Exoskeleton safety redundancy strategy and switching method based on anomaly detection

Abstract

The invention relates to an exoskeleton safety redundancy strategy and switching method based on anomaly detection, and the method comprises the steps: collecting a tension state between the active control intention of a user and the response execution of an exoskeleton system, defining as a potential anomaly when continuous asynchronous tension exists between the expected control direction of the user and the inertial behavior of the system; a redundancy strategy triggering precondition is adopted; after abnormal triggering, the system outputs a control suggestion, and whether the control right should be taken over by a redundancy strategy or not is judged by combining user reverse force application and posture deviation trend behavior performance, and master control continues to be maintained or a forced freezing state is entered; the intervention level of a redundancy control strategy is dynamically adjusted, and continuous control evolution from low intrusion to high takeover is achieved; if it is detected that an irreconcilable confrontation trend exists between continuous reverse intention input of the user and control output of the system in the control evolution process, and intervention ratio unbalance of the system is continuously expanded, the system enters a soft freezing or conservative attitude mode; and in the freezing state, judging whether the man-machine trust relationship is reconstructed or not.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of exoskeleton safety redundancy, in particular to an exoskeleton safety redundancy strategy and switching method based on anomaly detection. BACKGROUND

[0002] The prior art disclosed in Chinese patent document CN114129392A is an adaptive redundant drive exoskeleton rehabilitation robot capable of adjusting end fingertip force, and the core innovation point is that the fingertip force is collected through the finger connecting rod mechanism on the palm back platform, the steering engine is driven, and the pressure sensor is used, and then the control module is combined to perform adaptive switching of the active and passive modes, so that the stability of the gripping and the effectiveness of the training are ensured. However, it still has many deficiencies and drawbacks. First, the control logic of the scheme mainly focuses on the fingertip force and its duration, and its abnormality determination condition is single, which only depends on the setting of the fingertip force threshold and the safety threshold, and lacks multi-dimensional human-machine collaborative state recognition capability. In complex human-machine interaction scenarios, for example, when the user has motion delay, abnormal electromyographic signals, or joint motion mode mutation, simply relying on force threshold adjustment is easy to misjudge, which may lead to lag or even failure of redundant control intervention, and cannot make a pre-response to the potential high-risk state.

[0003] Secondly, although the invention proposes the switching of the active and passive modes, the switching mechanism is based on the threshold logic of binary, that is, reaching the threshold enters the active mode, and not reaching the threshold enters the passive mode. This mode conversion is too rigid and cannot reflect the gradual evolution process from low invasion to high intervention, which is easy to cause the user's intention to be suddenly deprived or over-authorized, and weakens the smoothness of human-machine interaction and the safety of the user. Thirdly, although the scheme emphasizes the realization of fingertip position compensation control through the disturbance observer (DOB), it is still a local optimization for the electrical and mechanical disturbance of the steering engine, lacks global analysis of the user's overall behavior trend, and cannot identify continuous indicators in action execution, such as trajectory repetition rate, direction consistency or speed change rate. As a result, the system can only correct after the end execution error occurs, but cannot predictively protect when abnormal signs appear. This after-compensation method has a lag in high-risk actions.

[0004] Fourth, the patent does not involve the dynamic maintenance of human-computer trust relationship, and the logic only stays in the mechanical judgment based on sensor data, that is, when the force exceeds the threshold, the training mode is suspended or switched. However, in the process of long-term rehabilitation training or complex task execution, the user's trust in the exoskeleton is an important factor affecting the use experience and rehabilitation effect. If the system frequently triggers or intervenes excessively, it will easily lead to the user's rejection. Fifth, the existing technology mainly designs for the local application scene of hand function rehabilitation, and its function is concentrated in finger grip training, lacking adaptability expansion to large-scale action scenes (such as gait assistance and lower limb exoskeleton control). Finally, in terms of flexibility of control strategy, the training mode selection of this scheme is biased towards static preset, and the user needs to train under fixed motion library and threshold conditions, and cannot realize dynamic adjustment according to individual differences or real-time state. Such a one-size-fits-all mode may lead to the risk of insufficient or excessive training for some patients. SUMMARY

[0005] The purpose of the present application is to provide an exoskeleton safety redundancy strategy and switching method based on anomaly detection, so as to solve some of the problems and deficiencies pointed out in the background art.

