Flexible mechanical arm multi-section precise driving method based on nickel-titanium memory alloy wire

Abstract

The invention relates to the technical field of flexible mechanical arm driving control, and discloses a flexible mechanical arm multi-section precise driving method based on nickel-titanium memory alloy wires, which comprises the following steps: generating structural modal resonance by using a first nickel-titanium memory alloy wire to construct an information medium, and picking up a response signal by using a second nickel-titanium memory alloy wire; the method comprises the following steps of: acquiring a response signal from an intrinsic state, performing transient characteristic analysis on a resonance spectrum of the response signal to distinguish whether resonance change is caused by intrinsic state change or external environment contact, and finally realizing self-adaptive switching of a control mode. The intrinsic physical state of the driving element is converted into the structural resonant frequency capable of being stably measured, so that the basis of closed-loop control is converted into an internal real-time physical state from an external lagged motion result, and therefore, the problems of control nonlinearity and lagging caused by the unknown state are solved.

Description

Technical Field

[0001] This invention relates to a multi-segment precision driving method for a flexible robotic arm based on nickel-titanium shape memory alloy wire, belonging to the field of flexible robotic arm drive and control technology. Background Technology

[0002] Currently, using nickel-titanium shape memory alloy wire as the driving element is a common technical approach. Its control structure typically involves a controller issuing heating commands and external sensors observing the robotic arm's motion to generate feedback. In this structure, drive execution and state observation are physically separated. However, when such flexible robotic arms are applied to dynamically changing working environments, the shortcomings of the above control structure in terms of real-time performance and stability become apparent. The overall shape change of the nickel-titanium shape memory alloy wire depends on the ratio of martensite to austenite phase transformation within it. This phase transformation ratio directly determines the equivalent Young's modulus of the alloy wire, which is the most direct intrinsic physical basis for its motion state. Existing external observation methods can only obtain the final result of the motion; the controller cannot directly obtain this decisive physical condition within the driving element. Because the physical state information is not readily available, there is a delay in information acquisition between the issuance of control commands and the actual evolution of the physical state of the actuator. This delay means that the drive commands issued by the controller cannot accurately correspond to the real-time physical state of the alloy wire. When rapid or fine adjustments are required, the commands issued by the control system to obtain the desired response are prone to motion overshoot or oscillation due to failure to match the actual phase transition process. To address this, those skilled in the art have made improvements such as optimizing the control algorithm or improving the measurement accuracy of external sensors. However, these improvements are still based on compensating for the delayed motion results and do not change the fundamental operating principle that the controller cannot directly acquire the real-time physical state of the drive element.

[0003] However, existing technologies are limited not only in the physical layout of sensors, but also in the control strategies that attempt to use the actuator's own parameters for feedback, which fail to fundamentally solve the problem in terms of information interpretation. For example, Chinese invention patent CN111993462A discloses a flexible robotic arm based on SMA-driven multi-segment rigid-flexible coupling. Its technical solution proposes to achieve feedback control by calculating the resistance of the SMA wire. The idea behind this method is to use the change in resistance value during the phase transition to indirectly reflect the driving state. However, resistance is a relatively general physical quantity. Although it can reflect the intrinsic phase of the SMA wire, it is not a reliable indicator of the driving state. The system changes state but cannot distinguish the root cause of the state change. Specifically, whether it is a phase change caused by the controller's active heating or a phase change caused by the robot arm's physical contact with the external environment, resulting in changes in stress, strain, or heat dissipation conditions, both will cause changes in resistance. The control logic confuses these disturbances of different natures. Once external contact occurs, the system is very likely to misjudge it as a position error of its own movement and make incorrect compensation adjustments, such as continuously increasing the drive current to resist external contact, thus exhibiting rigid resistance behavior. This is exactly contrary to the original intention of the flexible robot arm to be compliant interaction and fails to truly solve the core problem of state information confusion.

[0004] Specifically, this technical approach has the following unresolved issues: 1. There is a lack of a real-time acquisition path for the intrinsic physical state of the drive element between the control system and the drive element, resulting in asynchrony between the issuance of control commands and the actual evolution of the drive state; 2. The feedback of the control process relies on the delayed observation of external motion results, making it difficult to effectively suppress the instability of the control response caused by the phase transformation characteristics of the material. Therefore, how to establish a real-time information path for acquiring the intrinsic physical state of the nickel-titanium shape memory alloy wire actuator, and transform the basis of closed-loop control from the delayed external motion results to the real-time internal physical state, becomes the technical problem to be solved by this invention. Summary of the Invention

[0005] This invention provides a multi-segment precision driving method for a flexible robotic arm based on nickel-titanium shape memory alloy wire. Its main purpose is to solve the problem in the prior art that the control command and drive response are not synchronized due to the inability to obtain the intrinsic physical state of the drive element in real time, making it difficult to effectively suppress the hysteresis and nonlinear response of the control process.

[0006] To achieve the above objectives, this invention provides a multi-segment precision driving method for a flexible robotic arm based on nickel-titanium shape memory alloy wires. At least one segment of the flexible robotic arm is driven by a plurality of nickel-titanium shape memory alloy wires. The method includes: Step a, the excitation and information medium construction step, applies a composite driving signal containing macroscopic motion driving components and structural vibration excitation components to at least one first nickel-titanium shape memory alloy wire as the excitation source. The structural vibration excitation components are used to continuously excite structural modal resonance in the mechanical structure of the flexible robotic arm segment, thereby transforming the mechanical structure of the flexible robotic arm segment into a controlled vibration information medium that carries its intrinsic state and external interaction information in real time. Step b, the response pickup and source separation step, synchronously monitors at least one second nickel-titanium shape memory alloy wire, which is different from the first nickel-titanium shape memory alloy wire, and collects the periodic fluctuations of the electrical parameters of the second nickel-titanium shape memory alloy wire itself caused by structural modal resonance to obtain a response signal. The response signal is then subjected to spectrum analysis to obtain a resonance spectrum containing fundamental frequency and harmonic information. Based on the transient characteristics of the resonance spectrum within a time window, it is determined whether the change in resonance frequency is caused by intrinsic state changes or by external environmental contact. Step c, State Decoding and Adaptive Closed-Loop Control Step: When step b determines that the change in resonant frequency is caused by an intrinsic state change, the current intrinsic state is decoded based on the preset mapping relationship between the fundamental frequency of the resonant spectrum and the intrinsic phase transition state of the nickel-titanium shape memory alloy line, and motion closed-loop control is executed; when step b determines that the change in resonant frequency is caused by contact with the external environment, the closed-loop control is triggered to switch to the preset interactive control mode.

