Adaptive agent-oriented system control for physical system design and operation

Abstract

A multi-agent adaptive agent-oriented design and control of a target physical system. A set of specialized agents translate a high-level engineering objective into a set of tasks representing discrete mathematical or physical constraints for execution by a subset of the specialized agents. The set of specialized agents communicate independently to resolve task dependencies and to produce a set of outputs for integration and generation of a system design for the target physical system. The set of specialized agents may implement a digital twin to simulate candidate system designs for generating a final system design or to simulate the final system design for controlling the target physical system in real time, wherein the real-time control may be informed by direct or indirect variable measurements from physical or virtual sensors. The set of specialized agents may utilize mechanistic models, generative artificial intelligence, or quantum-enhanced inference.

Description

Atty. Doc. No. SICS-101-B-WO PATENTADAPTIVE AGENT-ORIENTED SYSTEM CONTROL FOR PHYSICAL SYSTEM DESIGN AND OPERATIONCROSS-REFERENCE TO RELATED APPLICATION(S)

[0001] This application claims priority to and the benefit of U.S. Application Serial No.63 / 779,953, entitled “AGENT-ORIENTED SYSTEMS CONTROL,” filed March 28, 2025, the entire disclosure of which is hereby incorporated by reference.TECHNICAL FIELD

[0002] The present disclosure relates to multi-agent systems for engineering and industrial control, and more particularly relates to a framework of specialized agents configured for autonomous decomposition of engineering objectives into mechanistic tasks, independent resolution of task dependencies, and iterative integration of agent outputs for physical system design and operational optimization.BACKGROUND

[0003] Modem industrial control and engineering design systems are increasingly required to manage high-dimensional data and complex physical constraints in real-time. In traditional engineering workflows, specialized tasks — such as thermal analysis, structural modeling, or chemical kinetic simulations — are often performed in isolation using disparate software tools and methodologies. As engineering objectives become more sophisticated, the need to integrate these specialized outputs into a cohesive system design grows. However, conventional centralized control architectures often encounter significant scaling limitations and processing bottlenecks when attempting to coordinate numerous interdependent tasks across different physical domains.

[0004] A significant technical challenge in complex system engineering involves the resolution of task dependencies and the integration of conflicting mathematical or physical constraints. In many large-scale systems, an adjustment in one domain (e.g., increasing the flow rate in a chemical reactor) may have non-linear impacts on another domain (e.g., the structural integrity or thermal stability of the vessel). Conventional systems often lack thecapacity to autonomously decompose a global engineering objective into discrete, manageable tasks while simultaneously accounting for the intricate web of dependencies between those tasks. Without a robust mechanism for independent communication and data exchange between specialized modeling components, these systems often rely on manual, iterative interventions that are prone to human error and significant delays.

[0005] Furthermore, a persistent gap exists between digital simulations and the operational reality of the target physical systems. While digital twins and predictive models are used to anticipate system behavior, they are highly sensitive to the accuracy of input data from physical sensors. Physical sensors in industrial environments are susceptible to drift, hardware failure, and environmental noise, which can lead to the propagation of errors throughout a control system. Additionally, many critical operational parameters are indirect variables that cannot be measured directly by hardware, requiring the use of virtual sensors, also referred to as soft sensors. Existing frameworks often struggle to validate the accuracy of these virtual measurements against physical reality in real-time, leading to a loss of situational awareness within the control logic.

[0006] These integration failures are further compounded when a candidate system design violates a global constraint, such as a safety limit or a fundamental physical law. In such instances, conventional control systems may fail to identify the specific sub-objective causing the violation or may lack a structured method for performing trade-off calculations between competing performance metrics. Without an integrated logical framework to manage the iterative refinement of engineering objectives and to synchronize specialized modeling outputs with real-time sensor feedback, existing industrial control systems remain limited in their ability to autonomously generate and optimize complex system designs.SUMMARY

[0007] In one aspect, a method is provided for adaptive agent-oriented system control for physical system design and operation. The method includes: (1) decomposing, by an initiator agent of a set of specialized agents, an engineering objective for a target physical system into a set of tasks representing discrete mathematical or physical constraints; (2) identifying, by the initiator agent, a set of task agents of the set of specialized agents configured to perform the set of tasks; (3) instructing, by the initiator agent, the set of task agents to perform the set of tasks to produce a set of outputs, wherein the set of task agents is configured to communicate with one another independently of the initiator agent for resolving task dependencies; and (4) generating, by an executive agent of the set of specialized agents, asystem design for the target physical system based on an integration of the set of outputs.

[0008] In another aspect, a non-transitory computer-readable medium is provided that stores instructions operable to cause one or more processors to perform the operations described above with respect to the method.

[0009] In another aspect, a system, or apparatus, is provided that comprises one or more memories and one or more processors configured to execute instructions stored in the one or more memories to perform the operations described above with respect to the method.

[0010] These and other aspects, features, elements, implementations, and embodiments of the methods, apparatus, procedures, and algorithms disclosed herein are described in further detail hereafter.BRIEF DESCRIPTION OF THE DRAWINGS

[0011] The disclosed technology is best understood from the following detailed description when read in conjunction with the accompanying drawings. It is emphasized that, according to common practice, the various features of the drawings may not be to scale. For example, the dimensions of various features may be expanded or reduced for clarity. Further, like reference numbers refer to like elements throughout the drawings unless otherwise noted.

[0012] FIG. 1 shows a diagram of an example quantum system by which the aspects, features, and elements disclosed herein may be implemented.

[0013] FIG. 2A shows a flow diagram illustrating an example of a quantum computing technique by which the aspects, features, and elements disclosed herein may be implemented.

[0014] FIG. 2B shows a flow diagram illustrating an example of a quantum annealing technique by which the aspects, features, and elements disclosed herein may be implemented.

[0015] FIG. 3 shows a diagram of an example hybrid quantum-classical (HQC) computer by which the aspects, features, and elements disclosed herein may be implemented.

[0016] FIG. 4 shows a block diagram of an example internal configuration of a classical computing device by which the aspects, features, and elements disclosed herein may be implemented.

[0017] FIG. 5 shows a diagram of an example of a system by which the aspects, features, and elements disclosed herein may be implemented, such as generating a system design from an engineering objective.

[0018] FIG. 6 shows a diagram of an example of a system by which the aspects, features, and elements disclosed herein may be implemented, such as simulating an outcome of a parameter adjustment and then adjusting the parameter in a target system accordingly.

[0019] FIG. 7 shows a diagram of an example of a system by which the aspects, features, and elements disclosed herein may be implemented, such as simulating a measurement of an indirect variable and then adjusting an associated parameter in a target system accordingly.

[0020] FIG. 8 shows a flow diagram illustrating an example of a technique for adaptive agent-oriented system control for physical system design and operation.DETAILED DESCRIPTION

[0021] To describe some implementations in greater detail, the following reference numerals are used to indicate various elements and features relevant to the illustrated implementations.

[0022] With respect to FIG. 1: quantum system 100; quantum computer 102; qubits 104; control unit 106; control signals 108; measurement unit 110; measurement signals 112; and feedback signals 114.

[0023] With respect to FIG. 2A: technique 200; operation 202; operation 204; section 206; operation 210; operation 212; operation 214; operation 216; operation 218; operation 220; operation 222; and operation 224.

[0024] With respect to FIG. 2B: technique 250; quantum computer 252; classical computer 254; domain boundary 256; computational problem 258; initial Hamiltonian 260; final Hamiltonian 262; operation 264; initial state 266; operation 268; annealing schedule 270; final state 272; measurement operation 274; results 276; operation 278; and output 280.

[0025] With respect to FIG. 3: HQC 300; classical computer 306; processor 308; memory 310; bus 314; input 316; output 318; state preparation signals 332; control signals 334; and measurement output 338.

[0026] With respect to FIG. 4: classical computing device 400; processor 404; memory 406; power source 408; input component 410; output component 412; and communication component 414.

[0027] With respect to FIG. 5: system 500; engineering objective 502; initiator agent 504; set of tasks 506; set of task agents 508; set of outputs 512; executive agent 514; and system design 516.

[0028] With respect to FIG. 6: system 600; target physical system 602; physical sensor 604; direct-variable measurement 606; set of task agents 608; parameter-adjustment outcome prediction 612; executive agent 614; and parameter adjustment 616.

[0029] With respect to FIG. 7: system 700; target physical system 702; physical sensor 704; direct-variable measurement 706; set of task agents 708; indirect-variable measurement710; parameter-adjustment outcome prediction 712; executive agent 714; and parameter adjustment 716.

[0030] With respect to FIG. 8: technique 800; step 802; step 804; step 806; and step 808.

[0031] FIG. 1 shows a diagram of an example quantum system 100 by which the aspects, features, and elements disclosed herein may be implemented. The quantum system 100 may be, be similar to, include, or be included in an apparatus for performing one or more methods, processes, algorithms, operations, tasks, and / or techniques, as described herein. For example, the initiator agent 504, task agents of the set of task agents 508, and the executive agents 514 of FIG. 5 may each comprise or be comprised in one or more instances of the quantum system 100.

[0032] The quantum system 100 includes a quantum computer 102. The quantum computer 102 includes a plurality of qubits 104, which may be implemented in any of the ways disclosed herein. There may be any number of qubits 104 in the quantum computer 102. For example, the qubits 104 may include or consist of no more than 2 qubits, no more than 4 qubits, no more than 8 qubits, no more than 16 qubits, no more than 32 qubits, no more than 64 qubits, no more than 128 qubits, no more than 256 qubits, no more than 512 qubits, no more than 1024 qubits, no more than 2048 qubits, no more than 4096 qubits, or no more than 8192 qubits. These are merely examples, and in practice, there may be any number of qubits 104 in the quantum computer 102.

[0033] There may be any number of gates in a quantum circuit. However, in some implementations, the number of gates may be at least proportional to the number of qubits 104 in the quantum computer 102. In some implementations, the gate depth may be no greater than the number of qubits 104 in the quantum computer 102, or no greater than some linear multiple of the number of qubits 104 in the quantum computer 102 (e.g., 2, 3, 4, 5, 6, or 7).

[0034] The qubits 104 may be interconnected in any graph pattern. For example, they may be connected in a linear chain, a two-dimensional grid, an all-to-all connection, any combination thereof, or any subgraph of any of the preceding.

[0035] As will become clear from the description below, although the quantum computer 102 is referred to herein as a “quantum computer,” this does not imply that all components of the quantum computer 102 leverage quantum phenomena. One or more components of the quantum computer 102 may, for example, be classical (i.e., non-quantum) components that do not leverage quantum phenomena.

[0036] The quantum computer 102 includes a control unit 106, which may include any ofa variety of circuitry and / or other machinery for performing the functions disclosed herein. The control unit 106 may, for example, consist entirely of classical components. The control unit 106 generates and provides as output one or more control signals 108 to the qubits 104. The control signals 108 may take any of a variety of forms, such as any kind of electromagnetic signals, such as electrical signals, magnetic signals, optical signals (e.g., laser pulses), or any combination thereof.

[0037] In implementations wherein some or all of the qubits 104 are implemented as photons (also referred to as a “quantum optical” implementation) that travel along waveguides, the control unit 106 may be a beam splitter (e.g., a heater or a mirror), the control signals 108 may be signals that control the heater or the rotation of the mirror, the measurement unit 110 may be a photodetector, and the measurement signals 112 may be photons.

[0038] The measurement unit 110 performs one or more measurement operations on the qubits 104 to read out measurement signals 112 (also referred to herein as “measurement results”) from the qubits 104, where the measurement signals 112 are signals representing the states of some or all of the qubits 104. In practice, the control unit 106 and the measurement unit 110 may be entirely distinct from each other, or contain some components in common with each other, or be implemented using a single unit (i.e., a single unit may implement both the control unit 106 and the measurement unit 110). For example, a laser unit may be used both to generate the control signals 108 and to provide stimulus (e.g., one or more laser beams) to the qubits 104 to cause the measurement signals 112 to be generated.

[0039] In general, the quantum computer 102 may perform various operations described herein any number of times. For example, the control unit 106 may generate one or more control signals 108, thereby causing the qubits 104 to perform one or more quantum gate operations. The measurement unit 110 may then perform one or more measurement operations on the qubits 104 to read out a set of one or more measurement signals 112. The measurement unit 110 may repeat such measurement operations on the qubits 104 before the control unit 106 generates additional control signals 108, thereby causing the measurement unit 110 to read out additional measurement signals 112 resulting from the same gate operations that were performed before reading out the previous measurement signals 112. The measurement unit 110 may repeat this process any number of times to generate any number of measurement signals 112 corresponding to the same gate operations. The quantum computer 102 may then aggregate such multiple measurements of the same gate operations in any of a variety of ways.

[0040] After the measurement unit 110 has performed one or more measurement operations on the qubits 104 after they have performed one set of gate operations, the control unit 106 may generate one or more additional control signals 108, which may differ from the previous control signals 108, thereby causing the qubits 104 to perform one or more additional quantum gate operations, which may differ from the previous set of quantum gate operations. The process described above may then be repeated, with the measurement unit 110 performing one or more measurement operations on the qubits 104 in their new states (resulting from the most recently performed gate operations).

[0041] In implementations wherein some or all of the qubits 104 are implemented as charge-type qubits (e.g., transmon, X-mon, G-mon) or flux-type qubits (e.g., flux qubits, capacitively shunted flux qubits) (also referred to as a “circuit quantum electrodynamic” (circuit QED) implementation), the control unit 106 may be a bus resonator activated by a drive, the control signals 108 may be cavity modes, the measurement unit 110 may be a second resonator (e.g., a low-Q resonator), and the measurement signals 112 may be voltages measured from the second resonator using dispersive readout techniques.