[0006] The technical scheme adopted by the present application to solve the above technical problems is as follows: an exoskeleton safety redundancy strategy and switching method based on anomaly detection, comprising: collecting the tension state between the user's active control intention and the response execution of the exoskeleton system, and when detecting that there is a persistent asynchronous tension between the user's expected control direction and the system's inertial behavior, defining it as a potential anomaly and as a redundancy strategy trigger precondition; After the anomaly is triggered, the system allows the main controller and the redundancy controller to output control suggestions at the same time, and combines the user's reverse force, posture offset trend behavior to judge whether the control right should be taken over by the redundancy strategy, continue to maintain the main control, or enter the forced frozen state; according to the explicitness of the user's current behavior and the response degree of the system intervention, a coupling relationship between behavior trend and control intervention degree is constructed, and the intervention level of the redundancy control strategy is dynamically adjusted to realize the continuous control evolution from low invasion to high takeover; If in the control evolution process, it is detected that there is an irreconcilable antagonistic trend between the user's continuous reverse intention input and the system control output, and the system intervention imbalance continues to expand, the control retreat is actively executed, and the system enters the soft frozen or conservative posture mode, and the further control instruction output is terminated; in the frozen state, the system judges whether the human-computer trust relationship is rebuilt by identifying whether the user's behavior shows active synchronicity or posture yielding action again.

[0007] Further, the tension state judgment includes: identifying the response time difference between the user's active force signal and the exoskeleton system's execution response, and the time difference exceeding the preset dynamic response threshold is considered as a potential abnormal state; when judging the abnormal state, the joint angle deviation, the muscle activation mode change and the motion response consistency are referred to for joint determination.

[0008] Further, in the parallel output stage of the main controller and the redundant controller, the control output weight of the two is dynamically allocated based on the execution trend of the user's action; when judging the control right attribution, the user's action compliance performance is referred to, and if the user continuously performs the action consistent with the output direction of the main controller within a unit time window, the priority of the main controller is maintained.

[0009] Further, the intervention level of the redundant control strategy is adjusted according to the user action trend stability index, which includes the action speed change rate, the trajectory repetition rate and the behavior direction consistency; when the redundant control strategy is intervened, the output intervention intensity is adjusted according to the overlap rate between the actual execution trajectory and the target trajectory.

[0010] Further, the action speed change rate is the standard deviation of the user's gait speed within a preset time window, and the redundant control strategy intervention level is increased by one level when the standard deviation continuously exceeds the set fluctuation threshold; the trajectory repetition rate is calculated by comparing the trajectory overlap area and the path difference of two consecutive action cycles, and if the repetition rate shows a downward trend in consecutive cycles, the redundant strategy level is increased.

[0011] Further, the behavior direction consistency is calculated by detecting the change of the included angle between the expected moving direction and the actual moving direction in each action cycle, and the average of the included angle change is taken as the consistency reference index; the initial level of the redundant control strategy is determined by the user's historical action trend stability score, which is the weighted average of multiple action stability indexes in the past control cycle.

[0012] Further, when calculating the overlap rate between the execution trajectory and the target trajectory, the actual trajectory is projected into the target trajectory corresponding time period for matching comparison based on the preset time synchronization window; the output intervention intensity adjustment takes the overlap rate as the reference signal, adopts the proportional regulation method to perform linear interpolation of the output value, and makes the controller output intensity change smoothly transition in consecutive multiple cycles; when the trajectory overlap rate is lower than the set threshold, and the user's behavior direction is continuously opposite to the system control output direction for more than two cycles, the redundant control strategy automatically enters the highest intervention level.

[0013] Further, the time synchronization window is asymmetric structure, the user trajectory is extended by multiple frames forward and backtracked by few frames backward, so that the projection matching focuses on the early trend of user behavior; the first derivative of the change rate is monitored in the continuous change of trajectory overlap rate, and when the downward trend of the overlap rate suddenly reverses or sharply steepens, the early promotion of redundant intervention level is triggered.

[0014] Further, before the actual trajectory is projected to the target trajectory, the time scale of the target trajectory is adaptively adjusted based on the current step frequency of the user, so that the rhythm is aligned in the high frequency motion state; when the overlap rate is lower than the historical average and the decline rate exceeds the set threshold, even if the set threshold is not reached, the redundant level preparation mode is started in advance.