[0007] Preferably, the structural vibration excitation component is a micro-amplitude AC signal superimposed on the macroscopic motion driving component used to generate macroscopic motion; the response pickup and source separation step specifically involves obtaining the response signal by monitoring the fluctuation of the resistance or voltage of the second nickel-titanium shape memory alloy wire, and determining the fundamental frequency peak of the resonance spectrum by performing a fast Fourier transform on the response signal.

[0008] Preferably, in the response pickup and source separation step, the determination of whether the change in resonant frequency is caused by an intrinsic state change or by external environmental contact specifically involves: calculating a transient impulse factor defined by the characteristics of the resonant spectrum. ,in ,in, The fundamental frequency value. The rate of change of the fundamental frequency value within a time window. This refers to the change in the energy ratio of harmonic information relative to the fundamental frequency. and The weighting coefficients are pre-calibrated based on the material and structure of the flexible robotic arm segment; when the transient impact factor... Exceeding the contact detection threshold preset based on experimental data When the change in resonant frequency is determined to be caused by contact with the external environment.

[0009] Preferably, the current intrinsic state includes the phase transformation fraction or equivalent Young's modulus of one or more of the first nickel-titanium shape memory alloy wires and the second nickel-titanium shape memory alloy wires.

[0010] Preferably, the method further includes a health status self-diagnosis step, which includes: storing one or more baseline resonance fingerprints representing the initial health status of the flexible manipulator, the baseline resonance fingerprints including the quality factor Q value of the resonance peak and the noise floor level of the non-resonant frequency band; continuously recording real-time resonance fingerprints determined by the response signal during the operation of the flexible manipulator; diagnosing the health status of the flexible manipulator by analyzing the long-term evolution trend of the real-time resonance fingerprints relative to the baseline resonance fingerprints, the long-term evolution trend including a systematic decrease in the quality factor Q value or a systematic increase in the noise floor level, and generating predictive maintenance information based on the diagnostic results.

[0011] Preferably, predictive maintenance information includes one or more of the following: drive line fatigue risk level or mechanical connection loosening risk level.

[0012] Preferably, the method further includes: in the response pickup and source separation step, determining the phase difference between the structural vibration excitation component in the composite drive signal and the response signal; and determining a dissipation factor characterizing the energy dissipation characteristics of the flexible robotic arm segment based on the phase difference; judging whether the flexible robotic arm segment is in an active drive-dominated mode or a passive disturbance-dominated mode based on the degree of deviation between the real-time value of the dissipation factor and its intrinsic baseline value based on the current fundamental frequency; and adaptively adjusting the control parameters of the motion closed-loop control based on the judgment result.

[0013] Preferably, the excitation and information medium construction step includes applying structural vibration excitation components to different nickel-titanium shape memory alloy lines in turn for distributed excitation; the response picking and information source separation step includes synchronously monitoring multiple unexcited nickel-titanium shape memory alloy lines to obtain an excitation-response relationship matrix; the state decoding and adaptive closed-loop control step further includes: based on the excitation-response relationship matrix, identifying multiple different structural vibration modes of the flexible robotic arm segment, and continuously tracking the resonant frequency of a specific structural vibration mode with a pre-assigned identity to decode the current intrinsic state.

[0014] Preferably, the multiple different structural vibration modes include first-order bending mode and torsional mode; a specific structural vibration mode with a pre-assigned identity is continuously tracked, specifically: when mode aliasing or mode switching causes the resonance peak of the specific structural vibration mode to no longer be the resonance peak with the strongest energy, it is still locked and tracked according to its mode identity.

[0015] Preferably, the method further includes a drive efficiency self-optimization step, which includes: continuously monitoring the heating current signal used to drive the macroscopic motion drive component of the first nickel-titanium shape memory alloy wire during the execution of motion closed-loop control; identifying the real-time heat exchange efficiency between the first nickel-titanium shape memory alloy wire and the surrounding environment online based on the statistical characteristics of the heating current signal within a time window, the statistical characteristics including the mean and variance of the heating current signal; and adaptively adjusting the control parameters of the motion closed-loop control according to the identified real-time heat exchange efficiency, so as to minimize drive energy consumption while maintaining the target intrinsic state.

[0016] Compared with the prior art, the beneficial effects of the present invention are: 1. By applying a driving signal containing an excitation signal component to a nickel-titanium shape memory alloy wire and simultaneously monitoring the changes in the electrical parameters of the remaining nickel-titanium shape memory alloy wires, a unified operating mechanism for driving and sensing behavior in terms of physical structure is established. This mechanism transforms the intrinsic phase transition state of the alloy wire, which directly determines the driving effect, into a structural modal resonant frequency that can be stably measured. This changes the basis of closed-loop control from external, delayed motion results to internal, real-time physical states, thus avoiding the nonlinearity and hysteresis problems caused by the unknown state in traditional control methods from the perspective of the working mechanism.

[0017] 2. Based on the determination of the structural modal resonant frequencies, the response signal is further subjected to spectral analysis to obtain the resonant spectrum, and the transient characteristics of the spectrum are incorporated into the judgment. When the change in resonant frequency presents as a smooth drift of the fundamental frequency, the system determines it as an intrinsic state change and executes motion control; when the spectrum characteristics show a pulse-like surge in harmonic energy or a sharp drop in the quality factor, it is determined to be caused by the contact between the arm body and the external environment. This method of correlating frequency changes with spectrum fingerprints allows a single response signal to simultaneously carry information about two different types of nature: its own state and external interaction. This gives the system the ability to distinguish between its own motion and external contact in addition to motion control, thus providing a clear trigger basis for subsequent switching to interactive control mode.

