A constant-axis bidirectional steady-state continuous calculation method

Abstract

The application discloses a kind of constant axis bidirectional steady-state continuous calculation method, belong to computing theory and computing method technical field.This method with natural physical field or high stable frequency source establishes four-dimensional composite space-time reference S (t) =[f0, φ (t), T (t), P];With the state variable of multiple channels is synchronized in the same clock cycle Lock-in generates state snapshot, converges and feeds back stream to form bidirectional closed loop;Based on snapshot estimation future deviation and injects constraint in advance, to prevent deviation maintains system steady state;For snapshot and output, uniform time stamp is stamped, after continuous domain parallel iteration operation, output spectrum energy is concentrated in continuous smooth result near fundamental wave.This method does not follow von neumann serial fetch-decode path, constitutes the physical domain continuous computing paradigm independent of binary instruction set, and can be widely used in physical domain real-time continuous control and high-performance parallel computing scene.

Description

Technical Field

[0001] This invention belongs to the technical field of computational theory and methods, specifically relating to a general computational method that departs from the binary serial instruction fetching and decoding logic of the von Neumann architecture and drives bidirectional closed-loop steady-state continuous computation with a four-dimensional composite spatiotemporal reference. This method is not bound to any specific hardware platform or application scenario, forming the foundation of a physical domain continuous computation paradigm independent of binary instruction sets. The constant-axis bidirectional steady-state continuous computation method disclosed in this invention is a fundamental general computational paradigm oriented towards the physical world, and is referred to as the "fundamental algorithm" in the commercial naming, technical promotion, and academic definition of this technical system. Background Technology

[0002] Existing computational methods, whether Turing machine models or von Neumann architectures, are based on discrete-time clock cycles and serial instruction execution logic. While this logic performs well in handling information domain tasks, it suffers from the following fundamental limitations when facing real-time continuous control tasks in the physical domain: First, discrete sampling and discrete output limit the control quality. Traditional calculation methods use fixed-period sampling and output, making system state changes completely invisible within the sampling interval. The output signal is a discrete square wave containing a large number of useless harmonic components. Second, the perception and execution link is unidirectional and serial. Following the time-sharing execution logic of "sampling-computation-output", the feedback data from the execution end can only be incorporated into the calculation process in the next sampling cycle. There is an inherent time difference between perception and execution, making it impossible to build a two-way closed loop. Third, there is a lack of a unified spatiotemporal reference. Different computing nodes run on independent clocks, and phase synchronization and timing alignment across nodes require additional communication overhead; Fourth, the calculation mode is passive error correction. Traditional methods correct errors only after they occur, and the system always operates in an oscillating cycle of "error-correction-overshoot-recorrection", which cannot actively maintain a steady state.

[0003] The aforementioned limitations are inherent technical defects in the hardware control layer when existing discrete computing architectures are applied to physically controlled systems. These limitations cannot be eliminated simply by optimizing mathematical algorithm models; the underlying computing paradigm must be reconstructed.

[0004] In summary, existing computational methods have inherent limitations when dealing with real-time continuous control tasks in the physical domain, and a completely new computational method is urgently needed. Summary of the Invention

[0005] Purpose of the invention This invention provides a two-way steady-state continuous computation method with a constant axis. It uses a natural physical field or a high-stability frequency source as a spatiotemporal reference to establish a four-dimensional composite spatiotemporal reference S(t)=[f0, φ(t), T(t), P]. Based on this, it realizes two-way closed-loop parallel interaction of perception and action, steady-state maintenance by pre-injection of prediction bias, and continuous domain operation with full-link timing alignment, thus forming a physical domain continuous computation paradigm independent of binary instruction sets.