[0015] Further, the opposite state of the user behavior direction and the control output direction is calculated by the unit change rate of the spatial direction difference vector, and if the rate exceeds the threshold, it is considered as a high mismatch risk segment; if the overlap rate oscillates rapidly, the system increases the change rate threshold of the linear interpolation output value to avoid unexpected jumps in output within a continuous period; Based on the above design, the discriminant function is used to identify the direction vector difference rate analysis and the sensitivity of the overlap rate change, which is used to control whether the system should perform strong intervention or limit output jump behavior, and its calculation formula is:

[0016] Wherein: The intervention sensitivity function value at the current time is used to determine whether to trigger redundant control transition or suppress interpolation jump; The unit time interval of the system control cycle; The spatial unit difference vector module value of the user intention direction and the system control direction at the current time, that is, the mismatch degree; The first derivative of the unit difference vector of the user intention direction in time; The normalized standard deviation value of the fluctuation degree of the current overlap rate change; The second derivative of the overlap rate in time, representing the acceleration characteristics of the oscillation trend; The function The derivation process includes: The user intention direction vector and the system control output direction vector are defined as unit vectors, on this basis, the difference vector between the two is calculated and the module value is taken, denoted as , to represent the direction mismatch degree; at the same time, to judge the evolution speed of this mismatch, the first derivative of the difference vector module value in time is calculated, which represents the direction mismatch rate, so that the first part of the risk measurement term is The item will increase rapidly when the behavior deviation of the user from the system is rapidly intensified, as an important dynamic index for the redundant control to cut in; On the other hand, in order to reflect the stability of the matching between the actual trajectory and the target trajectory in the system control process, the following is defined is the trajectory overlap rate at the current time (given by the main method); if the overlap rate appears sharp rise and fall changes in multiple periods, it will cause abnormal jump of the output interpolation, so the second-order derivative of the overlap rate is introduced to measure the acceleration of the trajectory matching trend; in order to improve the relative weight of the item in different states, the normalized standard deviation item of the overlap rate fluctuation is introduced The item reflects the disturbance amplitude of the trajectory matching curve in the recent period; after multiplying the two, the second part of the measurement item is obtained , which represents the system control instability risk caused by trajectory fluctuation; In summary, the above two risk items respectively depict the direction antagonistic risk and the trajectory disturbance risk, and the complete system intervention sensitive function can be constructed by linearly superimposing and normalizing per unit time; Finally, when the function value exceeds the system safety intervention threshold set in advance , the system will judge that the current state is a high-risk segment, and in either case of obvious direction mismatch or severe trajectory fluctuation, the highest level of redundant control intervention or the output change rate suppression mechanism can be triggered to ensure the safety of system operation and the stability of human-machine cooperation.

[0017] The beneficial effects of the present application are: the present application realizes dynamic safety guarantee and control continuity in the human-machine interaction process by introducing a safety redundant strategy and switching method based on anomaly detection in the exoskeleton control system. Through real-time collection and analysis of the tension state between the user's active control intention and the system's execution behavior, the present method can identify potential abnormalities at an early stage, including direction mismatch, action consistency decline, and trajectory overlap rate deterioration. Therefore, the system can not only establish the precondition for redundant control intervention before the anomaly occurs, but also respond according to the stability index of the user behavior trend and the historical behavior data, so that the control has a smooth transition from low intervention to high takeover, significantly improving the safety and reliability of the exoskeleton in complex motion environment.

[0018] Further, through trajectory projection comparison of time synchronization window, direction mismatch rate calculation and dynamic interpolation adjustment of trajectory overlap rate, the system avoids sudden jumps or unnecessary shocks in the control intensity adjustment process. When there is an irreconcilable antagonistic trend between user behavior and system control direction, the system can enter a frozen or conservative mode in time to prevent secondary risks caused by excessive intervention; and when the user shows synchronization or posture again, the control level can be gradually restored to achieve a balance between safety and flexibility. The present application not only improves the sensitivity and response speed of the exoskeleton system to abnormal states, but also establishes a recoverable human-machine trust closed loop, which helps to significantly improve the comfort and safety experience in the long-term wearing process. BRIEF DESCRIPTION OF DRAWINGS

[0019] Figure 1 The present application is an exoskeleton abnormality detection and redundancy switching main flowchart.