[0042] In implementations wherein some or all of the qubits 104 are implemented as superconducting circuits, the control unit 106 may be a circuit QED-assisted control unit or a direct capacitive coupling control unit or an inductive capacitive coupling control unit, the control signals 108 may be cavity modes, the measurement unit 110 may be a second resonator (e.g., a low-Q resonator), and the measurement signals 112 may be voltages measured from the second resonator using dispersive readout techniques.

[0043] In implementations wherein some or all of the qubits 104 are implemented as trapped ions (e.g., electronic states of, e.g., magnesium ions), the control unit 106 may be a laser, the control signals 108 may be laser pulses, the measurement unit 110 may be a laser and either a CCD or a photodetector (e.g., a photomultiplier tube), and the measurement signals 112 may be photons.

[0044] In implementations wherein some or all of the qubits 104 are implemented using nuclear magnetic resonance (NMR) (in which case the qubits may be molecules, e.g., in liquid or solid form), the control unit 106 may be a radio frequency (RF) antenna, the control signals 108 may be RF fields emitted by the RF antenna, the measurement unit 110 may be another RF antenna, and the measurement signals 112 may be RF fields measured by the second RF antenna.

[0045] In implementations wherein some or all of the qubits 104 are implemented as nitrogen-vacancy centers (NV centers), the control unit 106 may, for example, be a laser, amicrowave antenna, or a coil, the control signals 108 may be visible light, a microwave signal, or a constant electromagnetic field, the measurement unit 110 may be a photodetector, and the measurement signals 112 may be photons.

[0046] In implementations wherein some or all of the qubits 104 are implemented as two-dimensional quasiparticles called “anyons” (also referred to as a “topological quantum computer” implementation), the control unit 106 may be nanowires, the control signals 108 may be local electrical fields or microwave pulses, the measurement unit 110 may be superconducting circuits, and the measurement signals 112 may be voltages.

[0047] In implementations wherein some or all of the qubits 104 are implemented as semiconducting material (e.g., nanowires), the control unit 106 may be microfabricated gates, the control signals 108 may be RF or microwave signals, the measurement unit 110 may be microfabricated gates, and the measurement signals 112 may be RF or microwave signals.

[0048] Although not shown explicitly in FIG. 1, and not required in some implementations described herein, the measurement unit 110 may provide one or more feedback signals 114 to the control unit 106 based on the measurement signals 112. For example, quantum computers referred to as “one-way quantum computers” or “measurementbased quantum computers” utilize the feedback signals 114 from the measurement unit 110 to the control unit 106. The feedback signals 114 are also necessary for the operation of fault-tolerant quantum computing and error correction.

[0049] The control signals 108 may, for example, include one or more state preparation signals that, when received by the qubits 104, cause some or all of the qubits 104 to change their states. Such state preparation signals constitute a quantum circuit also referred to as an “ansatz circuit.” The resulting state of the qubits 104 is referred to herein as an “initial state” or an “ansatz state.” The process of outputting the state preparation signal(s) to cause the qubits 104 to be in their initial state is referred to herein as “state preparation” (see FIG. 2A, section 206). A special case of state preparation is “initialization,” also referred to as a “reset operation,” in which the initial state is one in which some or all of the qubits 104 are in the “zero” state, e.g., the default single-qubit state. More generally, state preparation may involve using the state preparation signals to cause some or all of the qubits 104 to be in any distribution of desired states. In some implementations, the control unit 106 may first perform initialization on the qubits 104 and then perform preparation on the qubits 104, by first outputting a first set of state preparation signals to initialize the qubits 104, and by then outputting a second set of state preparation signals to put the qubits 104 partially or entirely into non-zero states.

[0050] Another example of control signals 108 that may be output by the control unit 106 and received by the qubits 104 are gate control signals. The control unit 106 may output such gate control signals, thereby applying one or more gates to the qubits 104. Applying a gate to one or more qubits causes the set of qubits to undergo a physical state change that embodies a corresponding logical gate operation (e.g., single-qubit rotation, two-qubit entangling gate or multi-qubit operation) specified by the received gate control signal. As this implies, in response to receiving the gate control signals, the qubits 104 undergo physical transformations that cause the qubits 104 to change state in such a way that the states of the qubits 104, when measured (see below), represent the results of performing logical gate operations specified by the gate control signals. The term “quantum gate,” as used herein, refers to the application of a gate control signal to one or more qubits to cause those qubits to undergo the physical transformations described above and thereby to implement a logical gate operation.

[0051] It should be understood that a dividing line between state preparation (and the corresponding state preparation signals) and the application of gates (and the corresponding gate control signals) may be chosen arbitrarily. For example, some or all the components and operations that are illustrated in FIG. 1 and FIGS. 2A-2B as elements of “state preparation” may instead be characterized as elements of gate application. Conversely, for example, some or all of the components and operations that are illustrated in FIG. 1 and FIGS. 2A-2B as elements of “gate application” may instead be characterized as elements of state preparation. As an example, the system and method of FIG. 1 and FIGS. 2A-2B may be characterized as solely performing state preparation followed by measurement, without any gate application, where the elements that are described herein as being part of gate application are instead considered to be part of state preparation. Conversely, for example, the system and method of FIG. 1 and FIGS. 2A-2B may be characterized as solely performing gate application followed by measurement, without any state preparation, and where the elements that are described herein as being part of state preparation are instead considered to be part of gate application.

[0052] FIG. 2A shows a flow diagram illustrating an example of a quantum computing technique 200 by which the aspects, features, and elements disclosed herein may be implemented. The technique 200 may be executed or performed by the quantum system 100 of FIG. 1. In general, the quantum system 100 may implement a plurality of quantum circuits as follows. For each quantum circuit C in the plurality of quantum circuits (FIG. 2 A, operation 202), the quantum system 100 performs a plurality of “shots” on the qubits 104. The meaning of a shot will become clear from the description that follows. For each shot S inthe plurality of shots (FIG. 2 A, operation 204), the quantum system 100 prepares the state of the qubits 104 (FIG. 2 A, section 206). More specifically, for each quantum gate G in quantum circuit C (FIG. 2 A, operation 210), the quantum system 100 applies quantum gate G to the qubits 104 (FIG. 2 A, operation 212 and operation 214).

[0053] Then, for each of the qubits 104 (FIG. 2 A, operation 216), the quantum system 100 measures the qubit 104 to produce measurement output representing a current state of qubit 104 (FIG. 2A, operation 218 and operation 220).

[0054] The operations described above are repeated for each shot S (FIG. 2A, operation 222), and circuit C (FIG. 2A, operation 224). As the description above implies, a single “shot” involves preparing the state of the qubits 104 and applying all of the quantum gates in a circuit to the qubits 104 and then measuring the states of the qubits 104; and the quantum system 100 may perform multiple shots for one or more circuits.

[0055] FIG. 2B shows a flow diagram illustrating an example of a quantum annealing technique 250 by which the aspects, features, and elements disclosed herein may be implemented. The technique 250 may be executed or performed by both a quantum computer 252 to the left of the domain boundary 256, such as the quantum system 100 of FIG. 1, and a classical computer 254 to the right of the domain boundary 256, such as the classical computer 306 of FIG. 3. Operations shown on the left of the domain boundary 256 are typically performed by the quantum computer 252, and operations shown on the right of the domain boundary 256 are typically performed by the classical computer 254. The

[0056] Quantum annealing starts with the classical computer 254 generating an initial Hamiltonian 260 and a final Hamiltonian 262 based on a computational problem 258 to be solved, and providing the initial Hamiltonian 260, the final Hamiltonian 262, and an annealing schedule 270 as input to the quantum computer 252. The quantum computer 252 prepares an initial state 266 (FIG. 2B, operation 264), such as a quantum-mechanical superposition of all possible states (candidate states) with equal weights, based on the initial Hamiltonian 260. The classical computer 254 provides the initial Hamiltonian 260, a final Hamiltonian 262, and an annealing schedule 270 to the quantum computer 252. The quantum computer 252 starts in the initial state 266 and evolves its state according to the annealing schedule 270, following the time-dependent Schrodinger equation, a natural quantummechanical evolution of physical systems (FIG. 2B, operation 268). More specifically, the state of the quantum computer 252 undergoes time evolution under a time-dependent Hamiltonian, which starts from the initial Hamiltonian 260 and terminates at the final Hamiltonian 262. If the rate of change of the system Hamiltonian is slow enough, the systemstays close to the ground state of the instantaneous Hamiltonian. If the rate of change of the system Hamiltonian is accelerated, the system may leave the ground state temporarily but produce a higher likelihood of concluding in the ground state of the final problem Hamiltonian, i.e., diabatic quantum computation. At the end of the time evolution, the set of qubits on the quantum annealer is in a final state 272, which is expected to be close to the ground state of the classical Ising model that corresponds to the solution to the computational problem 258. An experimental demonstration of the success of quantum annealing for random magnets was reported immediately after the initial theoretical proposal.

[0057] The final state 272 of the quantum computer 252 is measured, thereby producing results 276 (i.e., measurements) (FIG. 2B, measurement operation 274). The measurement operation 274 may be performed, for example, in any of the ways disclosed herein, such as in any of the ways disclosed herein in connection with the measurement unit 110 of FIG. 1. The classical computer 254 performs postprocessing on the results 276 to produce output 280 representing a solution to the computational problem 258 (FIG. 2B, operation 278).

[0058] FIG. 3 shows a diagram of an example HQC 300 by which the aspects, features, and elements disclosed herein may be implemented. The HQC 300 may be, be similar to, include, or be included in an apparatus for performing one or more methods, processes, algorithms, operations, tasks, and / or techniques, as described herein. For example, the initiator agent 504, task agents of the set of task agents 508, and the executive agents 514 of FIG. 5 may each comprise or be comprised in one or more instances of the HQC 300.

[0059] The HQC 300 includes a quantum computer 102 (which may, for example, be implemented in the manner shown and described in connection with FIG. 1) and a classical computer 306, which may comprise or be comprised in the classical computing device 400 of FIG. 4. The classical computer 306 may be a machine implemented according to the general computing model established by John Von Neumann, in which programs are written in the form of ordered lists of instructions and stored within a classical (e.g., digital) memory 310 and executed by a classical (e.g., digital) processor 308 of the classical computer 306. The memory 310 is classical in the sense that it stores data in a storage medium in the form of bits, which have a single definite binary state at any point in time. The bits stored in the memory 310 may, for example, represent a computer program. The classical computer 306 typically includes a bus 314. The processor 308 may read bits from and write bits to the memory 310 over the bus 314. For example, the processor 308 may read instructions from the computer program in the memory 310, and may optionally receive input 316 from a source external to the classical computer 306, such as from a user input device, such as a mouse,keyboard, or any other input device. The processor 308 may use instructions that have been read from the memory 310 to perform computations on data read from the memory 310 and / or the input 316, and generate output from those instructions. The processor 308 may store that output back into the memory 310 and / or provide the output externally as output 318 via an output device, such as a monitor, speaker, or network device.

[0060] The quantum computer component 102 may include a plurality of qubits 104 (which may, for example, be implemented in the manner shown and described in connection with FIG. 1). A single qubit may represent a one, a zero, or any quantum superposition of those two qubit states. The classical computer 306 may provide classical state preparation signals 332 to the quantum computer 102, in response to which the quantum computer 102 may prepare the states of the qubits 104 in any of the ways disclosed herein, such as in any of the ways disclosed in connection with FIGS. 1 and 2A-2B.

[0061] Once the qubits 104 have been prepared, the processor 308 may provide classical control signals 334 to the quantum computer 102, in response to which the quantum computer 102 may apply the gate operations specified by the control signals 334 to the qubits 104, as a result of which the qubits 104 arrive at a final state. The measurement unit 110 in the quantum computer 102 (which may be implemented as described above in connection with FIGS. 1 and 2A-2B) may measure the states of the qubits 104 and produce measurement output 338 representing the collapse of the states of the qubits 104 into one of their eigenstates. As a result, the measurement output 338 includes or consists of bits and therefore represents a classical state. The quantum computer 102 provides the measurement output 338 to the processor 308. The processor 308 may store data representing the measurement output 338 and / or data derived therefrom in the memory 310.

[0062] The steps described above may be repeated any number of times, with what is described above as the final state of the qubits 104 serving as the initial state of the next iteration. In this way, the classical computer 306 and the quantum computer 102 may cooperate as co-processors to perform joint computations as a single computer system.

[0063] Although certain functions may be described herein as being performed by a classical computer and other functions may be described herein as being performed by a quantum computer, these are merely examples and do not constitute limitations of the present invention. A subset of the functions that are disclosed herein as being performed by a quantum computer may instead be performed by a classical computer. For example, a classical computer may execute functionality for emulating a quantum computer and provide a subset of the functionality described herein, albeit with functionality limited by theexponential scaling of the simulation. Functions that are disclosed herein as being performed by a classical computer may instead be performed by a quantum computer.

[0064] The techniques described above may be implemented, for example, in hardware, in one or more computer programs tangibly stored on one or more computer-readable media, firmware, or any combination thereof, such as solely on a quantum computer, solely on a classical computer, or on an HQC computer. The techniques disclosed herein may, for example, be implemented solely on a classical computer, in which the classical computer emulates the quantum computer functions disclosed herein.

