, AWG, camera triggering, and data acquisition.

[0070] 4. Generate sequence_hash, parameter_hash, and firmware_version for each task for data traceability.

[0071] 5. Check for timing resource conflicts, such as overlapping AOM switches, mismatch between camera exposure window and probe light pulse, and coil slope exceeding the limit.

[0072] In step S7, the result readout analysis module 17 generates physical state indices based on the experimental results obtained by executing the executable timing and parameters of the neutral atom quantum processing platform.

[0073] According to the example embodiment, the generated executable timing sequence and parameters can be input into the electronic control unit to drive the physical execution unit to perform experiments. The experimental results can be the raw experimental data obtained after the neutral atom quantum processing platform executes the executable timing sequence and parameters. For example, the experimental results may include experimental data of absorption imaging, experimental data of fluorescence imaging, experimental data of optical tweezers arrays, experimental data of Rydberg calibration, and experimental data of quantum gates (quantum circuits), etc.

[0074] Physical state indicators can be quantitative values ​​with clear physical meaning extracted from experimental results.

[0075] For example, experimental data for absorption imaging may include atomic images, probe images, dark images, camera exposure parameters, probe power, and probe detuning.

[0076] The result readout analysis module 17 can obtain the atomic density according to the following formula: OD(x,y) = -ln[(A(x,y)-D(x,y)) / (P(x,y)-D(x,y))]; Where x and y represent the horizontal and vertical coordinates, respectively. A(x, y) represents the pixel coordinates of the atomic image, P(x, y) represents the pixel coordinates of the probe image, and D(x, y) represents the pixel coordinates of the dark field image. OD(x, y) is the atomic density corresponding to the (x, y) pixel.

[0077] The results readout and analysis module 17 can calculate the number of atoms based on atomic density, pixel size, effective scattering cross section, probe light detuning, and saturation correction. It then uses a two-dimensional Gaussian or more general model to fit the cloud center, width, ellipticity, rotation angle, and fitting residuals. Physical state indicators include the number of atoms, cloud center, width, ellipticity, rotation angle, and fitting residuals.

[0078] For experimental data from fluorescence imaging, the data can include atomic images, background images, and regions of interest for single atoms. The physical state indices obtained from the results readout analysis module 17 include occupancy maps, loading probabilities, threshold confidence levels, and loss rates.

[0079] For experimental data on optical tweezers arrays, the data can include array positions, spatial light modulator / acousto-optic deflector parameters, and occupancy maps. The physical state indices analyzed by the results readout and analysis module 17 include optical tweezers uniformity, loading uniformity, rearrangement success rate, and defect rate.

[0080] For Rydberg calibration experimental data, the data can include excitation scan data, diatomic dynamics data, and readout results. The physical state indices analyzed by the results readout analysis module 17 include Rabi frequency, resonant detuning, blocking contrast, and interaction estimates.

[0081] For experimental data on quantum gates (quantum circuits), the experimental data can include quantum readout results, randomized benchmark tests, and scan data. The physical state indices analyzed by the result readout analysis module 17 include gate fidelity, state preparation and measurement errors, leakage errors, atom loss, contrast, and correlation.

[0082] Through the above embodiments, the technical solution of this application encapsulates the underlying hardware channels, control registers, and physical variables of the neutral atom quantum processing platform into physical semantic objects to obtain a physical semantic device registry. The technical solution of this application can generate experimental planning results through user instructions and the physical semantic device registry. The technical solution of this application can obtain an experimental task graph from the experimental planning results. The technical solution of this application can obtain an intermediate representation of the experimental tasks from the experimental task graph sequence. The technical solution of this application can obtain the executable timing and parameters of the neutral atom quantum processing platform from the intermediate representation of the experimental tasks. The technical solution of this application can generate physical state indicators from the experimental results.