[0006] Technical solution 1. Establishment of a constant-axis four-dimensional composite spacetime reference This method first establishes a four-dimensional composite spatiotemporal reference S(t)=[f0, φ(t), T(t), P], where f0 is the reference frequency, locked by a high-stability frequency source; φ(t) is the real-time phase, maintained by a phase-locked loop circuit; T(t) is the high-precision timing, integrating a precision time protocol and an external time signal; and P is the position vector of the controlled object or computing node in absolute space. This four-dimensional composite spatiotemporal reference provides a unified time index and spatial reference system for all calculation steps. All sensing data, control commands, and feedback signals are time-aligned and spatially calibrated under this reference. 2. Synchronous latching of the full state vector Under the unified clock cycle of a constant-axis spatiotemporal reference, all state variables from multiple sensor channels are synchronously latched within the same clock cycle, generating a complete state snapshot of the current moment. This state snapshot, with a unified timestamp and spatial coordinates, constitutes a complete state description of the controlled object or computing node in the physical world at the current moment. 3. Parallel closed-loop interaction of bidirectional data streams By merging the perceived data stream and the action feedback stream in parallel within the same clock cycle without time-division multiplexing, a two-way closed loop of perception and action is formed. Perceived data guides action output, and the results of the action are simultaneously fed back to the perception stage. The two are completely synchronized in time, eliminating the inherent time difference between perception and execution in traditional methods. 4. Steady-state maintenance based on prediction bias Based on a complete snapshot of the current state, a simplified predictive model is used to estimate the potential deviation of the system output from the setpoint over several future control cycles, and this predicted deviation is injected into the constraints of the control commands in advance. The optimization objective of the control commands is no longer to correct existing deviations, but to prevent future deviations from occurring, ensuring the system maintains steady-state operation throughout the process and avoiding the oscillating cycle of "deviation occurrence-correction-overshoot-recorrection" in traditional methods. 5. Continuous domain operations with end-to-end timing alignment Each set of state snapshots and control commands is stamped with a unified timestamp derived from the four-dimensional composite spatiotemporal reference S(t), ensuring that the data across the entire link is strictly aligned under the same time reference. Based on the unified time reference and prediction bias constraints, parallel iterative computation of model predictive control is performed in the continuous domain, outputting macroscopically continuous and smooth control commands.

[0007] Beneficial effects of the invention 1. This invention proposes for the first time a continuous computing paradigm that adapts to physically controlled systems and is free from von Neumann binary discrete serial operations, upgrading the computing method from "discrete sampling, discrete output, and passive error correction" to "continuous tracking, continuous output, and active prevention"; 2. This invention is the first to construct a four-dimensional composite spatiotemporal reference from a natural physical field or a highly stable frequency source, providing an absolute reference system for calculations and enabling distributed nodes to achieve synchronization without communication. 3. This invention is the first to achieve bidirectional closed-loop parallel interaction between perception and action at the computational method level, eliminating the inherent time difference; 4. This invention is the first to achieve steady-state maintenance based on prediction deviation at the computational method level, changing the control mode from corrective to preventive; 5. This method is a general calculation method that is not bound to any specific hardware platform, controlled object or application scenario, and can be widely used in various physical domain real-time continuous control scenarios. Detailed Implementation