[0065] The techniques described herein may be implemented in one or more computer programs executing on (or executable by) a programmable computer (such as a classical computer, a quantum computer, or an HQC) including any combination of any number of the following: a processor, a storage medium readable and / or writable by the processor (including, for example, volatile and non-volatile memory and / or storage elements), an input device, and an output device. Program code may be applied to input entered using the input device to perform the functions described and to generate output using the output device.

[0066] FIG. 4 is a block diagram of an example internal configuration of a classical computing device 400 capable of performing functions described herein. The classical computing device 400 may be, be similar to, include, or be included in an apparatus for performing one or more methods, processes, algorithms, operations, tasks, and / or techniques, as described herein. For example, the initiator agent 504, task agents of the set of task agents 508, and the executive agents 514 of FIG. 5 may each comprise or be comprised in one or more instances of the classical computing device 400.

[0067] The classical computing device 400 includes components or units, such as a processor 404, a memory 406, a power source 408, an input component 410, an output component 412, and a communication component 414, as well as other suitable components. One or more of the processor 404, the memory 406, the power source 408, the input component 410, the output component 412, and the communication component 414 may communicate with each other via the bus 402.

[0068] The processor 404 may include one or more chiplets, chips, system-on-chips (SoCs), network-on-chips (NoCs), chipsets, packages, or devices that individually or collectively constitute or include a processing system. The processing system includes a processor (or “processing”) circuitry in the form of one or multiple processors, microprocessors, processing units (such as central processing units (CPUs), graphics processing units (GPUs), neural processing units (NPUs), and / or digital signal processors(DSPs), processing blocks, application-specific integrated circuits (ASICs), programmable logic devices (PLDs), field-programmable gate arrays (FPGAs), or other discrete gate or transistor logic or circuitry (all of which may be generally referred to herein individually as “processors” or collectively as “the processor” or “the processor circuitry”). The processor 404 may include multiple processors interconnected in one or more manners, including hardwired or networked. Operations of the processor 404 can be distributed across multiple devices or units that can be coupled directly or across a local area or other suitable type of network. The processor 404 may include a cache, or cache memory, for local storage of operating data or instructions.

[0069] One or more of the processors may be individually or collectively configurable or configured to perform various operations described herein. In some implementations, a single processor may perform all of the operations described as being performed by the one or more processors. In some implementations, a group of processors collectively configurable or configured to perform a set of operations may include a first set of (one or more) processors configurable or configured to perform a first operation of the set and a second processor configurable or configured to perform a second operation of the set, or may include the group of processors all being configured or configurable to perform the set of operations. The first set of processors and the second set of processors may be the same set of processors or may be different sets of processors.

[0070] The memory 406 includes one or more memory components, which may each be volatile memory or non-volatile memory, that individually or collectively constitute a memory system. The memory system may include memory circuitry in the form of one or more memory devices, memory blocks, memory elements, or other discrete gate or transistor logic or circuitry, each of which may include tangible storage media such as random-access memory (RAM) or read-only memory (ROM), or combinations thereof (all of which may be generally referred to herein individually as “memories” or collectively as “the memory,” “the memory system,” or “the memory circuitry”). The memory 406 may include non-transitory memory, transitory memory, or a combination thereof. Volatile memory may include RAM (e.g., a dynamic RAM (DRAM) module, such as a double data rate (DDR) synchronous DRAM (SDRAM)). Non-volatile memory may include a disk drive, a solid-state drive, flash memory, or phase-change memory. In some implementations, the memory 406 may be distributed across multiple devices. For example, the memory 406 may include networkbased memory or memory in multiple clients or servers performing the operations of those multiple devices. The memory 406 may be referred to as one or more computer-readablestorage media. A computer-readable storage medium may include any storage unit (or multiple storage units) that stores data or instructions that are readable by a processing system. A computer-readable storage medium may include, for example, at least one of a data repository, a data storage unit, a computer memory, a hard drive, a disk, or a random-access memory.

[0071] One or more of the memories may be coupled (for example, operatively coupled, communicatively coupled, electronically coupled, or electrically coupled) with one or more of the processors and may individually or collectively store executable instructions (e.g., code such as software) that, when executed by one or more of the processors, may configure or otherwise cause one or more of the processors to perform various functions or operations described herein. “Software” shall be construed broadly to mean instructions, instruction sets, code, code segments, program code, programs, subprograms, software modules, applications, software applications, software packages, routines, subroutines, objects, executables, threads of execution, procedures, and / or functions, among other examples, whether referred to as software, firmware, middleware, microcode, hardware description language, or otherwise.

[0072] In some implementations, the executable instructions may include application data or an operating system, among other examples. The executable instructions may include one or more application programs, which may be loaded or copied, in whole or in part, from nonvolatile memory to volatile memory to be executed by the processor 404. For example, the executable instructions may include instructions for performing techniques described in this disclosure. In some implementations, the application data may include functional programs, such as computational programs, analytical programs, or database programs, among other examples. The operating system may be, for example, Microsoft Windows®, Mac OS X®, or Linux®; an operating system for a mobile device, such as a smartphone or tablet device; or an operating system for a non-mobile device, such as a mainframe computer.

[0073] Reference to “one or more memories” should be understood to refer to any one or more memories of a corresponding device, such as the memory described in connection with FIG. 4. For example, operations described as being performed by, or data described as being stored on, one or more memories can be performed by, or stored on, respectively, the same subset of the one or more memories or different subsets of the one or more memories.Additionally or alternatively, in some examples, one or more of the processors may be preconfigured to perform various functions or operations described herein without requiring configuration by software. For example, the memory 406 may include data or instructions that are hard-wired into the processing system.

[0074] In the description herein, language describing a system, an apparatus, or a device as taking an action (such as performing, determining, initiating, receiving, calculating, deciding, computing, processing, etc.) is to be understood as describing that some appropriate component of the system, apparatus, or device is taking the action. As used herein, the term “component” is intended to be broadly construed as hardware and / or a combination of hardware and software.

[0075] The power source 408 provides power to the classical computing device 400. For example, the power source 408 may be an interface to an external power distribution system. In an example, the power source 408 may be a battery, such as where the classical computing device 400 is a mobile device or is otherwise configured to operate independently of an external power distribution system. In some implementations, the classical computing device 400 may include or otherwise use multiple power sources. In some such implementations, the power source 408 can be a backup battery.

[0076] The input component 410 and / or the output component 412 may include one or more input interfaces and / or output interfaces configured for facilitating communication between the classical computing device 400 and one or more peripheral devices, such as, for example, one or more sensors, detectors, displays, input devices, or other devices configured for facilitating interaction with the classical computing device 400 or the environment around the classical computing device 400. An input device may, for example, include a positional input device, such as a mouse, touchpad, touchscreen, or the like; a keyboard; or another suitable human or machine interface device. An output device may, for example, include a display, such as a liquid crystal display, a cathode-ray tube, a light-emitting diode display, or other suitable display. In some implementations, the peripheral devices may include a geolocation component, such as a GPS location unit. In some examples, the peripheral devices may include a temperature sensor for measuring temperatures of components of the classical computing device 400, such as the processor 404.

[0077] The communication component 414 may include an interface for facilitating a connection or link to a network. The communication component 414 may include a wired network interface or a wireless network interface. The classical computing device 400 may communicate with other devices via the communication component 414 using one or more network protocols, such as using Ethernet, TCP, IP, power line communication, an IEEE 802.X protocol (e.g., Wi-Fi, Bluetooth, or ZigBee), infrared, visible light, general packet radio service (GPRS), global system for mobile communications (GSM), code-division multiple access (CDMA), Z-Wave, a cellular communication protocol, another protocol, or acombination thereof. For example, the classical computing device 400 can communicate with a database server.

[0078] The communication component 414 may include a transceiver, which may include a transmitter or a receiver. In some configurations, one or a combination of antenna(s), modem(s), multiple input multiple output (MIMO) detectors, receive processors, transmit processors, and / or the transmit MIMO processors may be included in the transceiver. The transceiver may be under control of or used by one or more processors, and in some aspects in conjunction with processor-readable code stored in the memory, to perform aspects of the methods, processes, techniques, and / or operations described herein.

[0079] FIG. 5 shows a diagram of an example of a system 500 by which the aspects, features, and elements disclosed herein may be implemented, and in particular, generating a system design 516, for a target physical system, based on an engineering objective 502. The system 500 comprises a set of specialized agents that includes an initiator agent 504, a set of task agents 508, and an executive agent 514.

[0080] An engineering objective 502 is a high-level goal or a set of requirements for the target physical system, such as a chemical reactor, an industrial manufacturing plant, an aerospace vehicle, or a power distribution grid. The engineering objective 502 defines a desired performance, functionality, or configuration of the target physical system in broad terms, such as a target production yield for a chemical reactor, a fuel efficiency rating for an engine, or a specific structural load capacity for a physical component. The engineering objective 502 serves as a primary input to the system 500 and is typically characterized by its multi-dimensional nature, often encompassing several overlapping or competing requirements. In the multi -agent framework described herein, the engineering objective 502 represents the ultimate problem statement that is governed by a set of discrete mathematical or physical constraints.

[0081] Because the engineering objective 502 describes a global outcome rather than specific execution steps, it must be parsed and translated by the initiator agent 504 into a set of discrete mathematical or physical constraints that govern the target physical system.Accordingly, the initiator agent 504 may comprise logic or instructions to decompose the engineering objective 502 into the set of tasks 506, where each task of the set of tasks 506 represents at least one discrete mathematical or physical constraint. To perform this decomposition, the initiator agent 504 may utilize a generative artificial intelligence model to parse the high-level requirements of the engineering objective 502 and identify the specific sub-problems or constraints that must be resolved. For example, the initiator agent 504 maycomprise a large language model (LLM), a rule-based expert system, a symbolic artificial intelligence system utilizing an engineering ontology, or a hierarchical task network (HTN) planner. In some implementations, the initiator agent 504 utilizes a rule-based engine to map the engineering objective 502 to the set of tasks 506 based on a deterministic logic tree. In some implementations, the initiator agent 504 may utilize a symbolic artificial intelligence (Al) framework to leverage a knowledge graph for identifying physical constraints.Additionally, the initiator agent 504 may comprise an automated planning system using a planning domain definition language (PDDL) to generate a sequence of actions to satisfy the engineering objective 502.

[0082] The initiator agent 504 is responsible for identifying and selecting individual task agents of the set of task agents 508 to perform the set of tasks 506. In some implementations, each task of the set of tasks 506 may be mapped to an individual task agent of the set of task agents 508. In some implementations, multiple tasks of the set of tasks 506 may be mapped to an individual task agent of the set of task agents 508. In some implementations, an individual task of the set of tasks 506 may be mapped to multiple task agents of the set of task agents 508. For example, the set of task agents may comprise a first task agent and a second task agent that are both configured to perform a same task of the set of tasks using different implementations of the underlying mathematical or physical constraints, wherein the executive agent 514 determines which of a first output of the first task agent or a second output of the second task agent more closely satisfies a predetermined constraint, and wherein the executive agent 514 discards or disregards the less satisfactory output when generating the system design 516.

[0083] In some implementations, the initiator agent 504 may identify and select the task agents of the set of task agents 508 for execution of the set of tasks 506 based on broadcast messages to a network of candidate agents. Specifically, the initiator agent 504 may broadcast a query to a network of candidate agents, indicating a respective task, and select at least one candidate agent of the network whose response indicates a match for the respective task. This broadcast query may be distributed via a message broker or a peer-to-peer discovery protocol and may include a task specification defining input data requirements, expected output formats, and specific physical constraints of the target physical system. The at least one response received from the at least one candidate agent may include a capability descriptor that characterizes the internal logic of the candidate agent, such as a specific mechanistic model or a version of a generative artificial intelligence model. The initiator agent 504 then evaluates the capability descriptor against the requirements of the respective task to ensure atechnical match. In some implementations, the selection of the at least one candidate agent is further based on a performance metric, such as a computational resource estimate, a predicted accuracy score, or a latency guarantee provided in the at least one response. This allows the initiator agent 504 to dynamically assemble the set of task agents 508 as a modular workforce tailored to the multi-dimensional requirements of the engineering objective 502.

[0084] In some implementations, the initiator agent 504 may identify and select the task agents of the set of task agents 508 for execution of the set of tasks 506 based on a registry of candidate agents. Specifically, the initiator agent 504 may query a registry with a respective task and select at least one candidate agent based on a registered characterization stored within the registry. The registry may be implemented as a centralized or distributed database, such as a service catalog or a vector database of agent capabilities, and may include entries that categorize candidate agents according to their ability to resolve specific physical or mathematical constraints. The registered characterization for a candidate agent may include a capability descriptor that defines the internal logic of the agent, such as a specific mechanistic model or a version of a generative artificial intelligence model. The initiator agent 504 then compares the requirements of the respective task against the registered characterization to ensure a technical match. In some implementations, the selection of the at least one candidate agent from the registry is further based on a performance metric, such as a computational resource estimate, a historical reliability score, or a latency guarantee associated with the registered characterization. This allows the initiator agent 504 to systematically assemble the set of task agents 508 as a modular workforce tailored to the multi-dimensional requirements of the engineering objective 502.

[0085] Returning to the various possible mappings between the set of tasks 506 and the set of task agents 508, e.g., one-to-one, one-to-many, and many-to-one, it may be advantageous for the individual task agents of the set of task agents 508 to communicate with each other in executing their respective tasks. For example, there may be, and often are, dependencies between tasks, such that execution, or performance, of one or more first tasks by one or more first task agents depends on one or more outputs of execution, or performance, of one or more second tasks by one or more second task agents. Accordingly, the set of specialized agents — which includes the set of task agents 508 as well as the initiator agent 504 and the executive agent 514 — may be arranged in a matrix topology, wherein resolving the task dependencies comprises cyclical data exchange between at least two task agents of the set of task agents.