[0083] The technical solution of this application introduces a physical semantic device registry as a structured intermediate layer between the large model and the underlying hardware, which can significantly reduce the risk of the large model misoperating the underlying hardware. Since the large model can only interact with the hardware through physical semantic objects with physical definitions, units, safety ranges, and risk levels, and cannot directly access raw hardware parameters such as DDS channel numbers and DAC voltage values, the possibility of the large model generating unit errors, dimensional confusion, and out-of-bounds parameters is fundamentally eliminated by architectural isolation. The mandatory constraints on the lower and upper limits of parameters by the safety range field in the physical semantic object ensure that even if the plan generated by the large model contains physically unreasonable parameter values, it will be intercepted during the legality verification stage of the physical semantic object and will not be passed to the underlying hardware for execution.

[0084] The technical solution of this application realizes a complete transformation link from user instructions to executable timing and parameter control flow, eliminating the need for experimental personnel to manually write hardware control scripts, and significantly reducing the operating threshold of the neutral atom experimental platform and the repetitive workload of R&D personnel.

[0085] The technical solution of this application generates physical state indicators by reading experimental results, providing quantitative input for subsequent closed-loop optimization and fault diagnosis. These physical state indicators are not raw images or free text, but rather structured numerical values ​​with physical meaning, quality indicators, and confidence levels. They can be directly used by the optimizer and large model planning module, avoiding misjudgment problems caused by unstructured feedback data.

[0086] Optionally, after step S1, the control method 1000 may further include step S11.

[0087] In step S11, the physical parameter graph module 18 generates a physical parameter graph based on the underlying hardware channels, control registers, physical variables, and dependencies.

[0088] According to the example embodiment, a physical parameter graph can be a graph structure representing causal, calibration, and dependency relationships between various physical quantities, hardware settings, and experimental states. The physical parameter graph is not pure knowledge base text, but a graph structure that can be queried by a program. Nodes in the physical parameter graph include hardware nodes, physical variable nodes, state variable nodes, experimental index nodes, and failure mode nodes; edges in the physical parameter graph represent causal, calibration, dependency, constraint, or observation relationships.

[0089] For example, in 3DMOT, the DDS frequency affects the AOM optical frequency shift, which in turn affects the cooling optical detuning. The cooling optical detuning affects the scattering force, trapping velocity, and temperature, ultimately influencing the MOT atom number, cloud size, and absorption imaging OD. As another example, the coil current, mapped by the calibration matrix to the magnetic field gradient and triaxial bias field, affects the cloud center, molasses temperature, and optical tweezers loading rate.

[0090] The physical parameter graph module 18 can traverse all physical semantic objects in the physical semantic device registry, register each physical variable as a physical variable node in the physical parameter graph, and add dependency edges according to its dependent variable fields. The physical parameter graph module 18 adds hardware nodes (corresponding to the various levels of hardware devices in the underlying hardware channel), state variable nodes (such as atomic cloud temperature, MOT atom count), experimental index nodes, and fault mode nodes to the physical parameter graph.

[0091] Step S3 may include steps S301-S302.

[0092] In step S301, the large model experiment planning module 13 queries the physical parameter diagram to obtain the adjustment status of the physical variables.

[0093] According to the example embodiment, the adjustment status can be the status information obtained by querying the physical parameter diagram, which is about whether a certain physical variable is currently adjustable, the range of adjustable options, and the constraints.

[0094] The adjustment status can include whether the current physical variable can be adjusted, whether it is within the safe range, whether it is within the recommended range, whether the upstream status it depends on is satisfied (such as whether the frequency locking status is locked), whether the deviation between the current set value and the last readback value is within the allowable range, the adjustment direction constraint of the variable (such as slow variables are only allowed to be adjusted slowly in one direction), the indirect influence range derived from the physical parameter diagram (such as adjusting the magnetic field gradient will simultaneously affect the atomic cloud size, molasses temperature and optical tweezer loading rate), response time and risk level, etc.

[0095] For example, the large model experiment planning module 13 queries the physical parameter diagram for each physical variable involved to confirm its adjustment status.

[0096] In step S302, the large model experiment planning module 13 generates the experiment planning results based on the adjustment status, user instructions, and physical semantic device registry.