[0008] Example 1: Motor control application based on a high-stability crystal oscillator reference This embodiment uses the real-time control of a permanent magnet synchronous motor as an example to illustrate the complete implementation process of this method; S1. Establishing a four-dimensional composite spatiotemporal reference: A high-stability crystal oscillator is used as the reference frequency source, outputting a reference frequency signal of f0 = 10MHz; the internal clock phase φ(t) of the controller is locked to this reference frequency through a phase-locked loop circuit; the global timing T(t) is obtained through the PTP precision time protocol; the rotor position is obtained through the motor encoder, and the spatial position vector P is determined in combination with the equipment installation position. The four-dimensional composite spatiotemporal reference S(t) = [f0, φ(t), T(t), P] is thus established. S2. Synchronous Latching of the Full State Vector: Under a uniform clock cycle at the reference frequency, the output data of the three-phase current sensor, rotor position encoder, speed sensor, temperature sensor, and vibration sensor are all latched into the register file at the same clock edge, generating a complete state snapshot of the current moment containing all state variables. This snapshot has a uniform timestamp T(t) and spatial coordinates P; S3. Parallel closed-loop interaction of bidirectional data streams: The above state snapshot (sensing data stream) and the actual motor response data after the execution of the previous control cycle (action feedback stream) are sent to the MPC calculation unit in parallel within the same clock cycle, without time-division serial processing, to form a bidirectional closed loop of sensing and action. S4. Steady-state maintenance based on prediction deviation: Based on the current state snapshot, a linear state equation recursive prediction model is used to estimate the potential deviation of motor speed and torque from the set value over the next three control cycles. The predicted deviation value is injected into the constraints of the MPC optimization problem in advance, so that the optimal excitation voltage vector solved by MPC aims to prevent deviation rather than correct it. S5. Continuous Domain Operation with End-to-End Timing Alignment: A unified timestamp derived from S(t) is applied to the current state snapshot and control commands. Based on the aligned data and the MPC model with injected prediction bias constraints, continuous domain parallel iterative operations are performed on a hardware parallel multiplier array, outputting a macroscopically continuous and smooth excitation control command. This command, after continuous domain Σ-Δ modulation, drives the power devices, generating a macroscopically continuous and smooth sinusoidal current in the motor windings. In this embodiment, all calculation steps of the method are decoupled from the binary serial instruction fetching and decoding logic of the von Neumann architecture. The calculations are not segmented by a fixed discrete clock cycle; instead, continuous, unsegmented steady-state calculations are achieved entirely based on a four-dimensional composite spatiotemporal reference S(t). Compared to traditional PWM discrete control, motor efficiency is increased to over 99%, harmonic content meets the requirements of the GB / T14549 national standard, and torque ripple and vibration are significantly reduced.

[0009] Example 2: Distributed power grid converter control based on natural geomagnetic field reference This embodiment takes the coordinated control of multiple distributed grid converters as an example to illustrate the implementation process of this method in a multi-node distributed scenario; Multiple converters are installed in different geographical locations. Each converter acquires the local geomagnetic field vector through a built-in fluxgate sensor. Combined with GNSS timing signals and inertial measurement unit data, a multi-sensor fusion algorithm is used to calculate its own spatial position vector P. All converters share the same high-stability frequency source or atomic clock timing signal as the reference frequency f0. Each converter executes steps S2 to S5 independently. However, since all nodes are time-aligned based on the same four-dimensional composite spatiotemporal reference S(t), the control commands output by each converter are naturally synchronized in time, achieving microsecond-level precision collaborative control without the need for inter-node communication. When the geomagnetic field at the location of a converter is subject to local disturbance, its multi-mode fusion unit automatically reduces the fusion weight of the geomagnetic signal and increases the weight of the inertial measurement unit and GNSS timing signal, maintaining the continuous stability of the spatial reference.

[0010] Example 3: Attitude Control of Deep Space Probe Based on Pulsar Timing Reference This embodiment takes the attitude control of a deep space probe after it leaves low Earth orbit as an example to illustrate the implementation process of this method in an environment without GNSS signals. The detector is equipped with an X-ray pulsar timing detector and a star sensor. The pulsar timing signal provides a high-precision time reference T(t) in deep space, while the star sensor provides stellar azimuth data for calculating the spatial position vector P. The detector's onboard atomic clock provides the reference frequency f0. A four-dimensional composite spatiotemporal reference S(t) is fully established in deep space. The reaction flywheel and thrusters on the probe are the controlled objects. The full state vector includes parameters such as the probe's three-axis attitude angles, angular velocities, flywheel speeds, and thruster temperatures. When the probe needs to adjust its attitude, the constant-axis bidirectional steady-state continuous calculation method completes full-state perception, bidirectional closed-loop interaction, prediction deviation pre-compensation, and continuous-domain parallel iterative calculation within nanoseconds, outputting macroscopically continuous and smooth attitude control commands, enabling the probe to maintain high-precision attitude stability even without ground-based telemetry and control support.

[0011] The three embodiments described above are merely illustrative and do not constitute any limitation on the scope of protection of this invention. Adaptive adjustments made by those skilled in the art within the framework of the core method steps of this invention for different controlled objects and application scenarios all fall within the scope of protection of this invention.