[0086] In an example cyclic data exchange, a first task agent of the set of task agents 508generates an intermediate output that serves as an input parameter for a second task agent of the set of task agents 508. The second task agent, in turn, processes this input to generate a subsequent output that may be fed back to the first task agent or distributed to additional agents within the set of task agents 508. This iterative loop continues until the specialized agents reach a state of convergence or a steady-state solution that satisfies the coupled mathematical or physical constraints of their respective tasks. For instance, in the design of a chemical reactor, a task agent responsible for fluid dynamics may exchange data cyclically with a task agent responsible for thermal management to resolve the non-linear relationship between flow rate and temperature distribution. By facilitating this autonomous negotiation through a matrix topology, the set of task agents 508 can resolve complex interdependencies and multi-physics interactions without requiring a centralized controller to orchestrate every individual data transfer or intermediate calculation.

[0087] In some implementations, the set of task agents 508 is configured to communicate with one another using a publish-subscribe architecture to resolve the task dependencies. For example, a task agent of the set of task agents 508 may publish its intermediate results or a final output from the set of outputs 512 to a message broker or a distributed event bus under a specific topic corresponding to a physical parameter or a mathematical constraint. Other task agents within the set of task agents 508, which require that parameter to resolve their own assigned tasks, subscribe to that topic to receive updates in real-time. This decoupled interaction allows the set of task agents 508 to synchronize their operations without a centralized orchestration layer, as each agent only consumes the specific data relevant to its domain-specific logic. Furthermore, the publish-subscribe architecture facilitates the addition or substitution of task agents within the system 500, as new agents can simply subscribe to existing data streams to begin contributing to the resolution of the engineering objective 502 without requiring direct reconfiguration of the other agents.

[0088] Each task agent of the set of task agents 508 is configured to perform one or more tasks of the set of tasks 506. Because the set of tasks 506 includes tasks that concern discrete mathematical or physical constraints, at least some of the task agents of the set of task agents 508 are configured to implement domain-specific logic, such as a mechanistic model representing a physical or chemical law, a heuristic, or a generative artificial intelligence model. This domain-specific logic allows each task agent to function as a modular expert, resolving the mathematical or physical constraints of its assigned task to produce a corresponding output within the set of outputs 512. For example, a task agent may implement a mechanistic model to calculate a mass balance or a structural stress limit, while another taskagent may utilize a generative artificial intelligence model to propose an optimized geometry based on a set of performance heuristics. This independent application of specialized logic ensures that each discrete constraint of the engineering objective 502 is addressed by an agent with the appropriate technical expertise.

[0089] Although certain task agents of the set of task agents 508 implement respective mechanistic models — such as a deterministic simulation of a physical law — they retain functional agency within the system 500. Rather than acting as a passive calculation script, each task agent is an autonomous entity capable of evaluating the state of its assigned constraint and making internal decisions to reach a valid output. In some implementations, this agency is characterized by the agent’s ability to independently tune its own parameters, to determine when its local solution has converged, and to proactively negotiate with other task agents to resolve data dependencies. For instance, a task agent managing a chemical reaction model may not merely execute a formula, but it may monitor incoming data from the matrix topology, identify if its current model parameters are physically inconsistent with the inputs provided by a neighboring agent, and iteratively adjust its internal logic to maintain the integrity of the constraints of the target physical system. Consequently, many of the task agents of the set of task agents 508 are not simply “dumb” execution units, but are specialized, intelligent solvers that possess the localized authority to navigate their specific mathematical or physical domains without constant oversight from the initiator agent 504. However, the framework of specialized agents described herein does not preclude “dumb” agents of the set of specialized agents from participating in producing outputs of the set of outputs 512 toward generating the system design 516 for the target physical system.

[0090] In some implementations, each task agent of the set of task agents 508 includes internal validation logic to ensure that its generated output is physically or mathematically consistent with the engineering objective 502. This internal validation logic allows a task agent to evaluate its own results against a predefined convergence criterion or a physical boundary condition. If a task agent determines that its current output violates a physical law — such as an impossible energy state or a structural failure — it autonomously re-executes its internal model with adjusted parameters. This localized self-correction capability ensures that the set of outputs 512 consists of technically viable solutions before they are integrated by the executive agent 514, as described later herein.

[0091] To facilitate the exchange of data between task agents of the set of task agents 508 utilizing different types of logic, such as a deterministic mechanistic model and a probabilistic generative artificial intelligence model, each task agent may include a datanormalization layer. This layer translates high-dimensional generative outputs, such as a proposed 3D component geometry, into a structured format that a mechanistic agent can ingest as an input parameter for a simulation, such as a finite element analysis. By bridging the gap between these different computational paradigms, the set of task agents 508 can collectively resolve the multi-dimensional constraints of the target physical system. This ensures that the qualitative proposals of a generative Al agent are strictly validated by the quantitative rigor of a mechanistic agent within the cyclic data exchange.

[0092] The set of outputs 512 represents the collective results produced by the set of task agents 508, where each individual output corresponds to the resolution of a specific mathematical or physical constraint defined in the set of tasks 506. These outputs serve as the technical bridge between the specialized, decentralized processing of the task agents and the global synthesis performed on the set of outputs 512 by the executive agent 514. Rather than being disparate pieces of data, the set of outputs 512 constitutes a synchronized snapshot of the target physical system’s sub -components or parameters. For example, in an industrial manufacturing environment, one output might define the optimal thermal gradient for a curing process, while another might define the mechanical torque requirements for a robotic assembly arm.

[0093] Each output within the set of outputs 512 may be a validated technical artifact according to either the validation logic integral to a respective task agent or to a correct-by-design model implemented by the respective task agent. In implementations where the task agents of the set of task agents 508 utilize internal validation logic and participate in cyclic data exchange, the resulting outputs are already checked for local physical consistency and inter-agent compatibility before they are finalized. This ensures that the set of outputs 512 does not contain contradictory data that would violate the fundamental laws governing the target physical system. Furthermore, these outputs are typically structured in a standardized, machine-readable format — such as a series of state vectors, boundary condition definitions, or optimized coordinate sets — allowing them to be ingested seamlessly by the integration logic of the executive agent 514.

[0094] In some implementations, the set of outputs 512 may also include metadata regarding the confidence intervals or the computational methods used to derive each result. This metadata allows the system 500 to maintain a transparent audit trail of how each high-level requirement of the engineering objective 502 was decomposed and resolved. By providing a comprehensive and validated representation of the design requirements, the set of outputs 512 functions as the finalized ingredients from which the executive agent generates1the system design 516.

[0095] Because the set of task agents 508 may operate at different computational scales, the set of outputs 512 may be populated asynchronously. The system 500 may utilize a temporary storage buffer or a state-management service to aggregate these outputs as they are completed. In some implementations, the initiator agent 504 may implement the temporary storage buffer, and further, the initiator agent 504 may select and instruct alternative task agents of the set of task agents 508 to perform the tasks corresponding to outputs that the initiator agent 504 determines are absent, corrupted, invalid, incorrect, or the like. In some implementations, the executive agent 514 may implement the temporary storage buffer, and the executive agent 514 may communicate with the initiator agent 504 when the executive agent 514 determines certain outputs are absent, corrupted, invalid, incorrect, or the like.

[0096] In some implementations, once a threshold of completeness on the set of outputs 512 is met — for example, as determined by the fulfillment of all mandatory tasks in the set of tasks 506 — the set of outputs 512 may be marked as ready for integration. This may be referred to as batch processing. However, in some implementations, integration of outputs may occur incrementally, e.g., before all outputs of the set of outputs 512 are available. This may be referred to as incremental processing or asynchronous integration. In some implementations, one or more specialized agents of the set of specialized agents not shown in FIG. 1 may perform the integration of the set of outputs 512, such as an “integration agent.” In some implementations, the executive agent 514 performs the integration of the set of outputs 512.

[0097] In an incremental processing model of the set of outputs 512, a specialized agent, such as the executive agent 514, begins the synthesis of the system design 516 as soon as a functional subset of the set of outputs 512 is received. This allows the system 500 to identify high-level design conflicts earlier in the cycle. For example, if the task agents that perform structural and thermal modeling finish their tasks early, the executive agent 514 can validate their specific intersection before other agents that perform propulsion or control modeling have finalized their respective outputs. This overlapping integration may reduce the total time required for generating the system design 516 and allows for faster feedback loops between the executive agent 514 and the set of task agents 508.

[0098] Feedback with the set of specialized agents is an important aspect of the adaptive and iterative nature of the framework of specialized agents described herein. For example, feedback between the executive agent 514 and one or more task agents of the set of task agents 508 enables continuously evaluating subsets of the set of outputs 512 against theglobal requirements of the engineering objective 502. Upon detection of a design conflict or a physical inconsistency, the executive agent 514 may generate a refinement signal or a set of updated boundary conditions. This signal is relayed back to the specific task agents of the set of task agents 508 whose outputs are implicated in the conflict, prompting an autonomous re-evaluation or adjustment of their local models. By resolving these discrepancies mid-cycle, the system 500 can prevent a compounding of errors that can occur in siloed or strictly sequential design processes.

[0099] For example, if the initial outputs for tasks concerning structural geometry and thermal dissipation are found to be mutually exclusive, the executive agent 514 can immediately coordinate a trade-off negotiation between the affected members of the set of task agents 508. This fail-fast approach ensures that the computational resources of the set of task agents 508 are not expended on downstream tasks that are predicated on an invalid foundation. As the set of outputs 512 continues to populate, the executive agent 514 provides increasingly granular constraints to the remaining agents, ensuring the entire system design 516 converges toward a globally optimized solution.

[0100] As another example, the system 500 may be configured to perform tasks redundantly, such as to determine which of a plurality of task agents produces outputs fastest, which produces outputs that are most accurate, or which produces outputs with the highest degree of precision. In this context, accuracy refers to how closely an output from the set of task agents 508 conforms to the true physical or mathematical requirements of the engineering objective 502, whereas precision refers to the repeatability and consistency of the logic of the task agent across varying input conditions or multiple execution cycles.

[0101] By employing redundancy, the initiator agent 504 or the executive agent 514 can perform a comparative analysis of different computational methodologies — such as comparing a high-speed generative Al model against a slower, high-fidelity mechanistic model — to select the most reliable result for inclusion in the set of outputs 512. This competitive execution enables the system 500 to dynamically prioritize specific members of the set of task agents 508 based on the immediate needs of the engineering objective 502, whether the priority is rapid iteration or extreme technical rigor. Furthermore, the redundant results can be used to calculate a confidence score for a given task, where a high degree of correlation between multiple task agents indicates a robust solution, while a high variance may trigger a refinement signal for further processing by the set of task agents 508.

[0102] One example of employing redundancy within the set of task agents 508 concerns selecting between redundant outputs according to which output more closely satisfies apredetermined constraint, such as a constraint determined by the initiator agent 504 during parsing and translation of the engineering objective 502. In this example, the initiator agent 504 determines the set of task agents 508 to include a first task agent and a second task agent that are both configured to perform a same task of the set of tasks 506 using different implementations, e.g., different models of the underlying mathematical or physical constraints. The initiator agent 504 instructs both task agents to perform the same task, and both produce their respective output. Subsequently, the executive agent 514 receives and analyzes the outputs, and determines that a first output of the first task agent more closely satisfies a predetermined constraint than does a second output of the second task agent.Accordingly, the executive agent 514 discards the second output from the set of outputs, e.g., the “less desirable” second output is not integrated into the set of outputs 512 for generation of the system design by the executive agent 514.

[0103] Another example of employing redundancy within the set of task agents 508 concerns selecting between redundant outputs according to which output causes a third output of a third task agent, whose third task depends alternatively from a first output of a first task agent or a second output of a second task agent, to more closely satisfy a predetermined constraint, such as a constraint determined by the initiator agent 504 during parsing and translation of the engineering objective 502. In this example, the initiator agent 504 determines the set of task agents 508 to include a first task agent and a second task agent that are both configured to perform a same task of the set of tasks 506 using different implementations, e.g., different models of the underlying mathematical or physical constraints. The initiator agent 504 instructs the first task agent to produce the first output and the second task agent to produce the second output. The set of task agents 508 further includes at least one third task agent whose third task depends alternatively on the first output or the second output, e.g., either the first output alone or the second output alone is sufficient for the at least one third task agent to perform the third task. The initiator agent 504 instructs the at least one third task agent to produce the third output, once using the first output and once using the second output. The at least one third task agent may consist of a single third task agent, in which case the single third task agent performs the third task twice (once using the first output and once using the second output). Alternatively, the at least one third task agent may comprise two identically configured third task agents, in which case one third task agent performs the third task using the first output and the other third task agent performs the third task using the second output. Subsequently, the executive agent 514 receives and analyzes the third outputs, and determines that the third output that was based on the firstoutput more closely satisfies a predetermined constraint than does the third output that was based on the second output. Accordingly, the executive agent 514 discards the second output and its associated third output from the set of outputs, e.g., the “less desirable” second output and its associated third output are not integrated into the set of outputs 512 for generation of the system design by the executive agent 514.