[0097] According to the example embodiment, the large-scale model experiment planning module 13 can call the large model, read the adjustment status, user instructions, and physical semantic device registry, and output the experiment planning results. For example, the large model can exclude unadjustable physical variables and high-risk physical variables to prevent the range of physical variables from exceeding the safe range.

[0098] Through the above embodiments, the technical solution of this application can exclude unadjustable and high-risk variables before generating experimental planning results using physical parameter diagrams. The technical solution of this application can confirm whether the upstream dependencies of physical variables are ready (e.g., whether frequency locking is normal) before generating experimental planning results by adjusting the state, rather than discovering conflicts only during the execution phase, effectively avoiding experimental failures in a semi-execution state.

[0099] The technical solution of this application enables large models to understand the physical causal relationships between physical variables through physical parameter diagrams, and the generated experimental planning results have stronger physical rationality, reliability and interpretability.

[0100] Optionally, after step S1, the control method 1000 may further include step S61.

[0101] In step S61, the security monitoring and transaction execution module 19 monitors the execution status of the neutral atom quantum processing platform according to the physical parameter diagram, while the neutral atom quantum processing platform executes executable timing and parameters, so as to obtain the monitoring results.

[0102] According to the example embodiment, the monitoring results can be hardware status information collected and analyzed in real time under the execution of executable timing and parameters on the neutral atom quantum processing platform, which is used to determine whether the execution is proceeding normally.

[0103] The monitoring results may include: the deviation between the hardware readback values ​​and the set values ​​of each physical variable, hardware telemetry (laser frequency locking error signal, optical power monitoring, actual value of coil current, coil temperature, vacuum gauge reading, camera readiness status), abnormal event records during execution (such as interlock triggering, TTL timing error, DDS frequency jump), and current experimental status assessment based on physical parameter diagrams (such as real-time estimation of MOT atom number trend based on scattered fluorescence readings).

[0104] For example, before the neutral atom quantum processing platform executes the executable timing and parameters, the security monitoring and transaction execution module 19 generates the timing and parameters of candidate configurations based on the experimental task graph.

[0105] If the physical variables in the timing and parameters of the candidate configuration meet the variable safety conditions, the security monitoring and transaction execution module 19 performs simulation compilation on the timing and parameters of the candidate configuration to monitor whether the timing and parameters of the candidate configuration meet the timing and parameter safety conditions.

[0106] If the timing and parameter safety conditions are met, the safety monitoring and transaction execution module 19 writes the candidate configuration into the electronic control unit of the neutral atom quantum processing platform.

[0107] During the execution of executable timing and parameters on the neutral atom quantum processing platform, the security monitoring and transaction execution module 19 reads the feedback results from the neutral atom quantum processing platform (such as the deviation between the hardware readback values ​​and the set values ​​of each physical variable, hardware telemetry, imaging quality, and objective function).

[0108] In the event of an anomaly in the feedback result, the security monitoring and transaction execution module 19 will restore the candidate configuration to the previous valid configuration or the current configuration.

[0109] After the neutral atom quantum processing platform executes the executable timing and parameters, the security monitoring and transaction execution module 19 determines the candidate configuration as the valid configuration if the candidate configuration meets the stability conditions.

[0110] Through the above embodiments, the technical solution of this application, based on real-time monitoring of physical parameter graphs, can detect hardware anomalies before and during the execution of executable timing and parameters, thus preventing hardware anomalies from propagating through the causal chain in the physical parameter graph within milliseconds to seconds, which could lead to completely invalid experimental results or even misleading conclusions.

[0111] Optionally, step S61 can specifically be as follows: when the neutral atom quantum processing platform executes executable timing and parameters, the security supervision and transaction execution module 19 monitors the execution status of the neutral atom quantum processing platform according to the preset risk level and the physical parameter diagram, so as to obtain the monitoring results.

[0112] According to the example embodiment, the preset risk level can be used to classify the degree of danger for each physical variable or hardware status in advance, and determine the priority, response strategy and access requirements of the physical variable or hardware status in monitoring.