[0018] 3. This method compares and analyzes the long-term evolution trend of the real-time resonant fingerprints continuously acquired during operation with the pre-stored baseline resonant fingerprints. Due to material fatigue of the drive line or fretting loosening of mechanical connections, fingerprint features such as the quality factor of the resonant peak or the noise floor will undergo irreversible systematic drift. Therefore, this analysis process utilizes the response signal that is inevitably generated to achieve motion control, accumulates it in the time dimension, and extracts the decay characteristics that characterize the health status of the system. In this way, without affecting the main drive control process, it provides the system with a predictive maintenance information generation path based on the actual operating load, changing the limitation of relying on fixed cycles or post-fault maintenance. Attached Figure Description

[0019] Figure 1 This is a schematic diagram of the closed-loop control and health self-diagnosis logic flow of the present invention; Figure 2 This is a performance comparison chart of the test group and the control group for external contact events of the present invention; Figure 3 This is a diagram of the multi-level hardware and information interaction architecture of the system implementation of this invention. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of the present invention clearer, the present invention will be described in further detail below. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] The invention claims a multi-segment precision driving method for a flexible robotic arm based on nickel-titanium shape memory alloy wire. This method integrates driving, sensing, and information processing into a unified operating mechanism. The method generally includes excitation and information medium construction steps, response acquisition and source separation steps, and state decoding and adaptive closed-loop control steps. By transforming the intrinsic physical state of the nickel-titanium shape memory alloy wire, which directly determines the driving effect, into a stably measurable structural modal resonant frequency, and by changing the basis of closed-loop control from externally delayed motion results to internal real-time physical states, the method addresses the nonlinear control caused by state uncertainty. The invention addresses the issues of performance and hysteresis. In a specific application scenario, such as a flexible robotic arm end effector driven by multiple nickel-titanium shape memory alloy wires used in minimally invasive laparoscopic surgery, precise and stable bending and posture maintenance are essential. Existing technologies, unable to directly perceive the internal phase transition process of the drive wires, often cause motion overshoot or oscillations when fine adjustments are required due to the asynchrony between control commands and the actual evolution of the actuator's physical state, posing a potential risk to surrounding tissues. To solve this problem, the method claimed in this invention is configured to first perform an excitation and information medium construction step, specifically, the controller directs the excitation to the excitation source. At least one first nickel-titanium shape memory alloy wire is subjected to a composite driving signal, which consists of a macroscopic motion driving component for generating bending motion and a structural vibration excitation component for exciting structural vibration. The macroscopic motion driving component can be a pulse width modulation (PWM) signal with a frequency between 0.1 Hz and 1 Hz, the change of its duty cycle being used to adjust the heating power, thereby controlling the phase transition process of the nickel-titanium shape memory alloy wire to generate the desired bending motion. The structural vibration excitation component is a micro-amplitude AC signal with a frequency range of 10 Hz to 500 Hz superimposed on the PWM signal, the amplitude of which is... The amplitude is set between 1% and 5% of the equivalent DC voltage of the macroscopic driving component. A specific numerical example is that when the equivalent voltage of the macroscopic driving component is 5V, the peak-to-peak value of the excitation signal is 0.1V. The basis for this amplitude setting is that, through offline calibration experiments, an energy window is determined that can continuously excite clearly identifiable structural modal resonances without causing perceptible motion disturbances to the robotic arm. In this way, the mechanical structure of the flexible robotic arm segment is transformed into a controlled vibration information medium that carries its intrinsic state and external interaction information in real time, which establishes an information path for subsequently obtaining the real-time intrinsic physical state of the actuator.

[0022] Based on the construction of the information medium, the system needs to accurately extract and separate effective information, especially when the change in resonant frequency may be caused by two different events: intrinsic state change or external environmental contact, which poses a risk of information confusion. Therefore, the system simultaneously executes response acquisition and source separation steps. This step utilizes at least one second nickel-titanium shape memory alloy wire, different from the first, as a sensor. A data acquisition card synchronously monitors the periodic resistance or voltage fluctuations at both ends of the second nickel-titanium shape memory alloy wire caused by structural modal resonance at a sampling frequency of not less than 1 kHz to obtain the response signal. To achieve source separation, this step employs a spectrum analysis method based on Fast Fourier Transform (FFT) to process the response signal acquired within a time window of 1024 sampling points, thereby obtaining a resonant spectrum containing fundamental frequency peak and harmonic information. Furthermore, this step calculates a transient impact factor defined by the characteristics of the resonant spectrum. To determine the root cause of the event, transient impact factor The calculation formula was determined as follows: ; where is the frequency value of the fundamental frequency determined by searching for the maximum energy peak in the resonance spectrum, in Hertz (Hz). The rate of change of the fundamental frequency value in the current time window relative to the previous time window is expressed in Hertz per second (Hz / s). It is the change in the ratio of the total harmonic information energy to the fundamental frequency energy over two consecutive time windows, and is a dimensionless value; and These are dimensionless weighting coefficients pre-calibrated based on the materials and structure of the flexible robotic arm segments; weighting coefficients and The calibration procedure is as follows: First, drive the robotic arm to perform slow reciprocating motion in free space, and record the changes caused by intrinsic state changes during this process. and The statistical distribution was analyzed, and its characteristic range was defined as the intrinsic variation region. Subsequently, probes were used to make instantaneous external contact with the robotic arm at different speeds and forces, and the effects caused by the external contact during this process were recorded. and The impulse response is used to define its characteristic range as the external contact region; finally, the weights that can optimally distinguish these two regions are determined by linear discriminant analysis (LDA). and The value of , for example, for a flexible arm segment mainly composed of polymer, It can be calibrated to 0.3. It can be calibrated to 0.7; when the calculated transient impact factor... Exceeding a contact determination threshold preset based on the above calibration experimental data For example Set as an intrinsic change zone If the frequency change is more than five standard deviations from the statistical mean, it is determined that the change in resonant frequency is caused by contact with the external environment. By analyzing the frequency change rate in relation to the change in spectral energy distribution, a single response signal can be decoded into two different types of information: information about its own state and information about external interaction. This provides a triggering basis for the adaptive switching of subsequent control modes.