[0104] The executive agent 514 may comprise logic or instructions to integrate the set of outputs 512 into the system design 516. To perform this synthesis, the executive agent 514 may utilize an optimization engine, such as a genetic algorithm, a particle swarm optimizer, or a gradient-descent solver, to reconcile competing parameters within the set of outputs 512. In some implementations, the executive agent 514 may comprise a multi-criteria decisionmaking (MCDM) framework, such as an analytic hierarchy process (AHP) or a technique for order of preference by similarity to ideal solution (TOPSIS), to arbitrate between mutually exclusive engineering requirements. Additionally, the executive agent 514 may utilize a generative artificial intelligence model, such as a multimodal transformer architecture, specifically trained to synthesize disparate technical data into a unified system design 516. The executive agent 514 may further incorporate a constraint satisfaction solver (CSS) or a symbolic reasoning engine to ensure the final system design 516 adheres to the high-level physical laws defined by the engineering objective 502.

[0105] In some implementations, the executive agent 514 identifies design conflicts that emerge when individual outputs from the set of task agents 508 are mutually exclusive or physically incompatible. For example, a first output from a first task agent might call for a material thickness that violates a weight constraint established in a second output from a second task agent. To resolve such a conflict, the executive agent 514 calculates a trade-off, e.g., a balanced set of parameters designed to satisfy the competing constraints of both outputs without compromising the global integrity of the engineering objective 502. This calculation may involve weighting the importance of each constraint or using multi -objective optimization to find a Pareto-optimal middle ground where the requirements of the target physical system are best met.

[0106] Following the calculation of this trade-off, the executive agent 514 may feedback the refined parameters to the initiator agent 504 for execution of iterative design loops. The initiator agent 504 may respond, for example, by instructing the first task agent and the second task agent to re-perform their respective tasks using the trade-off values as a new set of boundary conditions. This feedback loop ensures that the individual task agents within the set of task agents 508 iterate on their localized solutions until they are synchronized with thebroader system requirements. Rather than simply rejecting an invalid output, the system 500 uses this arbitration process to guide the specialized agents toward a convergent design, facilitating a resolution of multi-dimensional conflicts inherent in the balancing of competing physical laws within a complex system.

[0107] In some implementations, the executive agent 514 may determine that the integration of the set of outputs 512 fails to satisfy a global system constraint defined by the engineering objective 502, as interpreted by the initiator agent 504. Unlike a localized conflict between specific task agents of the set of task agents 508, a global failure indicates that the current decomposition or the original requirements are physically or mathematically unreachable within the current design space. Upon such a determination, the executive agent 514 communicates the failure to the initiator agent 504, which then determines a modified engineering objective to account for the failure. This modification may involve relaxing specific performance targets, adjusting the scope of the target physical system, or redefining the high-level goals of the engineering objective 502 to reach a feasible solution space.

[0108] Once the modified engineering objective is established, the system 500 initiates a comprehensive iteration of the design process. This includes the initiator agent 504 reperforming the decomposition of the modified engineering objective into a new or revised set of tasks 506, identifying and instructing the appropriate task agents of the set of task agents 508, and the executive agent 514 subsequently generating a new system design 516 based on the resulting new set of outputs 512. This iterative structure allows the system 500 to adapt to fundamental physical limitations that may only become apparent during the integration phase, ensuring that the final system design 516 is both optimized and physically realizable.

[0109] In some implementations, generating the system design 516 includes the utilization of a digital twin by at least one task agent of the set of task agents 508. A digital twin comprises a high-fidelity virtual representation of the target physical system that mirrors its real-world functional and physical characteristics. In this context, the at least one task agent ingests the integrated parameters or the functional subsets of the set of outputs 512 to simulate the performance of a candidate system design within a virtual environment. This simulation allows the system 500 to validate how the integrated components and constraints interact under operational conditions before the final system design 516 is committed.

[0110] By incorporating a digital twin into the logic of the set of task agents 508, the system 500 can perform stress tests, thermal analysis, or fluid dynamics simulations on the synthesized model of the target physical system according to the candidate system design. The results of these simulations provide a feedback mechanism to verify that the integratedoutputs from the set of task agents 508 collectively satisfy the multi-dimensional requirements of the engineering objective 502. If the digital twin simulation reveals a performance gap or a physical failure, for example, as determined by the at least one task agent executing the digital twin, the at least one task agent can generate a specific refinement signal for the executive agent 514 or the initiator agent 504 to trigger an iterative adjustment of the set of tasks 506. This helps to ensure that the final system design 516 is not just a theoretical assembly of parts, but a mathematically and physically verified blueprint of the target physical system.[OHl] FIG. 6 shows a diagram of an example of a system 600 by which the aspects, features, and elements disclosed herein may be implemented, such as simulating an outcome of a parameter adjustment and then adjusting the parameter in a target physical system 602 accordingly. The system 600 may comprise or be comprised in the system 500 of FIG. 5. The system 600 comprises a set of specialized agents that includes a set of task agents 608, which may be the set of task agents 508 of FIG. 5, and an executive agent 614, which may be the executive agent 514 of FIG. 5.

[0112] The target physical system 602 represents the physical environment, industrial process, or mechanical entity to which a system design, such as the system design 516 of FIG. 5, may be applied, and which may be further subject to the control, monitoring, and optimization of the system 600. Examples of the target physical system 602 include a chemical reactor, a biomanufacturing facility, an aerospace vehicle, an industrial manufacturing plant, or a power distribution grid. The target physical system 602 may be characterized by complex, often non-linear dynamics and a set of operational variables — such as temperature, pressure, flow rate, or chemical concentration — that define the state of the target physical system 602. The target physical system 602 may include a plurality of hardware components and control interfaces that allow the executive agent 614 to adjust operational parameters in order to satisfy the high-level requirements of an engineering objective, such as the engineering objective 502 of FIG. 5.

[0113] To facilitate real-time control, the target physical system 602 is associated with at least one physical sensor 604 that generates direct-variable measurements 606 for observable parameters of the target physical system 602. The physical sensor 604 may comprise a suitable instrument or transducer configured to detect a physical property of the target physical system 602 and convert that property into a direct-variable measurement 606. For example, depending on the engineering objective and the system design of the target physical system 602, the physical sensor 604 may include a thermocouple or resistance temperaturedetector for measuring thermal state, a piezoelectric transducer or strain gauge for measuring pressure or mechanical tension, an ultrasonic or electromagnetic flow meter for measuring fluid dynamics, or a gas chromatograph or electrochemical probe for measuring chemical concentrations.

[0114] The physical sensor 604 may be implemented as a standalone unit, an array of distributed sensors, or an embedded component integrated directly into the structural materials or flow paths of the target physical system 602. In some implementations, the physical sensor 604 may utilize micro-electromechanical systems (MEMS) or fiber-optic sensing techniques to provide high-resolution data within constrained or high-interference environments. The physical sensor 604 may be configured to output the direct-variable measurement 606 as a digital signal or a synchronized data stream, which may be processed by a signal conditioning layer, such as by dedicated hardware of the physical sensor 604 or by a task agent of the set of task agents 608, to remove noise or bias before the data is further processed by other task agents of the set of task agents 608. This enables the physical sensor 604 to provide the system 600 with an accurate and continuous representation of the observable physical parameters governing the operation of the target physical system 602.

[0115] The system 600 is configured to modify operational parameters of the target physical system 602 by integrating real-time physical data with predictive digital modeling. In this implementation, the physical sensor 604 determines a direct-variable measurement 606 of an operation of the target physical system 602. This direct-variable measurement 606 serves as an input to the set of task agents 608, wherein at least one task agent of the set of task agents 608 simulates at least a relevant portion of a digital twin of the system design. The simulation performed by the at least one task agent of the set of task agents 608 is informed by the direct-variable measurement 606 to generate a parameter-adjustment outcome prediction 612. The parameter-adjustment outcome prediction 612 predicts an outcome of adjusting a specific process parameter associated with the direct-variable measurement 606. Based on the parameter-adjustment outcome prediction 612, the executive agent 614 performs a parameter adjustment 616 of the operation of the target physical system 602. By utilizing the set of task agents 608 to evaluate the parameter-adjustment outcome prediction 612 prior to the physical execution of the parameter adjustment 616, the system 600 ensures that the parameter adjustment 616 maintains the target physical system 602 within a stable and optimized operational envelope.

[0116] As used herein, a digital twin comprises a high-fidelity virtual representation of a target physical system, such as the target physical system 602, that functions as asynchronized computational counterpart. The digital twin is maintained in a state of alignment with the real-world operational conditions of the target physical system 602 through the continuous ingestion of direct-variable measurements 606 provided by the physical sensor 604. This data synchronization ensures that the digital twin accurately mirrors the current physical and functional state of the target physical system 602 — such as internal dynamics, temperature gradients, or pressure levels — at any given time.

[0117] The digital twin enables the set of task agents 608 to perform predictive validations of proposed operational changes before any physical command is issued to the target physical system 602. Specifically, at least one task agent of the set of task agents 608 utilizes the digital twin to generate a parameter-adjustment outcome prediction 612, which forecasts how the target physical system 602 will respond to a proposed adjustment. This predictive mechanism allows the executive agent 614 to verify that the projected outcome adheres to the safety envelopes and operational boundary conditions defined by the engineering objective. By identifying issues ahead of time, such as potential mechanical strain, process instability, or violations of structural design limits within the virtual environment, the system 600 prevents the execution of parameter adjustments that could otherwise result in equipment damage, material degradation, or operational failure of the target physical system 602.

[0118] In some implementations, the system 600 is configured to monitor the operational health of the physical instrumentation associated with the target physical system 602 by detecting anomalies in one or more direct-variable measurements 606 determined by a physical sensor 604. For example, the system 600 determines, by the physical sensor 604, one or more direct-variable measurements 606 of an operation of the target physical system 602. The system 600 further simulates, by at least one task agent of the set of task agents 608, a portion of a digital twin of the system design, informed by the one or more direct-variable measurements 606, to determine an operational status of the physical sensor 604 relative to a predefined specification. The system 600 then determines, by the executive agent 614, based on the operational status, whether the physical sensor 604 is performing within the predefined specification.

[0119] As used herein, the operational status of the physical sensor 604 refers to a realtime characterization of the functional integrity and data reliability of the physical sensor 604 as determined by the simulation of the digital twin by at least one task agent of the set of task agents 608. This operational status may, for example, classify the physical sensor 604 into states such as nominal, degraded, or failed based on its ability to provide accurate direct-variable measurements 606. A nominal operational status indicates that the physical sensor 604 is functioning within expected parameters, whereas a degraded or failed operational status indicates that the data generated by the physical sensor 604 may be compromised by hardware wear, environmental interference, or electronic malfunction.

[0120] As used herein, the predefined specification represents a technical baseline or a set of performance requirements that the physical sensor 604 must satisfy to be considered reliable for use in controlling of the target physical system 602 by the system 600. This predefined specification may include metrics such as an allowable error margin, a specific calibration curve, a signal -to-noise ratio threshold, or a maximum response latency. For example, if the target physical system 602 is a high-pressure chemical reactor, the predefined specification for a pressure-based physical sensor 604 may require a precision of plus or minus 0.1 percent of the full-scale reading and a response time of less than ten milliseconds. These specifications may be stored within the system 600 and may be utilized by the digital twin to evaluate the validity of incoming direct-variable measurements 606.

[0121] As used herein, anomalies represent detected deviations wherein the direct-variable measurements 606 obtained by the physical sensor 604 do not align with the state of the target physical system 602 as predicted by the digital twin. One example of an anomaly is sensor drift, where the physical sensor 604 provides direct-variable measurements 606 that slowly diverge from the true physical state over time due to material degradation or environmental fouling. Another example is a “stuck-af ’ fault, wherein the physical sensor 604 continues to output a constant direct-variable measurement 606 even as the digital twin simulation indicates that the physical state of the target physical system 602 is changing. Furthermore, an anomaly may include the detection of excessive signal noise or intermittent dropouts that exceed the limits of the predefined specification.

[0122] In some implementations, at least one task agent of the set of task agents 608 may implement a quantum-enhanced GAN ensemble to identify these anomalies. By utilizing quantum-circuit-based distributions to model the latent space of the target physical system 602, these specialized task agents can detect subtle anomalies — such as minute shifts in signal distribution — that might be indistinguishable from normal operating noise for classical diagnostic systems. When the executive agent 614 receives an operational status indicating such an anomaly, the executive agent 614 determines that the physical sensor 604 is no longer performing within the predefined specification and may trigger a refinement signal to adjust the set of tasks, such as the set of tasks 506 of FIG. 5, or to transition the target physical system 602 to a safe operational mode by effectuating one or more parameter adjustments616.

[0123] FIG. 7 shows a diagram of an example of a system 700 by which the aspects, features, and elements disclosed herein may be implemented, such as simulating a measurement of an indirect variable and then adjusting an associated parameter in a target physical system 702 accordingly. The system 700 may comprise or be comprised in the system 500 of FIG. 5. The system 700 comprises a set of specialized agents that includes a set of task agents 708, which may be the set of task agents 508 of FIG. 5, and an executive agent 714, which may be the executive agent 514 of FIG. 5.

[0124] The target physical system 702 represents the physical environment, industrial process, or mechanical entity to which a system design, such as the system design 516 of FIG. 5, may be applied, and which may be further subject to the control, monitoring, and optimization of the system 700. Examples of the target physical system 702 include a chemical reactor, a biomanufacturing facility, an aerospace vehicle, an industrial manufacturing plant, or a power distribution grid. The target physical system 702 may be characterized by complex, often non-linear dynamics and a set of operational variables — such as temperature, pressure, flow rate, or chemical concentration — that define the state of the target physical system 702. The target physical system 702 may include a plurality of hardware components and control interfaces that allow the executive agent 714 to adjust operational parameters in order to satisfy the high-level requirements of an engineering objective, such as the engineering objective 502 of FIG. 5.