[0113] Preset risk levels can include read-only level, low-risk write level, fast variable write level, slow variable write level, and high-risk level.

[0114] The physical variables or hardware states corresponding to the read-only level can include images, logs, frequency locking errors, vacuum, temperature, optical power readback, and camera status. The permission requirement for the read-only level is free to read, but not write.

[0115] The physical variables or hardware status corresponding to the Fast Variable Write level can include repetition count, camera region of interest, analysis threshold, chart generation, and reporting parameters. The permission requirements for the Fast Variable Write level are that it can be executed automatically and logged.

[0116] The physical variables or hardware states corresponding to the low-risk write level may include laser detuning, acousto-optic modulator power, magneto-optical trap magnetic field gradient, compensation magnetic field, and small-range compensation for optical tweezer depth. The permission requirements for the low-risk write level are that they can be executed automatically within the recommended range; beyond the recommended range but within a safe range, secondary confirmation or task policy authorization is required.

[0117] The physical variables or hardware states corresponding to the slow variable write level can include extensive modifications to rubidium sources, nuclear reactors, ultraviolet duty cycles, vacuum-related variables, and master lock frequencies. The permission requirement for the slow variable write level is that, by default, only suggestions are generated, requiring manual confirmation before execution.

[0118] The physical variables or hardware states corresponding to high-risk levels may include disabling hardware interlocks, raising protection thresholds, starting and stopping high-power lasers, and disabling vacuum protection. The permission requirements for high-risk levels are that direct execution is prohibited; only manual procedures can be prompted.

[0119] Through the above embodiments, the technical solution of this application can differentiate the allocation of monitoring resources according to preset risk levels, ensuring that high-risk physical variables are strictly protected in real time while avoiding excessive monitoring of low-risk physical variables that would cause excessive load.

[0120] Optionally, after step S7, the control method 1000 may further include steps S8-S9.

[0121] In step S8, the structured physical state signal module 20 generates a structured physical state signal based on the experimental results and physical state indicators.

[0122] According to the example implementation, a structured physical state signal can be a state representation that compresses experimental results and physical state indices into a standardized format. The structured physical state signal can include numerical indices, error bars, quality indicators, anomaly indicators, telemetry states, diagnostic hypotheses, confidence levels, and suggested actions. The structured physical state signal can be output using a dual-channel approach combining a structured signal and a natural language summary. The structured signal (e.g., JSON fields) can be used for parameter optimization and constraint handling, while the natural language summary is used for large model planning and reporting.

[0123] For example, the structured physical state signal module 20 can generate standardized structured physical state signals based on experimental results (e.g., original images, original data curves, floating-point arrays, and telemetry data in the original experimental data) and physical state indicators (e.g., fitting residuals, threshold confidence, etc.).

[0124] In step S9, the parameter optimization module 21 determines the next set of experimental parameters based on the structured physical state signal, the physical parameter diagram, and the experimental planning results.

[0125] According to the example embodiment, the next set of experimental parameters can be the executable timing and parameters proposed by the optimization module for the next round of experiments after the current round of experiments is completed, based on historical results and the current state.

[0126] For example, the parameter optimization module 21 can employ a preset optimization algorithm (such as grid scanning, random scanning, Latin hypercube sampling, Bayesian optimization, covariance matrix adaptive evolution strategy, genetic algorithm, multi-objective optimization, robust optimization algorithm, or physical model-based optimization algorithm) to determine the next set of experimental parameters based on the structured physical state signal, physical parameter map, and experimental planning results. The parameter optimization module 21 queries the physical parameter map to exclude physical variables irrelevant to the current task or high-risk physical variables.

[0127] The optimization strategies of the preset optimization algorithm may include noise handling, constraint handling, multi-objective handling, hierarchical optimization, variable grouping, and rollback verification.

[0128] Noise handling may include repeatedly sampling each parameter point and updating it using the median or robust mean. Constraint handling may include applying hard constraints to atomic density saturation, camera saturation, frequency locking anomalies, vacuum anomalies, and coil temperature anomalies.