[0012] Patent System Association Statement This patent is the foundational mother patent of the constant axis technology system, protecting the most abstract and highest-level general continuous calculation method core paradigm, and is the basic and guiding core patent of this technology system; Subsequent patents extending this technology system, including but not limited to "A Constant Axis Physical Domain Processor and Control Method", "A Constant Axis Dual-Field Coupled Continuous Domain Paradigm Motor and Power Generation Control Method", "A Real-Time Music Generation and Performance Processor Based on Constant Axis Continuous Domain Control", and "A Global Wireless Energy Transmission Networking System and Method Based on Constant Axis Global Steady-State Reference and Natural Field Adaptive Coupling", are all specific implementations and optimizations of this patented technology solution in specific hardware carriers and specific application scenarios. This patent is not bound to any specific hardware platform, controlled device or application scenario. Its protection scope fully covers all physical domain real-time continuous calculation implementation methods that adopt the core logic of the present invention, namely, four-dimensional composite spatiotemporal reference, two-way closed-loop parallel interaction, prediction deviation steady state maintenance and continuous domain parallel iterative operation. This invention serves as the sole fundamental legal carrier of the original algorithm (a two-way steady-state continuous computation paradigm with a constant axis), governing the entire series of derivative technologies and industrial application systems.

Claims

1. A method for calculating steady-state continuity in two directions with a constant axis, characterized in that, Includes the following steps: S1. Establish a four-dimensional composite spatiotemporal reference S(t)=[f0, φ(t), T(t), P], where f0 is the reference frequency, φ(t) is the real-time phase, T(t) is the high-precision timing, and P is the spatial position vector of the controlled object or computing node; this four-dimensional composite spatiotemporal reference provides a unified time index and spatial reference system for all subsequent calculation steps; S2. Under the unified clock cycle of the four-dimensional composite spatiotemporal reference, all state variables from multiple sensor channels are synchronously latched within the same clock cycle to generate a complete state snapshot at the current moment. S3. The sensing data stream and the action feedback stream are merged in parallel within the same clock cycle without time-division serial processing, forming a two-way closed loop of sensing and action. S4. Based on the complete state snapshot at the current moment, the future deviation trend is estimated by simplifying the prediction model, and the predicted deviation is injected into the constraints of the control command in advance, so that the control command maintains the steady state of the system in a way that prevents deviation rather than corrects deviation. S5. Add a unified timestamp derived from the four-dimensional composite spatiotemporal reference to each set of state snapshots and control commands to ensure that the data of the entire link are strictly aligned under the same time reference. On this basis, complete the parallel iterative operation in the continuous domain and output the macroscopic continuous and smooth control commands. The continuous domain parallel iterative operation is independent of the binary serial fetch and decode execution logic, and does not perform segmented operations based on a fixed discrete clock beat. It achieves continuous, unsegmented steady-state computation based entirely on a four-dimensional composite spatiotemporal reference.

2. The method for calculating steady-state continuity with a constant axis in two directions according to claim 1, characterized in that: The simplified prediction model can be any one of the following: a linear state equation recursive prediction model, a Kalman simplified prediction model, or an iterative recursive prediction model.

3. The method for calculating steady-state continuity with a constant axis in two directions according to claim 1, characterized in that: The control command is any one of the following: motor excitation control command, generator set regulation command, spacecraft attitude control command, robot motion control command, power grid converter regulation command, or audio drive control command.

4. The method for calculating steady-state continuity with a constant axis in two directions according to claim 1, characterized in that: The reference frequency f0 in the four-dimensional composite spatiotemporal reference is provided by a high-stability crystal oscillator or atomic clock, the real-time phase φ(t) is maintained by a phase-locked loop circuit, the high-precision timing T(t) is integrated with a precision time protocol and an external timing signal, and the spatial position vector P is calculated by a multi-sensor fusion algorithm.

5. The method for calculating steady-state continuity in two directions with a constant axis according to claim 1, characterized in that: The sensing data stream includes at least three of the following: current, voltage, rotational speed, position, temperature, vibration, magnetic field, or attitude data of the controlled object.

6. The method for calculating steady-state continuity in two directions with a constant axis according to claim 1, characterized in that: The method can be applied to any real-time control scenario that requires the output of macroscopic, continuous, and smooth physical quantities of the controlled object.