[0023] Finally, based on the separated information sources, the system executes state decoding and adaptive closed-loop control steps; when the response pickup and information source separation steps determine that the change in resonant frequency is caused by an intrinsic state change, i.e. The system is based on the fundamental frequency of the resonance spectrum. A pre-defined mapping relationship between the intrinsic phase transition states of the nickel-titanium shape memory alloy wire and the intrinsic phase transition states is used to decode the current intrinsic state. The intrinsic state can be specifically the phase transition fraction or equivalent Young's modulus of one or more of the driving wires. This pre-defined mapping relationship is established through an offline calibration procedure: in a controlled temperature environment, a series of quasi-static heating currents are applied to the nickel-titanium shape memory alloy wire, and its phase transition fraction or Young's modulus is measured using standard methods such as differential scanning calorimetry (DSC) or dynamic mechanical analysis (DMA), while the structural resonant fundamental frequency measured by the method of this invention is recorded simultaneously. This yields a lookup table or polynomial fitting function for the fundamental intrinsic state. In closed-loop control, this real-time decoded intrinsic state is input as a feedback signal to a proportional-integral-derivative (PID) controller. This controller compares the real-time intrinsic state with the target setpoint and dynamically adjusts the heating current applied to the macroscopic motion drive component of the first nickel-titanium shape memory alloy wire based on the error, thereby driving the movement of the robotic arm. On the other hand, when the response pickup and source separation steps determine that the change in resonant frequency is caused by contact with the external environment, i.e. If this happens, the closed-loop control will immediately switch to a preset interactive control mode. For example, the system can switch to a zero-force control mode, immediately stopping the heating of the drive line to make the robotic arm compliant, or switch to a constant-force control mode, adjusting the drive current to maintain a constant transient impact factor. value.

[0024] Where there is no conflict, to further enhance the robustness of the system, the method claimed in this invention may optionally include one or more additional technical procedures; one of which is a health status self-diagnosis step, which, upon the first use of the robotic arm, records and stores one or more baseline resonance fingerprints representing its initial health status, specifically including the quality factor of the resonance peak. The system continuously records the real-time resonant fingerprint determined by the response signal during subsequent operation, and analyzes the long-term evolution trend of the real-time resonant fingerprint relative to the baseline resonant fingerprint, such as the quality factor. The first method involves a systematic monotonic decrease in the value or a systematic increase in the noise floor level to diagnose degradation such as material fatigue in the drive line or fretting loosening of mechanical connections. Predictive maintenance information, such as the fatigue risk level of the drive line or the loosening risk level of the mechanical connection, is generated based on the diagnostic results. The second method is a disturbance decoupling procedure based on a dissipation factor. In the response picking step, this procedure uses a phase-locked loop circuit to determine the phase difference between the structural vibration excitation component in the composite drive signal and the response signal, and determines a dissipation factor characterizing the energy dissipation properties of the flexible robotic arm segment based on this phase difference. The real-time value of this dissipation factor is then compared with its intrinsic value based on the current fundamental frequency. By comparing line values, the system can determine whether the boom segment is in an active mode dominated by internal drive or a passive disturbance mode dominated by external fluid impact or tissue creep. Based on this determination, the system adaptively adjusts the control parameters of the PID controller, for example, reducing the integral gain in the passive disturbance mode to adapt to the dynamic environment. Thirdly, there is a self-calibration procedure based on multimodal analysis. In the excitation step, the procedure alternately applies the structural vibration excitation component to different nickel-titanium shape memory alloy lines for distributed excitation. In the response picking step, it simultaneously monitors multiple unexcited nickel-titanium shape memory alloy lines to obtain an excitation-response relationship. The system uses a matrix; based on the topological features of this matrix, it can identify multiple different structural vibration modes of the flexible robotic arm segment, such as first-order bending and torsional modes, and assign identity labels to these modes. In the state decoding step, the system can continuously track the resonant frequency of a specific structural vibration mode with a pre-assigned identity. Even when mode aliasing or mode switching occurs, causing the resonant peak of that specific mode to no longer be the strongest, it can still lock and track it based on its modal identity, thus ensuring the continuity and accuracy of state decoding. The fourth step is a drive efficiency self-optimization step, which continuously optimizes the drive efficiency during the motion closed-loop control execution. The system monitors the heating current signal used to drive the macroscopic motion component of the first nickel-titanium shape memory alloy wire, and identifies the real-time heat exchange efficiency between the first nickel-titanium shape memory alloy wire and the surrounding environment based on the statistical characteristics of this signal within a sliding time window, including its mean and variance. For example, when the robotic arm is working in a flowing liquid, the heat dissipation conditions improve, and to maintain the same intrinsic state, the controller needs to output a current signal with a higher mean and a larger variance. Based on the identified real-time heat exchange efficiency, the system can adaptively adjust the parameters of the PID controller to minimize drive energy consumption while maintaining the target intrinsic state.

[0025] Example 1: In a flexible robotic arm application for intracardiac catheter ablation surgery, the end effector needs to adhere to and maintain stable contact force on the beating cardiac wall tissue. This condition requires the control system to have position control capability during free space movement and compliance force control capability when contacting dynamic tissue. However, traditional control methods relying on external sensors suffer from response delays and judgment ambiguities when switching between these two states. In this application, when the end effector of the flexible robotic arm moves towards the target ablation point in the blood environment, the system executes the excitation and information medium construction steps, applying a composite drive signal containing macroscopic motion drive components and structural vibration excitation components to the first nickel-titanium shape memory alloy wire, and simultaneously executing response pickup and signal source separation steps. The controller, based on the fundamental frequency-intrinsic state lookup table, detects the fundamental frequency of the resonance spectrum that has been smoothly drifted due to the phase transition of the drive wire itself. The current intrinsic state is decoded and used as feedback for closed-loop motion control, driving the robotic arm to move towards the predetermined target point; during this process, the transient impact factor calculated in the response pickup and source separation steps is used. The contact threshold is always below the preset threshold. The system remains in motion control mode. It should be noted that the micro-amplitude vibration excitation continuously injected in the excitation and information medium construction steps enables the response pickup and source separation steps to obtain a resonance spectrum with high time resolution, which provides a data basis for capturing subsequent transient contact events.