[0129] Multi-objective processing can include simultaneous optimization of atomic number, stability, temperature, loading rate, defect rate, readout fidelity, etc. Hierarchical optimization can include first optimizing imaging and MOT, then optimizing molasses or CMOT, and then optimizing optical tweezers loading, Rydberg calibration, and quantum gates.

[0130] Variable grouping can include automatic optimization of fast variables, and generating suggestions or awaiting manual approval for slow variables. Rollback verification can include repeating the old optimal point and baseline configuration after each round of optimization to prevent false optimalities caused by slow drift.

[0131] Through the above embodiments, the technical solution of this application can solve the problems of unstructured experimental feedback and unreliable use of large models in the prior art by using structured physical state signals.

[0132] The technical solution of this application achieves closed-loop optimization with dynamic constraint awareness by querying the physical parameter graph when generating the next set of parameters. If the upstream dependency state of a certain physical variable changes during the optimization process (such as the frequency locking state changing from locked to unlocked), the physical variable is automatically excluded from the optimization range when generating the next set of parameters, instead of continuing to blindly explore the space containing the physical variable. This prevents the problem of meaningless results generated by continuous optimization iteration when the hardware state has changed, and improves the robustness and resource utilization efficiency of the optimization process.

[0133] Optionally, after step S8, the control method 1000 may further include step S81.

[0134] In step S81, the fault diagnosis and maintenance suggestion module 22 generates fault diagnosis and maintenance suggestions based on physical state indicators, structured physical state signals and physical parameter diagrams.

[0135] According to the example embodiment, fault diagnosis can be a process of analyzing the causes of experimental anomalies, outputting possible causes of the fault and their confidence levels. Fault diagnosis may include observed symptoms, a list of diagnostic evidence, a list of candidate causes, an automated verification experimental protocol, suggested actions, risk levels, expected scope of impact, and a responsible module.

[0136] Observable symptoms can be derived from anomalous signs and index deviations in structured physical state signals. The list of diagnostic evidence can come from supporting evidence derived from telemetry data and physical state indices. A list of candidate causes can be obtained by traversing the candidate cause nodes backward from the symptom nodes in the physical parameter graph; each candidate cause includes a cause description and an initial confidence level. An automated validation experiment protocol can generate a minimal, automatically executable validation experiment for each candidate cause.

[0137] Maintenance recommendations can be suggestions for addressing diagnosed or potential hardware or experimental issues, including automated calibration procedures and maintenance work orders requiring manual intervention.

[0138] For example, automated calibration operations may include automatically performing magnetic field zero-point scanning, updating the imaging reference map, and adjusting the compensation magnetic field. Maintenance work orders requiring manual intervention may include fields such as work order identifier, symptom description, diagnosis, required tools / consumables, operating procedures, risk warnings, responsible person, and estimated duration.

[0139] For example, an experimental anomaly might be "suddenly unable to observe atomic clouds or a significant decrease in loading rate." The candidate causes output by the fault diagnosis and maintenance suggestion module 22 might be: abnormal frequency locking of the cooling light (or repump light), no output of the acousto-optic modulator RF signal, shutter signal (or transistor-transistor logic signal), and no current in the coil. The automatically executable calibration operations output by the fault diagnosis and maintenance suggestion module 22 include: reading the frequency locking error, reading the RF readback value, reading the shutter status, reading the coil current, and taking a fluorescence snapshot. The corresponding output suggestions are: pause optimization, roll back timing, and generate a frequency-locked acousto-optic modulator checklist.

[0140] The experimental anomaly was described as "a slow decrease in the number of atoms, but the cloud shape remained normal." The fault diagnosis and maintenance suggestion module 22 outputs possible causes such as: decreased rubidium source flux, decreased two-dimensional magneto-optical trap flux, vacuum variations, and insufficient ultraviolet duty cycle. The automatically executable calibration operations output by the fault diagnosis and maintenance suggestion module 22 include: fitting the loading curve, reading the vacuum trend, and comparing the rubidium source current or temperature. The corresponding output suggestion is: check the rubidium distributor, ultraviolet light, and two-dimensional magneto-optical trap.