[0125] In some implementations, the system 700 is configured to manage the target physical system 702 based on latent or unobservable parameters that cannot be captured through direct physical instrumentation, e.g., by a physical sensor 704. For example, the system 700 may determine, by a physical sensor 704, which may be the physical sensor 604 of FIG. 6, a direct-variable measurement 706 of an operation of the target physical system 702. The system 700 further generates, by a virtual sensor of the set of specialized agents, an indirect-variable measurement 710 that is informed by the direct-variable measurement 706. The system 700 then simulates, by at least one task agent of the set of task agents 708, a portion of a digital twin of the system design, such as the system design 516 of FIG. 5, informed by the indirect-variable measurement 710, to predict a parameter-adjustment outcome prediction 712 of adjusting a process parameter associated with the indirect-variable measurement 710. Based on the parameter-adjustment outcome prediction 712, the executive agent 714 performs a parameter adjustment 716 of the operation of the target physical system 702.

[0126] As used herein, an indirect-variable measurement 710, also referred to as a virtualsensor measurement or a soft-sensor measurement, represents a calculated or inferred estimate of a physical or chemical state that is not directly measured (or not directly measurable within time or cost constraints) by the physical sensor 704. The generation of the indirect-variable measurement 710 allows the system 700 to gain situational awareness of critical variables that are otherwise hidden. For example, in a biomanufacturing process within the target physical system 702, the physical sensor 704 may measure direct variables such as temperature or pH level, while the virtual sensor, e.g., one or more task agents of the set of task agents 708, utilizes those direct-variable measurements 706 to infer an indirect-variable measurement 710 such as biomass concentration, nutrient depletion rate, or metabolic activity.

[0127] The use of the indirect-variable measurement 710 to inform the digital twin ensures that the parameter-adjustment outcome prediction 712 accounts for the underlying state of the target physical system 702. As used herein, the parameter-adjustment outcome prediction 712 refers to a simulated forecast of the response of the target physical system 702 to a proposed change in operational state as determined by the simulation of the digital twin. By simulating the response of the target physical system 702 relative to the inferred indirect variable, the digital twin can predict how a change in a process parameter — such as a feed rate, agitation speed, or reactant concentration — will impact the latent dynamics of the process. This enables the executive agent 714 to execute the parameter adjustment 716 based on a validated understanding of the internal state of the target physical system 702, thereby maintaining the process within an optimized operational envelope even when direct measurement of the process parameter is unavailable or impractical.

[0128] In some implementations, the generation of the indirect-variable measurement 710 or the simulation of the digital twin may be performed by a quantum-enhanced task agent within the set of task agents 708. This specialized task agent may utilize a quantum-circuitbased distribution or a hybrid quantum -cl as si cal architecture to model the complex, nonlinear correlations between the direct-variable measurement 706 and the latent state of the target physical system 702. By leveraging quantum-accelerated inference, the set of task agents 708 can produce highly accurate indirect-variable measurements 710 and parameteradjustment outcome predictions 712 in real-time, allowing for more precise control of the target physical system 702 by the executive agent 714 without requiring physical modifications to the hardware of the physical sensor 704.

[0129] In some implementations, the system 700 is configured to perform self-validationand iterative refinement of the virtual sensors used to monitor the target physical system 702. For example, the system 700 determines, by a set of physical sensors 704, a set of direct-variable measurements 706 of an operation of the target physical system 702. A virtual sensor of the set of specialized agents, such as a given task agent of the set of task agents 708, determines an indirect-variable measurement 710 of the operation of the target physical system 702. This indirect-variable measurement 710 is informed by at least one first measurement of the set of direct-variable measurements 706. To ensure the technical integrity of the model of the virtual sensor, the executive agent 714 determines an accuracy of the indirect-variable measurement 710 based on at least one second measurement of the set of direct-variable measurements 706, where the at least one second measurement is different from the at least one first measurement used to inform the virtual sensor. Based on this accuracy, the executive agent 714 modifies an implementation of the virtual sensor to improve the performance of the set of task agents 708.

[0130] As used herein, the accuracy of the indirect-variable measurement 710 refers to a quantitative assessment of the degree of agreement or correspondence between the value of the indirect-variable measurement 710 and the true physical condition of the target physical system 702 as cross-referenced against independent physical data. By utilizing a second measurement that was excluded from the initial calculation of the virtual sensor, the executive agent 714 can perform a validation of the predictive model of the virtual sensor against a different physical indicator. For instance, if a virtual sensor estimates a chemical concentration (the indirect variable) within the target physical system 702 based on a first direct-variable measurement 706, such as temperature, the executive agent 714 may validate that estimate against a second direct-variable measurement 706, such as a downstream pressure reading or a flow rate, that is physically correlated to the concentration but was not used by the virtual sensor to generate the initial estimate.

[0131] Modifying the implementation of the virtual sensor may involve the executive agent 714 adjusting the underlying logic, model parameters, or algorithmic weightings utilized by the virtual sensor. If the executive agent 714 determines that the accuracy falls below a predefined specification, the executive agent 714 may trigger a recalibration of the virtual sensor or instruct the virtual sensor to adopt a different computational paradigm, such as transitioning from a mechanistic model to a generative artificial intelligence model. In some implementations, this modification may involve the executive agent 714 communicating with an initiator agent, such as the initiator agent 504 of FIG. 5, to identify and select a replacement virtual sensor from a registry that possesses a capability descriptorbetter suited to the current operational dynamics of the target physical system 702.

[0132] In some implementations, the modification of the virtual sensor implementation may include the utilization of quantum-enhanced task agents within the set of task agents 708. If the executive agent 714 determines that classical modeling techniques are insufficient to maintain a required accuracy level due to high-dimensional interactions within the target physical system 702, the executive agent 714 may modify the implementation to utilize quantum-circuit-based distributions. By leveraging hybrid quantum-classical architectures or time-bin photonic interferometers, the system 700 can more accurately model the latent spaces and non-linear correlations of the target physical system 702. This ensures that the indirect-variable measurement 710 remains a reliable basis for the parameter-adjustment outcome prediction 712 and the subsequent parameter adjustment 716.

[0133] FIG. 8 shows a flow diagram illustrating an example of a technique 800 for adaptive agent-oriented system control for physical system design and operation. The technique 800 may be implemented by a system including some or all of the systems, components, or aspects depicted in FIGS. 1-7.

[0134] The step 802 comprises: decomposing, by an initiator agent of a set of specialized agents, an engineering objective for a target physical system into a set of tasks representing discrete mathematical or physical constraints. The initiator agent may be the initiator agent 504 of FIG. 5. The engineering objective may be the engineering objective 502 of FIG. 5. The set of tasks may be the set of tasks 506 of FIG. 5.

[0135] The step 804 comprises: identifying, by the initiator agent, a set of task agents of the set of specialized agents configured to perform the set of tasks. The set of task agents may be the set of task agents 508 of FIG. 5.

[0136] The step 806 comprises: instructing, by the initiator agent, the set of task agents to perform the set of tasks to produce a set of outputs, wherein the set of task agents is configured to communicate with one another independently of the initiator agent for resolving task dependencies. The set of outputs may be the set of outputs 512 of FIG. 5.

[0137] The step 808 comprises: generating, by an executive agent of the set of specialized agents, a system design for the target physical system based on an integration of the set of outputs. The executive agent may be the executive agent 514 of FIG. 5. The system design may be the system design 516 of FIG. 5.

[0138] The above-described techniques can be implemented as a method, a system, and a non-transitory computer-readable medium, for example, as described below.

[0139] In an example implementation as a method, the method comprises: decomposing,by an initiator agent of a set of specialized agents, an engineering objective for a target physical system into a set of tasks representing discrete mathematical or physical constraints; identifying, by the initiator agent, a set of task agents of the set of specialized agents configured to perform the set of tasks; instructing, by the initiator agent, the set of task agents to perform the set of tasks to produce a set of outputs, wherein the set of task agents is configured to communicate with one another independently of the initiator agent for resolving task dependencies; and generating, by an executive agent of the set of specialized agents, a system design for the target physical system based on an integration of the set of outputs.

[0140] In some implementations, the method further comprises: performing, by the set of task agents, the set of tasks to produce the set of outputs.

[0141] In some implementations, the method further comprises: integrating, by the executive agent, the set of outputs to produce the integration.

[0142] In some implementations, the set of specialized agents is arranged in a matrix topology, and resolving the task dependencies comprises cyclical data exchange between at least two task agents of the set of task agents.

[0143] In some implementations, the set of task agents is further configured to communicate with one another using a publish-subscribe architecture to resolve the task dependencies.

[0144] In some implementations, at least one of the discrete mathematical or physical constraints is defined by a mechanistic model representing a physical or chemical law or heuristic.

[0145] In some implementations, identifying the set of task agents comprises: broadcasting a query, to a network of candidate agents, indicating a respective task; receiving at least one response from at least one candidate agent; and selecting, based on the at least one response, at least one candidate agent of the network whose response indicates a match for the respective task.

[0146] In some implementations, identifying the set of task agents comprises: querying a registry with a respective task; receiving a response from the registry; and selecting, based on the response, at least one candidate agent of the registry whose registered characterization matches the respective task.

[0147] In some implementations, the method further comprises: decomposing the engineering objective comprises: parsing, by a generative artificial intelligence model, the engineering objective into the set of tasks.

[0148] In some implementations, the set of task agents comprises a first task agent and a second task agent both configured to perform a same task of the set of tasks using different implementations, the method further comprising: determining, by the executive agent, that a first output of the first task agent more closely satisfies a predetermined constraint than does a second output of the second task agent; and discarding, by the executive agent, the second output from the set of outputs.

[0149] In some implementations, the set of task agents comprises a first task agent and a second task agent both configured to perform a same task of the set of tasks using different implementations, the method further comprising: determining, by the executive agent, that a third output, of a third task agent, that depends alternatively from a first output of the first task agent or a second output of the second task agent, more closely satisfies a predetermined constraint when the third output is based on the first output; and discarding, by the executive agent, the second output from the set of outputs.

[0150] In some implementations, the method further comprises: identifying, by the executive agent, a design conflict between a first output of a first task agent of the set of task agents and a second output of a second task agent of the set of task agents; calculating, by the executive agent, a trade-off between the first output and the second output; and instructing, by the initiator agent, the first task agent and the second task agent to re-perform their respective tasks based on the trade-off.

[0151] In some implementations, the method further comprises: determining, by the executive agent, a failure of the integration of the set of outputs to satisfy a global system constraint; determining, by the initiator agent, a modified engineering objective to account for the failure; and re-performing the decomposing, identifying, instructing, and generating steps based on the modified engineering objective to generate the system design.

[0152] In some implementations, generating the system design comprises: simulating, by at least one task agent of the set of task agents, a digital twin of a candidate system design for the target physical system based on the integration.

[0153] In some implementations, the method further comprises: determining, by a physical sensor, a measurement of a direct variable of an operation of the target physical system; simulating, by at least one task agent of the set of task agents, a portion of a digital twin of the system design, informed by the measurement of the direct variable, to predict an outcome of adjusting a process parameter associated with the direct variable; and adjusting, by the executive agent based on the outcome, a process parameter of the operation of the target physical system.

[0154] In some implementations, the method further comprises: determining, by a physical sensor, a measurement of a direct variable of an operation of the target physical system; simulating, by at least one task agent of the set of task agents, a portion of a digital twin of the system design, informed by the measurement of the direct variable, to determine an operational status of the physical sensor relative to a predefined specification; and determining, by the executive agent based on the operational status, that the physical sensor is not performing within the predefined specification.

[0155] In some implementations, the method further comprises: determining, by a physical sensor, a measurement of a direct variable of an operation of the target physical system; determining, by a virtual sensor of the set of specialized agents, a measurement of an indirect variable of the operation of the target physical system, wherein the measurement of the indirect variable is informed by the measurement of the direct variable; simulating, by at least one task agent of the set of task agents, a portion of a digital twin of the system design, informed by the measurement of the indirect variable, to predict an outcome of adjusting a process parameter associated with the indirect variable; and adjusting, by the executive agent based on the outcome, a process parameter of the operation of the target physical system.

[0156] In some implementations, the method further comprises: determining, by a set of physical sensors, a set of measurements of a set of direct variables of an operation of the target physical system; determining, by a virtual sensor of the set of specialized agents, a measurement of an indirect variable of the operation of the target physical system, wherein the measurement of the indirect variable is informed by at least one first measurement of the set of measurements of direct variables; determining, by the executive agent, an accuracy of the measurement of the indirect variable based on at least one second measurement of the set of measurements of the direct variables, wherein the at least one second measurement is different from the at least one first measurement; and modifying, by the executive agent, an implementation of the virtual sensor based on the accuracy.

[0157] In another example implementation as a non-transitory computer-readable medium, the non-transitory computer-readable medium stores instructions operable to cause one or more processors to perform operations comprising: decomposing, by an initiator agent of a set of specialized agents, an engineering objective for a target physical system into a set of tasks representing discrete mathematical or physical constraints; identifying, by the initiator agent, a set of task agents of the set of specialized agents configured to perform the set of tasks; instructing, by the initiator agent, the set of task agents to perform the set of tasks to produce a set of outputs, wherein the set of task agents is configured to communicate withone another independently of the initiator agent for resolving task dependencies; and generating, by an executive agent of the set of specialized agents, a system design for the target physical system based on an integration of the set of outputs.