[0026] At the instant the robotic arm's end effector makes initial contact with the beating heart wall tissue, the resonance spectrum of the acquired response signal undergoes a momentary change due to the nonlinear impact of the contact; specifically, this change manifests as a change in the fundamental frequency. A jump occurs, and simultaneously, due to the introduction of new energy dissipation pathways and nonlinear constraints by the contact, the quality factor of the resonance peak decreases. The decrease in value and the increase in the proportion of harmonic energy together lead to a decrease in the transient impact factor. The calculated result increases within one sampling period and exceeds the contact determination threshold. The transient impact factor here By using the instantaneous rate of change of the frequency of the resonant spectrum With harmonic energy change These two characteristic quantities, reflecting different physical processes, are weighted and fused. The judgment logic no longer relies on tracking a single frequency value, but rather distinguishes between events that can cause fundamental frequency changes—both intrinsic state changes and external contact—by analyzing the overall changes in the spectral fingerprint. Once the transient impact factor... Exceeding the threshold Upon triggering the state decoding and adaptive closed-loop control steps, the control system immediately switches from the position control mode based on intrinsic state feedback to the preset interactive control mode. In this interactive control mode, the controller's objective is no longer to track a geometric position, but to maintain a constant contact state by adjusting the heating current of the macroscopic motion driving component. For example, it can reduce the transient impact factor. As an indicator of the contact state, it is maintained within a stable range corresponding to the target contact force; thus, when the cardiac wall tissue moves away due to pulsation, the system fine-tunes the drive to make the arm follow, and when the cardiac wall tissue moves closer, it yields smoothly.

[0027] Example 2: To objectively verify the effectiveness of the method claimed in this invention in distinguishing between intrinsic state changes and contact with the external environment, and in achieving adaptive control switching, the following experiment was conducted. A single-segment flexible robotic arm was used as the test object. This robotic arm was driven by three 0.2mm diameter nickel-titanium shape memory alloy wires, fixed to a six-dimensional force sensor base. The force sensor had a data acquisition frequency of 1kHz and a resolution of 0.01N, used to record contact force for effect verification. One of the nickel-titanium shape memory alloy wires was designated as the first nickel-titanium shape memory alloy wire, used to receive the composite drive signal, and the other was designated as the second nickel-titanium shape memory alloy wire, used to pick up the response signal. The experiment employed a laser displacement sensor to measure the spatial displacement of the robotic arm's end effector, achieving a measurement accuracy of 0.05 mm. A linear slide with a rigid probe at the end effector, driven by a servo motor, was used to apply controlled external environmental contact. The entire experiment was controlled and recorded by a real-time controller, with the data acquisition card's sampling frequency set to 2 kHz. This sampling frequency was chosen to be twice the highest effective frequency component (500 Hz) in the response signal, to meet the requirement of distortion-free signal sampling. The experiment included a comparison between an experimental group using the method of this invention and a control group using existing technology. The experimental group fully implemented all steps of this invention, including those based on transient impact factors. The event judgment and control mode switching were performed; the control group adopted a PID position closed-loop control method based on an external laser displacement sensor. Its control objective was to maintain the commanded position of the robotic arm end, and it did not have the ability to distinguish between intrinsic state changes and external contact. The test process was divided into two stages. The first stage was the intrinsic state change test. The controller issued a command to the robotic arms of both groups to complete a 30-degree bend within 10 seconds. The second stage was the external environment contact test. The controller commanded the robotic arm to maintain a fixed bending posture. At this time, the linear slide drove the probe to make perpendicular contact with the end of the robotic arm at a speed of 5 mm / s.

[0028] In the first phase of intrinsic state change testing, the experimental group's system moved along a preset trajectory. During this period, the response pickup and source separation steps continuously calculated the resonance spectrum and observed the fundamental frequency. As the drive current increases, the frequency smoothly and continuously drifts from the initial 125Hz to the final 140Hz, with the corresponding frequency change rate... The change in the harmonic energy ratio is always kept below 5 Hz / s. The fluctuations are weak, and the calculated transient impact factor The value remains below the preset contact detection threshold. The system maintained the motion closed-loop control mode, with the error between the end trajectory and the commanded trajectory being less than 0.5 mm. The control group also completed the motion, and its position error was at the same level as the test group. In the second stage of the external environment contact test, when the probe contacted the end of the control group's robotic arm, its PID controller treated the displacement caused by the external contact as a position error, and then increased the drive current to try to resist the external disturbance, resulting in a contact force peak of 3.2 N between the robotic arm and the probe, exhibiting rigid resistance behavior. For the test group, at the instant of contact, the system detected an instantaneous change in the resonance spectrum, with the fundamental frequency... The frequency jumps abruptly from 135Hz to 155Hz within 10ms, resulting in a pulse-like increase in higher-order harmonic energy and causing a transient impulse factor. The calculated value instantaneously exceeded the contact determination threshold. The state decoding and adaptive closed-loop control steps are triggered immediately, the system control mode switches from position closed-loop control to interactive control mode, the controller stops increasing the drive current, the robotic arm becomes compliant, and the final stable contact force between the robotic arm and the probe is 0.3N; see Table 1 for a comparison of the key data of the two groups in the second stage of testing.

[0029] Table 1: Comparison of key data from external environment contact tests.

[0030]

[0031] Experimental data show that the transient impact factor constructed by the experimental group using the method of this invention is... It can distinguish between the smooth drift of the resonant frequency caused by changes in intrinsic state and the instantaneous impact of the resonant spectrum caused by contact with the external environment, and based on this, it can determine the adaptive switching of the trigger control mode, thereby avoiding rigid confrontation when in contact with the external environment and keeping the interaction force at a low level.

[0032] To further verify the necessity and uniqueness of the source separation based on the transient characteristics of the resonance spectrum proposed in this invention, a comparative example 1 is set up as follows.