[0141] Through the above embodiments, the technical solution of this application can generate fault diagnosis and maintenance suggestions based on physical parameter diagrams, thereby improving the fault location capability from "relying on human experience and subjective judgment" to "structured reasoning based on physical causal relationships". It can quickly distinguish different candidate causes, shorten the fault location time from hours to minutes, and improve the diagnostic accuracy from relying on personal experience to a repeatable process based on physical evidence.

[0142] Optionally, after step S81, the control method 1000 may further include steps S82-S85.

[0143] In step S82, the knowledge base automatic update module 23 generates an experimental summary based on the experimental results, physical state indicators, structured physical state signals, physical parameter diagrams, experimental planning results, fault diagnosis, and maintenance suggestions.

[0144] According to the example implementation, an experiment summary can be a structured summary of the execution process of a complete experimental task, including the objectives, execution process, results, and conclusions.

[0145] Experiment summaries may include task identifier, task objective, execution time, parameter range, number of executions, optimal configuration parameter values, failure points (such as which candidate configurations triggered constraints or rollbacks), reasons for and number of bad data removals, conclusions (whether the objective was achieved, the extent of improvement, and limitations), records of human intervention, and final decision (success / failure / partial success).

[0146] For example, after the user-instructed task is completed (whether successfully or not), the knowledge base automatic update module 23 extracts key information from experimental results, physical state indicators, structured physical state signals, physical parameter diagrams, experimental planning results, fault diagnosis, and maintenance suggestions, and calls the large model to generate an experimental summary. The experimental summary contains both structured fields and natural language descriptions.

[0147] In step S83, if the user command is successfully executed, the knowledge base automatic update module 23 saves the experiment summary, experiment task diagram, and executable timing and parameters as a reusable protocol.

[0148] According to an example embodiment, a reusable protocol can be a standardized process record that can be reused in future tasks, which can solidify the complete process of a successful experimental task.

[0149] Reusable protocols may include: protocol identifier, protocol name, applicable conditions (e.g., "applicable to 87 rubidium platform, MOT optimization before optical tweezers loading, the task objective is to improve the stability of the number of atoms measured at the end of 100ms loading time"), experimental task diagram, intermediate representation of experimental task, compiled executable timing and parameters, optimal configuration parameter values ​​and parameter confidence range, and most recent verification time.

[0150] In step S84, the knowledge base automatic update module 23 saves the experiment summary as a failure case if the user command fails to be executed.

[0151] According to the example implementation, a failure case can be a structured record of the failure process of an experimental task.

[0152] Failure cases may include: case identifier, observed symptoms, list of diagnostic evidence, root cause (confirmed manually or by the system), verification process (which verification experiments were performed and what the results were), remedial actions (including operation steps, parameter modification values ​​and execution time), risk level, reusable diagnostic rule flag (if true, the diagnostic logic of the failure case will be added to the physical parameter diagram), and case remarks.

[0153] In step S85, the knowledge base automatic update module 23 updates the knowledge base according to the reusable protocol and failure cases.

[0154] According to the example implementation, the knowledge base can be an information system for storing and indexing various documents and records related to experiments, supporting enhanced query generation during large-scale model planning. For example, the knowledge base can be implemented through a vector database, graph database, or relational database, supporting retrieval by keywords, experiment type, parameter name, symptom description, and time range, and providing an interface for enhanced query generation to the large-scale causal model planning module.

[0155] For example, the knowledge base automatic update module 23 can update the physical semantic device registry, reusable protocols, failure cases, and other content to the knowledge base so that the latest protocols and failure cases can be retrieved when querying large models.

[0156] Through the above embodiments, the technical solution of this application can realize the continuous automatic accumulation and reuse of experimental knowledge by solidifying the experiments in which user instructions are successfully executed into reusable protocols and writing back the diagnostic logic of failed cases to the physical parameter diagram.