[0158] In another example implementation as a system, the system comprises one or more memories; and one or more processors configured to execute instructions stored in the one or more memories to: decompose, by an initiator agent of a set of specialized agents, an engineering objective for a target physical system into a set of tasks representing discrete mathematical or physical constraints; identify, by the initiator agent, a set of task agents of the set of specialized agents configured to perform the set of tasks; instruct, by the initiator agent, the set of task agents to perform the set of tasks to produce a set of outputs, wherein the set of task agents is configured to communicate with one another independently of the initiator agent for resolving task dependencies; and generate, by an executive agent of the set of specialized agents, a system design for the target physical system based on an integration of the set of outputs.

[0159] In some implementations, the methods and systems described herein comprise an anomaly-detection system configured to autonomously monitor the manufacturing process and identify deviations from expected operational behavior. For example, as biomanufacturing and chemical manufacturing processes scale, they become increasingly susceptible to multifaceted disturbances, such as sensor drift, biological contamination, or mechanical degradation. To address this, the described anomaly-detection system leverages an architecture comprising at least one generator and at least one discriminator. By integrating this anomaly-detection system with one or more virtual sensors and synthetic data generation modules, an overall control system of a target physical system achieves a highly responsive, autonomous optimization loop capable of distinguishing between transient process noise and genuine systemic anomalies.

[0160] In some implementations, the anomaly-detection system may comprise diverse neural network architectures tailored to the specific dynamics of the manufacturing process. The at least one generator and the at least one discriminator may be implemented utilizing GANs or associative adversarial networks (AANs). Additionally, autoencoders may be employed to compress and reconstruct operational data to establish a baseline of normal operations. Depending on the specific nature of the at least one measurement, these networks may incorporate sequential deep-learning models, such as long short-term memory (LSTM) networks or gated recurrent units (GRUs). Sequential models may capture the time-dependent behaviors and complex temporal dependencies inherent in batch or continuous manufacturingprocesses. For spatially structured data or complex multivariate inputs, convolutional neural networks (CNNs) may also be integrated to extract hierarchical features from data.

[0161] According to certain aspects disclosed herein, at least one of the generator and the discriminator is trained on an operational dataset. The operation dataset may, for example, be supervised, unsupervised, or semi-supervised, or a combination of any such labeling methods. The operation dataset may comprise vast quantities of historical, real-time, and synthetic data representing the “normal” or baseline behavior of the target physical system under optimal or acceptable conditions. Because anomalies in complex manufacturing environments are often rare, unpredictable, or entirely novel, supervised learning approaches relying on labeled anomaly data are frequently insufficient. Instead, the anomaly-detection system utilizes unsupervised learning to map the multi-dimensional manifold of normal system behavior. Anomalies (whether singular, isolated events or a plurality of compounding deviations) are subsequently detected based on their statistical or structural divergence from this learned normal behavior.

[0162] In some implementations, the neural networks (e.g., the generator and discriminator) may be configured to minimize false positive anomaly alerts. False positives are a significant challenge in large-scale manufacturing, often leading to unnecessary process shutdowns or operator fatigue. To mitigate this, the generator and discriminator may work in concert within an adversarial or reconstructive framework. For example, when an autoencoder or generator is trained to reconstruct the normal operational data, real-time measurements are fed into the system, and the generator attempts to reconstruct the data based on its learned understanding of normal behavior. The discriminator then evaluates the difference (e.g., the reconstruction error) between the actual real-time measurements and the generator's reconstruction. By utilizing an adversarial training process, the discriminator can identify true anomalies while the generator becomes robust at modeling the variance of normal operations, thereby reducing the rate of false positives.

[0163] The architecture of the anomaly-detection system may be implemented using various computational modalities, including purely classical, purely quantum, or hybrid classical-quantum frameworks, depending on the complexity of the process and available computational resources. In a classical -classical implementation, both the data distributions and the machine learning models (the generator and discriminator) are executed on classical computing hardware. This approach utilizes deep learning frameworks to process the operation dataset and execute the anomaly detection.

[0164] In certain hybrid implementations, the anomaly-detection system may partitiontasks between classical and quantum resources. For instance, in some implementations, quantum hardware or quantum-inspired tensor networks may be utilized to model the probability distributions of the underlying manufacturing process of a target physical system. The outputs of these quantum distributions are then fed into classical neural networks, which serve as the generator and discriminator to perform the final classification and anomaly scoring. This enables the system to capture process correlations that classical systems might miss, while leveraging the speed and reliability of classical hardware for the adversarial machine learning tasks.

[0165] Alternatively, classical systems may be used to manage, store, and provide the operation dataset and its underlying distributions, while the machine learning components (the generator and / or discriminator) are implemented using quantum computing architectures. In this configuration, quantum neural networks or variational quantum circuits are trained to discriminate between normal and anomalous data; the quantum discriminator can evaluate the classical distributions across a vastly expanded parameter space, enabling the detection of subtle, multi-variable anomalies that would be computationally intractable for classical neural networks to identify.

[0166] To train the anomaly-detection system, some implementations involve generating a benchmark dataset of time-series process data. This dataset may be produced by utilizing a high-fidelity dynamic simulation model of an integrated manufacturing system, which maps both upstream processes (such as cell cultivation and fermentation) and downstream processing operations (such as centrifugation, filtration, and chromatography). The simulation model may be configured to output time-series data spanning both normal operating regimes and anomalous operating regimes. By simulating these conditions in silico, the system can synthesize a high-volume dataset that captures the complex, non-linear dynamics and spatiotemporal gradients of the manufacturing process without requiring costly, risky, or disruptive physical experiments.

[0167] The system may train an ensemble of GANs. In some implementations, this ensemble may be trained exclusively on the time-series data corresponding to the normal operating regime. Each GAN within the ensemble comprises a generator neural network and a discriminator neural network. During training, the generator may receive a latent vector sampled specifically from a quantum probability distribution. Utilizing quantum hardware or quantum-inspired tensor networks to sample this latent space allows the generator to access a highly complex, multidimensional parameter space that classical probability distributions cannot efficiently replicate. The generator may map this quantum-sampled latent vector tosynthetic process data, while the discriminator is trained to distinguish between this generated synthetic data and the actual normal process data from the benchmark dataset. Because the ensemble is exposed solely to normal operational data, the discriminators become hyperspecialized in recognizing the precise statistical boundaries of optimal system behavior.

[0168] The trained discriminators of the GAN ensemble may be isolated and stored for deployment within the active control architecture. During the real-time operational phase, the ensemble of discriminators may be applied continuously to an incoming time-series data stream acquired from the live manufacturing process. Each discriminator can independently process the live input stream and compute an anomaly score representing the input's deviation from its uniquely learned distribution of normal operation. The anomaly-detection system monitors these concurrent evaluations and is configured to label the input stream as anomalous if at least one discriminator in the ensemble scores the stream outside the established bounds of normal operation. This multi-discriminator, ensemble-based scoring mechanism can provide sensitivity to subtle, multivariate process deviations, allowing the system to rapidly identify novel fault conditions and proactively optimize the manufacturing environment.

[0169] Various physical embodiments of a quantum computer may be suitable for use according to the implementations described herein. In general, the fundamental data storage unit in quantum computing is the quantum bit, or qubit. The qubit is a quantum-computing analog of a classical digital computer system bit. A classical bit is considered to occupy, at any given point in time, one of two possible states corresponding to the binary digits (bits) 0 or 1. By contrast, a qubit is implemented in hardware by a physical medium with quantummechanical characteristics. Such a medium, which physically instantiates a qubit, may be referred to herein as a “physical instantiation of a qubit,” a “physical embodiment of a qubit,” a “medium embodying a qubit,” or similar terms, or simply as a “qubit,” for ease of explanation. It should be understood, therefore, that references herein to “qubits” within descriptions of implementations described herein refer to physical media that embody qubits.

[0170] Each qubit has an infinite number of different potential quantum-mechanical states. When the state of a qubit is physically measured, the measurement produces one of two different basis states resolved from the state of the qubit. Thus, a single qubit can represent a one, a zero, or any quantum superposition of those two qubit states; a pair of qubits can be in any quantum superposition of four orthogonal basis states; and three qubits can be in any superposition of eight orthogonal basis states. The function that defines the quantum-mechanical states of a qubit is known as its wavefunction. The wavefunction alsospecifies the probability distribution of outcomes for a given measurement. A qubit, which has a quantum state of dimension two (i.e., has two orthogonal basis states), may be generalized to a d-dimensional “qudit,” where d may be any integral value, such as 2, 3, 4, or higher. In the general case of a qudit, measurement of the qudit produces one of d different basis states resolved from the state of the qudit. Any reference herein to a qubit should be understood to refer more generally to a d-dimensional qudit with any value of d.

[0171] Although certain descriptions of qubits herein may describe such qubits in terms of their mathematical properties, each such qubit may be implemented in a physical medium in any of a variety of different ways. Examples of such physical media include superconducting material, trapped ions, photons, optical cavities, individual electrons trapped within quantum dots, point defects in solids (e.g., phosphorus donors in silicon or nitrogenvacancy centers in diamond), molecules (e.g., alanine, vanadium complexes), or aggregations of any of the foregoing that exhibit qubit behavior, that is, comprising quantum states and transitions therebetween that can be controllably induced or detected.

[0172] For any given medium that implements a qubit, any of a variety of properties of that medium may be chosen to implement the qubit. For example, if electrons are chosen to implement qubits, then the x component of its spin degree of freedom may be chosen as the property of such electrons to represent the states of such qubits. Alternatively, the y component, or the z component of the spin degree of freedom may be chosen as the property of such electrons to represent the state of such qubits. This is merely a specific example of the general feature that for any physical medium that is chosen to implement qubits, there may be multiple physical degrees of freedom (e.g., the x, y, and z components in the electron spin example) that may be chosen to represent 0 and 1. For any particular degree of freedom, the physical medium may be controllably put in a state of superposition, and measurements may then be taken in the chosen degree of freedom to obtain readouts of qubit values.

[0173] Certain implementations of quantum computers, referred to as gate model quantum computers, comprise quantum gates. In contrast to classical gates, there is an infinite number of possible single-qubit quantum gates that change the state vector of a qubit.Changing the state of a qubit state vector typically is referred to as a single-qubit rotation, and may also be referred to herein as a state change or a single-qubit quantum-gate operation. A rotation, state change, or single-qubit quantum-gate operation may be represented mathematically by a unitary 2-by-2 matrix with complex elements. A rotation corresponds to a rotation of a qubit state within its Hilbert space, which may be conceptualized as a rotation of the Bloch sphere. (As is well-known to those having ordinary skill in the art, the Blochsphere is a geometrical representation of the space of pure states of a qubit.) Multi-qubit gates alter the quantum state of a set of qubits. For example, two-qubit gates rotate the state of two qubits as a rotation in the four-dimensional Hilbert space of the two qubits. (As is well-known to those having ordinary skill in the art, a Hilbert space is an abstract vector space possessing the structure of an inner product that allows length and angle to be measured. Furthermore, Hilbert spaces are complete: there are enough limits in the space to allow the techniques of calculus to be used.)

[0174] A quantum circuit may be specified as a sequence of quantum gates. As described in more detail below, the term “quantum gate,” as used herein, refers to the application of a gate control signal (defined below) to one or more qubits to cause those qubits to undergo certain physical transformations and thereby to implement a logical gate operation. To conceptualize a quantum circuit, the matrices corresponding to the component quantum gates may be multiplied together in the order specified by the gate sequence to produce a 2n-by-2n complex matrix representing the same overall state change on n qubits. A quantum circuit may thus be expressed as a single resultant operator. However, designing a quantum circuit in terms of constituent gates allows the design to conform to a standard set of gates, and thus enables greater ease of deployment. A quantum circuit thus corresponds to a design for actions taken upon the physical components of a quantum computer.

[0175] A given variational quantum circuit may be parameterized in a suitable devicespecific manner. More generally, the quantum gates making up a quantum circuit may have an associated plurality of tuning parameters. For example, in implementations based on optical switching, tuning parameters may correspond to the angles of individual optical elements.

[0176] In some implementations of quantum circuits, the quantum circuit includes both one or more gates and one or more measurement operations. Quantum computers implemented using such quantum circuits are referred to herein as implementing “measurement feedback.” For example, a quantum computer implementing measurement feedback may execute the gates in a quantum circuit and then measure only a subset (i.e., fewer than all) of the qubits in the quantum computer, and then decide which gate(s) to execute next based on the outcome(s) of the measurement s). In particular, the measurement(s) may indicate a degree of error in the gate operation(s), and the quantum computer may decide which gate(s) to execute next based on the degree of error. The quantum computer may then execute the gate(s) indicated by the decision. This process of executing gates, measuring a subset of the qubits, and then deciding which gate(s) to executenext may be repeated any number of times. Measurement feedback may be useful for performing quantum error correction, but is not limited to use in performing quantum error correction. For every quantum circuit, there is an error-corrected implementation of the circuit with or without measurement feedback.

[0177] Some implementations described herein generate, measure, or utilize quantum states that approximate a target quantum state (e.g., a ground state of a Hamiltonian). As will be appreciated by those trained in the art, there are many ways to quantify how well a first quantum state “approximates” a second quantum state. In the following description, any concept or definition of approximation known in the art may be used without departing from the scope hereof. For example, when the first and second quantum states are represented as first and second vectors, respectively, the first quantum state approximates the second quantum state when an inner product between the first and second vectors (called the “fidelity” between the two quantum states) is greater than a predefined amount (typically labeled c). In this example, the fidelity quantifies how “close” or “similar” the first and second quantum states are to each other. The fidelity represents a probability that a measurement of the first quantum state will give the same result as if the measurement were performed on the second quantum state. Proximity between quantum states can also be quantified with a distance measure, such as a Euclidean norm, a Hamming distance, or another type of norm known in the art. Proximity between quantum states can also be defined in computational terms. For example, the first quantum state approximates the second quantum state when a polynomial time-sampling of the first quantum state gives some desired information or property that it shares with the second quantum state.