[0033] Comparative Example 1: Comparative Example 1 uses the exact same hardware configuration and test platform as the test group in Example 2, including a single-segment flexible robotic arm driven by three nickel-titanium shape memory alloy wires, a six-dimensional force sensor and a laser displacement sensor for data verification. The only difference between Comparative Example 1 and the test group in Example 2 is the control method. Although the controller of Comparative Example 1 also monitors the resonant frequency... However, its control logic is set to a simplified feedback control approach that a person skilled in the art might adopt when facing this problem, namely, the controller relies solely on the resonant fundamental frequency. The deviation between the real-time value and the target value is adjusted based on feedback, and the control objective is to maintain a constant resonant fundamental frequency corresponding to the desired arm posture. However, the controller lacks the logic to distinguish the source of resonant frequency changes based on transient impact factors, and therefore lacks the ability to adaptively switch to interactive control mode. The test process was also divided into two phases. In the first phase, the intrinsic state change test, the controller of Comparative Example 1 adjusted the heating current to change the resonant fundamental frequency generated by the driven robotic arm. It can follow the preset frequency change trajectory (from 125Hz to 140Hz) to complete the specified 30-degree angle bending motion. Its end trajectory error is at the same level as the test group in Example 2, which shows that the control method has basic motion control capability under working conditions without external contact.

[0034] In the second phase of the external environment contact test, the robotic arm was instructed to maintain a fixed bending posture with a resonant fundamental frequency of 135Hz. At the instant the probe driven by the linear slide made perpendicular contact with the end effector of the robotic arm at a speed of 5mm / s, the contact introduced a new mechanical constraint, causing the resonant fundamental frequency of the arm segment to... Within 10ms, the frequency jumped abruptly from 135Hz to 155Hz. The controller of Comparative Example 1 interpreted this frequency jump as an intrinsic state error caused by excessive drive and immediately executed its preset control logic, which attempted to soften the arm body by significantly reducing the heating current of the drive line in order to reduce the resonant frequency back to the target value of 135Hz. However, this frequency change was caused by external physical contact (a rigid constraint) rather than internal phase transition state (a flexible constraint). Therefore, reducing the heating current could not effectively counteract this effect. Ultimately, the drive capability of the drive line was erroneously canceled, causing the robotic arm to passively yield under the continuous advancement of the probe and unable to maintain its preset posture. The end effector produced a position error of 5mm, which was equivalent to the displacement of the probe. Its control task failed. The key data of Comparative Example 1 and the test group in Example 2 were compared, and the results are shown in Table 2.

[0035] Table 2: Comparison of key data from external environment contact tests.

[0036]

[0037] The experimental results of Comparative Example 1 show that, in the absence of the mechanism proposed in this invention to analyze the transient characteristics of the resonant spectrum to distinguish the source of disturbance, even if the resonant frequency is used as the feedback signal, the controller cannot make the correct judgment and response to two completely different events: intrinsic state change and external environmental contact. When faced with external contact, the method mistakenly attributes the frequency change caused by external constraints to the deviation of the internal driving state and makes completely opposite control adjustments, ultimately causing the system to be unable to maintain the preset attitude and the control task to fail.

[0038] Example 3: This example combines Figures 1 to 3 This paper describes a multi-segment precision drive method for a flexible robotic arm based on nickel-titanium shape memory alloy wire, as follows: Figure 1 As shown, the controller sends a composite drive signal to the excitation and information medium construction stage, generating controlled vibration information within the flexible robotic arm. The response pickup and source separation stage collects the fluctuations in the electrical parameters of the response signal and separates the resonance spectrum and real-time resonance fingerprint from it. The resonance spectrum is sent to the state decoding and adaptive closed-loop control stage, which queries the preset mapping relationship fundamental frequency intrinsic state to decode the intrinsic state and generates macroscopic motion drive components to control the flexible robotic arm. The real-time resonance fingerprint is sent to the health status self-diagnosis stage, which compares the baseline resonance fingerprint to generate predictive maintenance information for the operator / maintenance system.

[0039] like Figure 2 As shown in the figure, this chart compares the key performance indicators of the experimental group's method of the present invention and the control group's conventional method in external contact events. Regarding the transient impact factor St peak value (times), the experimental group exhibits a significant peak value approximately 3.5 times the threshold, while the control group is not applicable due to the lack of this judgment mechanism. Regarding the contact force peak value N, the experimental group's contact force peak value is approximately 0.3 N, while the control group exhibits a contact force peak value as high as 3.2 N. Figure 3 As shown, the core embedded real-time controller integrates functional modules such as composite drive signal generation, resonance spectrum analysis and state decoding, adaptive closed-loop control logic, and health status self-diagnosis. The controller interacts with onboard data storage containing a baseline resonance fingerprint library and a state mapping model library via a data bus, and sends drive signals to the first nickel-titanium shape memory alloy wire of the flexible robotic arm body, while receiving response signals from the second nickel-titanium shape memory alloy wire sensor. The embedded controller can also connect to a host computer workstation via LAN / serial communication to realize human-machine interaction and monitoring interface functions, and can further communicate with a cloud server cluster via WAN to utilize its remote task planning and analysis module.

[0040] Example 4: To establish a preset mapping relationship between resonant frequency and intrinsic state, and to calibrate individual deviations caused by manufacturing tolerances or material batch differences, the following calibration procedure is performed: A single-segment flexible robotic arm to be calibrated is installed on a test platform, the ambient temperature of which is maintained at 25°C by a temperature control unit. The relative humidity is 50%. The platform integrates a force sensor with a measurement range of 0N to 10N and a resolution of no less than 0.001N, and a laser displacement sensor with a measurement range of 0mm to 50mm and an accuracy of 0.01mm. Both are used together to calculate the effective stiffness of the arm segment in a specific direction by applying a small known displacement and measuring the corresponding reaction force, and then converting it into the equivalent Young's modulus of the nickel-titanium shape memory alloy wire. A thermocouple is placed on the surface of the nickel-titanium shape memory alloy wire being measured to monitor its temperature. The calibration process is automatically executed by the host computer, and its data acquisition system is synchronized with the real-time controller of the robotic arm. After the calibration process starts, it is first ensured that all nickel-titanium shape memory alloy wires are in the fully martensitic phase at room temperature, at which point the initial resonant frequency is recorded. The initial equivalent Young's modulus obtained by force displacement measurement The calibration curve begins at a specific point; subsequently, the calibration program applies a heating current to the first nickel-titanium shape memory alloy wire, which serves as the main driving line. This current is controlled by a PWM signal, whose duty cycle increases linearly at a rate of 0.5% every 5 seconds. During this heating process, the system continuously executes the excitation and information medium construction steps and the response pickup steps, calculating and recording the current fundamental frequency of the resonance spectrum in real time at an update frequency of no less than 100Hz. Simultaneously, the calibration platform periodically drives the actuator at a frequency of 1Hz, causing a disturbance displacement of 0.1mm at the end of the robotic arm, and synchronously records the reaction force measured by the force sensor, calculating the current equivalent Young's modulus of the arm segment based on mechanical formulas. Thus, the system obtains a series of time-synchronized data pairs. , The process continues until the temperature monitored by the thermocouple reaches and stabilizes above the austenitic phase transformation completion temperature.