[0157] Optionally, after step S81, the control method 1000 may further include step S86.

[0158] In step S86, the electronic experiment record and report generation module 24 generates experiment records and reports based on the experiment results, physical state indicators, structured physical state signals, physical parameter diagrams, experiment planning results, fault diagnosis and maintenance suggestions.

[0159] According to the example embodiment, experimental records and reports can be documented outputs of experimental procedures and results. Experimental records and reports can include structured experimental records and readable reports.

[0160] Each experimental record may include task identifier, task objective, execution platform, approval status, hardware version, firmware version, timing version, parameter snapshot, experimental results, experimental result storage path, physical status indicators, structured physical status signals, physical parameter diagrams, experimental planning results, fault diagnosis, and maintenance suggestions.

[0161] For example, the electronic experiment record and report generation module 24 extracts data from experimental results, physical state indicators, structured physical state signals, physical parameter diagrams, experimental planning results, fault diagnosis, and maintenance suggestions, and generates structured records according to a predefined experiment record template. Simultaneously, the electronic experiment record and report generation module 24 calls a large model, using the structured records as input, to generate a readable report in natural language.

[0162] Through the above embodiments, the technical solution of this application can automatically generate experimental records and reports, thus meeting the requirements of scientific research for repeatability and auditability.

[0163] According to another aspect of this application, this application also provides a non-volatile computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, can implement the control method as described above.

[0164] According to another aspect of this application, this application also provides an electronic device, including: one or more processors; and a storage device for storing one or more programs, which, when executed by one or more processors, enable the one or more processors to implement the control method described above.

[0165] According to another aspect of this application, this application also provides a computer program product, including: a computer program stored on a computer-readable storage medium; the computer program includes program instructions that, when executed by a computer, cause the computer to perform the control method as described above.

[0166] Finally, it should be noted that the above description is merely a preferred embodiment of this application and is not intended to limit this application. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions of the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. A control method for a neutral atom quantum processing platform, characterized in that, The control method includes: The underlying hardware channels, control registers, and physical variables of the neutral atom quantum processing platform are encapsulated into physical semantic objects to obtain a physical semantic device registry. A physical parameter diagram is generated based on the underlying hardware channel, the control register, the physical variables, and the dependencies. Determine whether the received user instruction is classified as a command. If so, query the physical parameter diagram to obtain the adjustment status of the physical variables; Based on the adjustment status, the user instructions, and the physical semantic device registry, generate the experimental planning results; The experimental planning results are converted into an experimental task diagram; The experimental task diagram is serialized into an intermediate representation of the experimental task. The intermediate representation of the experimental task is converted into timing and parameters executable by the neutral atom quantum processing platform; Based on the experimental results obtained by executing the executable timing and parameters using the neutral atom quantum processing platform, physical state indices are generated. Based on the experimental results and the physical state indicators, a structured physical state signal is generated; Based on the structured physical state signal, the physical parameter diagram, and the experimental planning results, determine the next set of experimental parameters; The physical semantic device registry is a queryable and verifiable mapping table of the physical semantic objects. The process of encapsulating the underlying hardware channels, control registers, and physical variables of the neutral atom quantum processing platform into physical semantic objects to obtain a physical semantic device registry includes: Traverse all hardware resources of the neutral atom quantum processing platform and generate a physical semantic object instance for each controllable physical quantity according to a predefined configuration template, thereby obtaining the physical semantic device registry; The step of generating experimental planning results based on the adjustment state, the user instructions, and the physical semantic device registry includes: The large model is invoked to read the adjustment status, the user instructions, and the physical semantic device registry, and the experimental planning results are output. The large model prompts are injected with the contents of the physical semantic device registry, guiding the large model to use only the physical variables defined in the physical semantic device registry during planning, and to follow the unit and range constraints of the physical variables defined in the physical semantic device registry.