[0178] Not all quantum computers are gate model quantum computers. Implementations of the techniques described herein are not limited to being implemented using gate model quantum computers. As an alternative example, implementations of the techniques described herein may be implemented, in whole or in part, using a quantum computer that is implemented using a quantum annealing architecture, which is an alternative to the gate model quantum computing architecture. More specifically, quantum annealing is a metaheuristic for finding the global minimum of a given objective function over a given set of candidate solutions (candidate states), by a process using quantum fluctuations.

[0179] As yet another alternative example, implementations of the techniques described herein may be implemented, in whole or in part, using a quantum computer that is implemented using a one-way quantum computing architecture, also referred to as a measurement-based quantum computing architecture, which is another alternative to the gatemodel quantum computing architecture. More specifically, the one-way or measurementbased quantum computer (MBQC) is a method of quantum computing that first prepares an entangled resource state, usually a cluster state or graph state, then performs single qubit measurements on it. It is “one-way” because the resource state is destroyed by the measurements.

[0180] The outcome of each individual measurement is random, but they are related in such a way that the computation always succeeds. In general, the choices of basis for later measurements need to depend on the results of earlier measurements, and hence the measurements cannot all be performed at the same time.

[0181] Any of the functions disclosed herein may be implemented using means for performing those functions. Such means include, but are not limited to, any of the components disclosed herein, such as the computer-related components described herein.

[0182] The foregoing disclosure provides illustration and description but is not intended to be exhaustive or to limit the aspects to the precise forms disclosed. Modifications and variations may be made in light of the above disclosure or may be acquired from practice of the aspects. As used herein, the term “component” is intended to be broadly construed as hardware or a combination of hardware and at least one of software or firmware. “Software” shall be construed broadly to mean instructions, instruction sets, code, code segments, program code, programs, subprograms, software modules, applications, software applications, software packages, routines, subroutines, objects, executables, threads of execution, procedures, or functions, among other examples, whether referred to as software, firmware, middleware, microcode, hardware description language, or otherwise. As used herein, a “processor” is implemented in hardware or a combination of hardware and software. It will be apparent that systems or methods described herein may be implemented in different forms of hardware or a combination of hardware and software. The actual specialized control hardware or software code used to implement these systems or methods is not limiting of the aspects. Thus, the operation and behavior of the systems or methods are described herein without reference to specific software code, because those skilled in the art will understand that software and hardware can be designed to implement the systems or methods based, at least in part, on the description herein.

[0183] As used herein, the terminology “instructions” may include directions or expressions for performing any technique, or any portion or portions thereof, disclosed herein, and may be realized in hardware, software, or any combination thereof. For example, instructions may be implemented as information, such as a computer program, stored inmemory that may be executed by a processor to perform any of the respective methods, algorithms, aspects, techniques, or combinations thereof, as described herein. Instructions, or a portion thereof, may be implemented as a special-purpose processor, or circuitry, that may include specialized hardware for carrying out any of the techniques, algorithms, aspects, or combinations thereof, as described herein. In some implementations, portions of the instructions may be distributed across multiple processors on a single device, on multiple devices, which may communicate directly or across a network such as a local area network, a wide area network, the Internet, or a combination thereof.

[0184] As used herein, the terminology “example,” “embodiment,” “implementation,” “aspect,” “feature,” or “element” indicates serving as an example, instance, or illustration. Unless expressly indicated, any example, embodiment, implementation, aspect, feature, or element is independent of each other example, embodiment, implementation, aspect, feature, or element and may be used in combination with any other example, embodiment, implementation, aspect, feature, or element.

[0185] As used herein, the terminology “determine” and “identify,” or any variations thereof, includes selecting, ascertaining, computing, looking up, receiving, determining, establishing, obtaining, or otherwise identifying or determining in any manner whatsoever using one or more of the devices shown and described herein. As used herein, “satisfying a threshold” may, depending on the context, refer to a value being greater than the threshold, greater than or equal to the threshold, less than the threshold, less than or equal to the threshold, equal to the threshold, or not equal to the threshold, among other examples.

[0186] As used herein, the terminology “or” is intended to mean an inclusive “or” rather than an exclusive or (“xor”) and may be used interchangeably with “and / or,” unless explicitly stated otherwise (for example, if used in combination with “either” or “only one of’), or clearly is used otherwise from context. In addition, the articles “a” and “an” as used in this application and the appended claims should generally be construed to mean “one or more” unless specified otherwise or clear from context to be directed to a singular form. Further, as used herein, the article “the” is intended to include one or more items referenced in connection with the article “the” and may be used interchangeably with “the one or more.” Furthermore, as used herein, the terms “set,” “subset,” “group,” and “subgroup” are intended to include one or more items, unless explicitly stated otherwise or indicated by context (e.g., an empty set), and may be used interchangeably with “one or more.” As used herein, a phrase referring to “at least one of’ a list of items refers to any combination of those items, including single members. As an example, “at least one of: a, b, or c” is intended to cover a, b, c, a + b,a + c, b + c, and a + b + c, as well as any combination with multiples of the same element (for example, a + a, a + a + a, a + a + b, a + a + c, a + b + b, a + c + c, b + b, b + b +b, b +b + c, c + c, and c + c + c, or any other ordering of a, b, and c).

[0187] Also, as used herein, the terms “has,” “have,” “having,” and similar terms are intended to be open-ended terms that do not limit an element that they modify (for example, an element “having” A may also have B). Further, the phrase “based on” is intended to mean “based on or otherwise in association with” unless explicitly stated otherwise. Accordingly, unless explicitly stated otherwise, the phrase “based on” is intended to mean “based at least in part on.”

[0188] Even though particular combinations of features are recited in the claims or disclosed in the specification, these combinations are not intended to limit the disclosure of various aspects. Many of these features may be combined in ways not specifically recited in the claims or disclosed in the specification. The disclosure of various aspects includes each dependent claim in combination with every other claim in the claim set. Further, for simplicity of explanation, although the figures and descriptions herein may include sequences or series of steps or stages, elements of the techniques disclosed herein may occur in various orders or concurrently. Additionally, elements of the techniques disclosed herein may occur with other elements not explicitly presented and described herein. Furthermore, not all elements of the techniques described herein may be required to implement a technique in accordance with this disclosure. Although aspects, features, and elements are described herein in particular combinations, each aspect, feature, or element may be used independently or in various combinations with or without other aspects, features, and elements.

[0189] The above-described aspects, examples, and implementations have been described in order to allow easy understanding of the disclosure and are not limiting. On the contrary, the disclosure covers various modifications and equivalent arrangements included within the scope of the appended claims, which scope is to be accorded the broadest interpretation so as to encompass all such modifications and equivalent structures as is permitted under the law.

Claims

What is claimed is:

1. A method, comprising:decomposing, by an initiator agent of a set of specialized agents, an engineering objective for a target physical system into a set of tasks representing discrete mathematical or physical constraints;identifying, by the initiator agent, a set of task agents of the set of specialized agents configured to perform the set of tasks;instructing, by the initiator agent, the set of task agents to perform the set of tasks to produce a set of outputs, wherein the set of task agents is configured to communicate with one another independently of the initiator agent for resolving task dependencies; andgenerating, by an executive agent of the set of specialized agents, a system design for the target physical system based on an integration of the set of outputs.

2. The method of claim 1, further comprising:performing, by the set of task agents, the set of tasks to produce the set of outputs.

3. The method of claim 1, further comprising:integrating, by the executive agent, the set of outputs to produce the integration.

4. The method of claim 1, wherein:the set of specialized agents is arranged in a matrix topology, andresolving the task dependencies comprises cyclical data exchange between at least two task agents of the set of task agents.

5. The method of claim 1, wherein:the set of task agents is further configured to communicate with one another using a publish-subscribe architecture to resolve the task dependencies.

6. The method of claim 1, wherein:at least one of the discrete mathematical or physical constraints is defined by a mechanistic model representing a physical or chemical law or heuristic.

7. The method of claim 1, wherein identifying the set of task agents comprises: broadcasting a query, to a network of candidate agents, indicating a respective task; receiving at least one response from at least one candidate agent; andselecting, based on the at least one response, at least one candidate agent of the network whose response indicates a match for the respective task.

8. The method of claim 1, wherein identifying the set of task agents comprises:querying a registry with a respective task;receiving a response from the registry; andselecting, based on the response, at least one candidate agent of the registry whose registered characterization matches the respective task.

9. The method of claim 1, wherein decomposing the engineering objective comprises:parsing, by a generative artificial intelligence model, the engineering objective into the set of tasks.

10. The method of claim 1, wherein the set of task agents comprises a first task agent and a second task agent both configured to perform a same task of the set of tasks using different implementations, the method further comprising:determining, by the executive agent, that a first output of the first task agent more closely satisfies a predetermined constraint than does a second output of the second task agent; anddiscarding, by the executive agent, the second output from the set of outputs.

11. The method of claim 1, wherein the set of task agents comprises a first task agent and a second task agent both configured to perform a same task of the set of tasks using different implementations, the method further comprising:determining, by the executive agent, that a third output, of a third task agent, that depends alternatively from a first output of the first task agent or a second output of the second task agent, more closely satisfies a predetermined constraint when the third output is based on the first output; anddiscarding, by the executive agent, the second output from the set of outputs.

12. The method of claim 1, further comprising:identifying, by the executive agent, a design conflict between a first output of a first task agent of the set of task agents and a second output of a second task agent of the set of task agents;calculating, by the executive agent, a trade-off between the first output and the second output; andinstructing, by the initiator agent, the first task agent and the second task agent to reperform their respective tasks based on the trade-off.

13. The method of claim 1, further comprising:determining, by the executive agent, a failure of the integration of the set of outputs to satisfy a global system constraint;determining, by the initiator agent, a modified engineering objective to account for the failure; andre-performing the decomposing, identifying, instructing, and generating steps based on the modified engineering objective to generate the system design.

14. The method of claim 1, wherein generating the system design comprises:simulating, by at least one task agent of the set of task agents, a digital twin of a candidate system design for the target physical system based on the integration.

15. The method of claim 1, further comprising:determining, by a physical sensor, a measurement of a direct variable of an operation of the target physical system;simulating, by at least one task agent of the set of task agents, a portion of a digital twin of the system design, informed by the measurement of the direct variable, to predict an outcome of adjusting a process parameter associated with the direct variable; and adjusting, by the executive agent based on the outcome, a process parameter of the operation of the target physical system.

16. The method of claim 1, further comprising:determining, by a physical sensor, a measurement of a direct variable of an operation of the target physical system;simulating, by at least one task agent of the set of task agents, a portion of a digital twin of the system design, informed by the measurement of the direct variable, todetermine an operational status of the physical sensor relative to a predefined specification; anddetermining, by the executive agent based on the operational status, that the physical sensor is not performing within the predefined specification.

17. The method of claim 1, further comprising:determining, by a physical sensor, a measurement of a direct variable of an operation of the target physical system;determining, by a virtual sensor of the set of specialized agents, a measurement of an indirect variable of the operation of the target physical system, wherein the measurement of the indirect variable is informed by the measurement of the direct variable;simulating, by at least one task agent of the set of task agents, a portion of a digital twin of the system design, informed by the measurement of the indirect variable, to predict an outcome of adjusting a process parameter associated with the indirect variable; and adjusting, by the executive agent based on the outcome, a process parameter of the operation of the target physical system.

18. The method of claim 1, further comprising:determining, by a set of physical sensors, a set of measurements of a set of direct variables of an operation of the target physical system;determining, by a virtual sensor of the set of specialized agents, a measurement of an indirect variable of the operation of the target physical system, wherein the measurement of the indirect variable is informed by at least one first measurement of the set of measurements of direct variables;determining, by the executive agent, an accuracy of the measurement of the indirect variable based on at least one second measurement of the set of measurements of the direct variables, wherein the at least one second measurement is different from the at least one first measurement; andmodifying, by the executive agent, an implementation of the virtual sensor based on the accuracy.

19. Anon-transitory computer-readable medium storing instructions operable to cause one or more processors to perform operations comprising:decomposing, by an initiator agent of a set of specialized agents, an engineering objective for a target physical system into a set of tasks representing discrete mathematical or physical constraints;identifying, by the initiator agent, a set of task agents of the set of specialized agents configured to perform the set of tasks;instructing, by the initiator agent, the set of task agents to perform the set of tasks to produce a set of outputs, wherein the set of task agents is configured to communicate with one another independently of the initiator agent for resolving task dependencies; andgenerating, by an executive agent of the set of specialized agents, a system design for the target physical system based on an integration of the set of outputs.

20. A system, comprising:one or more memories; andone or more processors configured to execute instructions stored in the one or more memories to:decompose, by an initiator agent of a set of specialized agents, an engineering objective for a target physical system into a set of tasks representing discrete mathematical or physical constraints;identify, by the initiator agent, a set of task agents of the set of specialized agents configured to perform the set of tasks;instruct, by the initiator agent, the set of task agents to perform the set of tasks to produce a set of outputs, wherein the set of task agents is configured to communicate with one another independently of the initiator agent for resolving task dependencies; andgenerate, by an executive agent of the set of specialized agents, a system design for the target physical system based on an integration of the set of outputs.

Images