[0041] After completing the data acquisition for the heating process, the calibration program stops heating and allows the robotic arm to cool naturally in the environment, while continuously recording data pairs during the cooling process. , This system captures the hysteresis characteristics of nickel-titanium shape memory alloy wires during phase transition. All collected data points are stored as a two-dimensional array. The system uses a polynomial fitting algorithm to fit the data from both the heating and cooling stages, generating two mathematical function models describing the relationship between the resonant frequency and the equivalent Young's modulus. These models are then stored as lookup tables in the robot arm's controller. In subsequent operation, the state decoding and adaptive closed-loop control steps, based on whether the system is currently in a heating or cooling stage, call the corresponding function model and input the real-time fundamental frequency. The current equivalent Young's modulus of the drive line is calculated and used as the feedback signal for closed-loop control.

[0042] Example 5: When a flexible robotic arm performs a large-curvature bending posture transformation, the boundary conditions of its mechanical structure change accordingly, which may lead to mode aliasing of different orders of structural vibration modes, such as the first-order bending mode and the torsional mode, where their resonant frequencies approach or even cross. This condition affects the continuity and accuracy of state decoding. If the system only tracks the frequency based on the strongest resonant peak, it may lock onto the wrong mode during mode switching, causing the intrinsic state decoding based on the preset mapping relationship to fail. To address this situation, the method claimed in this invention is configured to execute a mode identification and locking procedure based on distributed excitation and response. In the excitation and information medium construction step, the system, at a preset time interval, such as... Every 100ms, a distributed excitation scan is performed, that is, the structural vibration excitation components are applied to different nickel-titanium shape memory alloy lines in turn according to a predetermined order, and each excitation lasts for 10ms. During each single excitation, the response picking and source separation steps synchronously monitor all unexcited nickel-titanium shape memory alloy lines to obtain an excitation-response relationship matrix. Different structural vibration modes exhibit specific response patterns in this matrix due to their inherent spatial mode shapes. For example, the first-order bending mode has the highest signal intensity when excited and responded on the same side, while the response intensity of the torsional mode is more evenly distributed in this matrix. These response patterns are identified and stored in the controller along with the identity tags of each mode in the initial calibration process of Example 3.

[0043] During the operation of the robotic arm, after obtaining the current resonance spectrum through spectrum analysis, the state decoding and adaptive closed-loop control steps do not immediately determine the strongest peak as the fundamental frequency. Instead, they first perform a modal identity matching. The system matches the real-time acquired excitation response relationship matrix with the pre-stored modal response modes, thereby assigning each significant resonance peak in the resonance spectrum its corresponding modal identity. Even when modal aliasing occurs, causing the resonance peak energy of the first-order bending mode, which was originally the tracking target, to no longer be the highest globally, the control system still continuously locks and tracks the frequency value of the specific modal resonance peak based on its set identity label to decode the intrinsic state. Through this procedure, the system's decoding basis for the current intrinsic state changes from a potentially abrupt energy maximum value to a specific structural vibration mode with clear physical meaning that is continuously identified and tracked.

[0044] Example 6: During a long-term cyclic life test of a flexible robotic arm used for detecting internal defects in pipelines, it needs to perform hundreds of thousands of reciprocating bending and stretching movements within a simulated pipeline. This process tests the long-term reliability of the nickel-titanium shape memory alloy wire used for drive and the mechanical connection structure. In the initial stage of the test, i.e., at 0 cycles, the system performs a complete self-test scan, records and stores the baseline resonance fingerprint representing its initial health state, and measures the quality factor of its first-order bending mode. The value is 35.2, and the average noise floor level in the 10Hz to 500Hz frequency band is -65dB. During the continuous execution of the predetermined bending and extending motion cycle by the robotic arm, the system utilizes the response signal inevitably generated by each motion control, and automatically records the current real-time resonant fingerprint every 10,000 cycles without affecting the main control flow, and stores it in non-volatile memory. When the test reaches 500,000 cycles, the recorded historical data shows that the quality factor of the first-order bending mode is 35.2. The value has shown a systematic downward trend, with a current value of 28.5. At the same time, the average noise floor level has risen to -58dB. The analysis module compares this real-time resonant fingerprint with the baseline resonant fingerprint to identify this irreversible systematic drift. In view of this, the system determines that there is a risk of fatigue accumulation in the drive line or that there is fretting looseness at the mechanical connection. Based on this, it automatically generates a predictive maintenance message, which prompts the operator to inspect or replace the boom segment.