2. The control method according to claim 1, characterized in that, After converting the intermediate representation of the experimental task into timing and parameters executable by the neutral atom quantum processing platform, the control method further includes: When the neutral atom quantum processing platform executes the executable timing and parameters, the execution status of the neutral atom quantum processing platform is monitored according to the physical parameter diagram to obtain monitoring results.

3. The control method according to claim 2, characterized in that, When the executable timing and parameters are executed on the neutral atom quantum processing platform, the execution status of the neutral atom quantum processing platform is monitored according to the physical parameter diagram to obtain the monitoring results, including: When the neutral atom quantum processing platform executes the executable timing and parameters, the execution status of the neutral atom quantum processing platform is monitored according to the physical parameter diagram based on the preset risk level, so as to obtain the monitoring results.

4. The control method according to claim 1, characterized in that, After generating the structured physical state signal based on the experimental results and the physical state indicators, the control method further includes: Based on the physical state indicators, the structured physical state signals, and the physical parameter diagram, fault diagnosis and maintenance suggestions are generated.

5. The control method according to claim 4, characterized in that, After generating fault diagnosis and maintenance recommendations based on the physical state indicators, the structured physical state signals, and the physical parameter diagram, the control method further includes: Based on the experimental results, the physical state indicators, the structured physical state signals, the physical parameter diagram, the experimental planning results, the fault diagnosis, and the maintenance suggestions, an experimental summary is generated. If the user instruction is executed successfully, the experiment summary, the experiment task graph, and the executable timing and parameters are saved as a reusable protocol. If the user instruction fails to execute, the experiment summary will be saved as a failure case. Update the knowledge base based on the reusable protocol and the failure cases.

6. The control method according to claim 4, characterized in that, After generating fault diagnosis and maintenance recommendations based on the physical state indicators, the structured physical state signals, and the physical parameter diagram, the control method further includes: Based on the experimental results, physical state indicators, structured physical state signals, physical parameter diagrams, experimental planning results, fault diagnosis, and maintenance recommendations, experimental records and reports are generated.

7. A control system for a neutral atom quantum processing platform, characterized in that, The control system includes: The physical semantic device registry module encapsulates the underlying hardware channels, control registers, and physical variables of the neutral atom quantum processing platform into physical semantic objects to obtain the physical semantic device registry. The physical parameter graph module generates a physical parameter graph based on the underlying hardware channel, the control register, the physical variables, and the dependencies. The natural language target parsing module determines whether the received user instructions are classified as commands. If so, the large-scale model experiment planning module queries the physical parameter graph to obtain the adjustment status of the physical variables; the large-scale model experiment planning module generates the experiment planning result based on the adjustment status, the user instruction, and the physical semantic device registry. The experimental task graph generation module converts the experimental planning results into an experimental task graph. The experimental task intermediate representation module serializes the experimental task graph into an experimental task intermediate representation. The experimental compiler module converts the intermediate representation of the experimental task into timing and parameters executable by the neutral atom quantum processing platform; The result readout and analysis module generates physical state indices based on the experimental results obtained by executing the executable timing and parameters on the neutral atom quantum processing platform. The structured physical state signal module generates a structured physical state signal based on the experimental results and the physical state indicators. The parameter optimization module determines the next set of experimental parameters based on the structured physical state signal, the physical parameter diagram, and the experimental planning results. The physical semantic device registry is a queryable and verifiable mapping table of the physical semantic objects. The physical semantic device registry module traverses all hardware resources of the neutral atom quantum processing platform and generates a physical semantic object instance for each controllable physical quantity according to a predefined configuration template, thereby obtaining the physical semantic device registry. The large-scale model experiment planning module calls the large model, reads the adjustment status, the user instructions, and the physical semantic device registry, and outputs the experiment planning results. The large model prompts are injected with the contents of the physical semantic device registry, guiding the large model to use only the physical variables defined in the physical semantic device registry during planning, and to follow the unit and range constraints of the physical variables defined in the physical semantic device registry.

8. A non-volatile computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the control method as described in any one of claims 1-6.

Images