[0045] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0046] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for precise multi-segment driving of a flexible robotic arm based on nickel-titanium shape memory alloy wires, wherein at least one segment of the flexible robotic arm is driven by a plurality of nickel-titanium shape memory alloy wires, characterized in that... The method includes: Step a, the excitation and information medium construction step, applies a composite driving signal containing macroscopic motion driving components and structural vibration excitation components to at least one first nickel-titanium shape memory alloy wire as the excitation source. The structural vibration excitation components are used to continuously excite structural modal resonance in the mechanical structure of the flexible robotic arm segment, thereby transforming the mechanical structure of the flexible robotic arm segment into a controlled vibration information medium that carries its intrinsic state and external interaction information in real time. Step b, the response pickup and source separation step, synchronously monitors at least one second nickel-titanium shape memory alloy wire, which is different from the first nickel-titanium shape memory alloy wire, and collects the periodic fluctuations of the electrical parameters of the second nickel-titanium shape memory alloy wire itself caused by structural modal resonance to obtain a response signal. The response signal is then subjected to spectrum analysis to obtain a resonance spectrum containing fundamental frequency and harmonic information. Based on the transient characteristics of the resonance spectrum within a time window, it is determined whether the change in resonance frequency is caused by intrinsic state changes or by external environmental contact. Step c, State Decoding and Adaptive Closed-Loop Control Step: When step b determines that the change in resonant frequency is caused by an intrinsic state change, the current intrinsic state is decoded based on the preset mapping relationship between the fundamental frequency of the resonant spectrum and the intrinsic phase transition state of the nickel-titanium shape memory alloy line, and motion closed-loop control is executed; when step b determines that the change in resonant frequency is caused by contact with the external environment, the motion closed-loop control is triggered to switch to the preset interactive control mode.

2. The multi-segment precision driving method for a flexible robotic arm based on nickel-titanium shape memory alloy wire according to claim 1, characterized in that, The structural vibration excitation component is a micro-amplitude AC signal superimposed on the macroscopic motion driving component used to generate macroscopic motion; the response pickup and source separation steps are specifically as follows: the response signal is obtained by monitoring the fluctuation of the resistance or voltage of the second nickel-titanium shape memory alloy wire, and the fundamental frequency peak of the resonance spectrum is determined by performing a fast Fourier transform on the response signal.

3. The multi-segment precision driving method for a flexible robotic arm based on nickel-titanium shape memory alloy wire according to claim 1, characterized in that, In the response pickup and source separation steps, the determination of whether the change in resonant frequency is caused by an intrinsic state change or by external environmental contact specifically involves calculating a transient impulse factor defined by the characteristics of the resonant spectrum. ,in ,in, The fundamental frequency value. The rate of change of the fundamental frequency value within a time window. This refers to the change in the energy ratio of harmonic information relative to the fundamental frequency. and The weighting coefficients are pre-calibrated based on the material and structure of the flexible robotic arm segment; when the transient impact factor... Exceeding the contact detection threshold preset based on experimental data When the change in resonant frequency is determined to be caused by contact with the external environment.

4. The multi-segment precision driving method for a flexible robotic arm based on nickel-titanium shape memory alloy wire according to claim 1, characterized in that, The current intrinsic state includes the phase transformation fraction or equivalent Young's modulus of one or more of the first nickel-titanium shape memory alloy wires and the second nickel-titanium shape memory alloy wires.

5. The multi-segment precision driving method for a flexible robotic arm based on nickel-titanium shape memory alloy wire according to claim 1, characterized in that, The method also includes a health status self-diagnosis step, which includes: storing one or more baseline resonance fingerprints representing the initial health status of the flexible manipulator, the baseline resonance fingerprints including the quality factor Q value of the resonance peak and the noise floor level of the non-resonant frequency band; continuously recording real-time resonance fingerprints determined by the response signal during the operation of the flexible manipulator; diagnosing the health status of the flexible manipulator by analyzing the long-term evolution trend of the real-time resonance fingerprints relative to the baseline resonance fingerprints, the long-term evolution trend including a systematic decrease in the quality factor Q value or a systematic increase in the noise floor level, and generating predictive maintenance information based on the diagnostic results.

6. The multi-segment precision driving method for a flexible robotic arm based on nickel-titanium shape memory alloy wire according to claim 5, characterized in that, Predictive maintenance information includes one or more of the following: drive line fatigue risk level or mechanical connection loosening risk level.

7. The multi-segment precision driving method for a flexible robotic arm based on nickel-titanium shape memory alloy wire according to claim 1, characterized in that, The method further includes: in the response pickup and source separation step, determining the phase difference between the structural vibration excitation component in the composite drive signal and the response signal; and determining a dissipation factor characterizing the energy dissipation characteristics of the flexible robotic arm segment based on the phase difference; judging whether the flexible robotic arm segment is in an active drive-dominated mode or a passive disturbance-dominated mode based on the degree of deviation between the real-time value of the dissipation factor and its intrinsic baseline value based on the current fundamental frequency; and adaptively adjusting the control parameters of the motion closed-loop control based on the judgment result.

8. The multi-segment precision driving method for a flexible robotic arm based on nickel-titanium shape memory alloy wire according to claim 1, characterized in that, The excitation and information medium construction steps include applying structural vibration excitation components to different nickel-titanium shape memory alloy lines in turn for distributed excitation; the response picking and source separation steps include synchronously monitoring multiple unexcited nickel-titanium shape memory alloy lines to obtain an excitation-response relationship matrix; the state decoding and adaptive closed-loop control steps also include: based on the excitation-response relationship matrix, identifying multiple different structural vibration modes of the flexible robotic arm segment, and continuously tracking the resonant frequency of a specific structural vibration mode with a pre-assigned identity.

9. A multi-segment precision driving method for a flexible robotic arm based on nickel-titanium shape memory alloy wire according to claim 8, characterized in that, Multiple different structural vibration modes include first-order bending mode and torsional mode; continuously track a specific structural vibration mode with a pre-assigned identity, specifically: when mode aliasing or mode switching causes the resonance peak of a specific structural vibration mode to no longer be the strongest resonance peak, still lock and track it according to its mode identity.

10. The multi-segment precision driving method for a flexible robotic arm based on nickel-titanium shape memory alloy wire according to claim 1, characterized in that, The method also includes a drive efficiency self-optimization step, which includes: continuously monitoring the heating current signal used to drive the macroscopic motion drive component of the first nickel-titanium shape memory alloy wire during the execution of motion closed-loop control; identifying the real-time heat exchange efficiency between the first nickel-titanium shape memory alloy wire and the surrounding environment online based on the statistical characteristics of the heating current signal within a time window, the statistical characteristics including the mean and variance of the heating current signal; and adaptively adjusting the control parameters of the motion closed-loop control according to the identified real-time heat exchange efficiency.

Images