[0020] Figure 2 The present application is a multi-source signal fusion and redundancy dynamic intervention main flowchart.

[0021] Figure 3 The present application is a trajectory-direction dual sensitivity advance redundancy intervention relationship diagram.

[0022] Figure 4 The present application is an exoskeleton rehabilitation redundancy hierarchical control main flowchart of embodiment 1.

[0023] Figure 5 The present application is a step frequency self-adaptive and trajectory sensitive redundancy adjustment function flowchart of embodiment 2. DETAILED DESCRIPTION

[0024] The specific embodiments of the present application will be described in detail below with reference to the accompanying drawings.

[0025] The present application is an exoskeleton abnormality detection and redundancy switching main flowchart. Figure 1The application is based on an abnormality detection-based exoskeleton safety redundancy strategy and switching method. In the operation process of an exoskeleton system, the user's active control intention signal is collected in real time through sensors arranged on the user's body and key joints or driving units of the exoskeleton. The intention signal can be derived from the muscle activation pattern detected by the electromyography sensor, the movement trend obtained by the joint angle sensor, and the force sensor feedback generated by the user during the execution of the action. At the same time, the system synchronously acquires the response signal of the exoskeleton execution unit, including the angle position, movement speed, output torque and overall dynamics state of the execution joint, thereby establishing the coupling relationship between the user's active control intention and the execution behavior of the exoskeleton system. On the basis of the above coupling relationship, the tension state between the two is calculated, that is, whether there is a difference between the force or displacement trend corresponding to the user's expected control direction and the inertial response direction of the exoskeleton system. When the difference is non-synchronous and the tension is continuously accumulated in a continuous time window, the system determines that it is a potential abnormal state. The potential abnormal state serves as a precondition for triggering the subsequent redundancy strategy, which can ensure that the system enters the abnormal management process in time when the user's intention and the exoskeleton's response are out of sync, thereby providing a reliable basis for the intervention of the redundancy control strategy and avoiding control failure and safety risks caused by delayed detection or misjudgment.

[0026] After the abnormal state is triggered, the system enters a dual-controller parallel operation mode, that is, the main controller and the redundancy controller generate control suggestions at the same time in the same control cycle. The main controller derives the output according to the user's normal movement intention, while the redundancy controller generates a modified or safety-priority control scheme according to the abnormality detection mechanism. Then the system comprehensively judges the user's real-time generated counterforce signal, the trend characteristics of the overall posture deviation and the behavior consistency shown in the user's continuous action to determine whether the control right should be taken over by the redundancy strategy, the main control should be continued or the forced frozen state should be entered. In this judgment process, the system analyzes the explicitness of the user's action (such as the consistency degree of the action direction and speed) and the response matching degree of the system intervention output, and constructs the coupling relationship between the behavior trend and the control intervention degree. This coupling relationship is defined as a dynamic weight mapping model, which can adjust the intervention level of the redundancy control strategy according to the stability of the user's behavior and the output effect of the system, thereby achieving a balance between maintaining the user's initiative and ensuring the safety of the system. When the user shows explicit compliance action, the system remains in a low-invasive state, and the redundancy controller only makes slight modifications. When the user's action is uncertain or continuously conflicts with the system control, the intervention level of the redundancy strategy gradually increases, gradually transitioning from a low-invasive state to a high-takeover state, forming a continuous control evolution mechanism that avoids the control abruptness caused by a single jump.

[0027] In the control evolution process, the system continuously monitors the matching relationship between the user's intention input signal and the exoskeleton execution unit output. When it is detected that the user's reverse intention input and the system control output always exist in the opposite direction or mechanically counteract each other in multiple consecutive control periods, and through comprehensive calculation it is found that this antagonistic trend is irreconcilable and manifests as the system intervention imbalance gradually, and the difference between the user input and the system output continues to expand, the system will actively trigger the control retreat mechanism, that is, terminate the further intervention of the redundant control strategy, make the exoskeleton system from the normal working mode into the soft frozen state or the conservative posture mode, in which the controller stops applying new driving instructions to the execution joint, only maintains the minimum necessary posture support to avoid user falling or causing danger, while reducing the joint output torque or movement speed to ensure safety; in the frozen state, the system will not immediately restore normal work, but through sensors to continuously identify whether the user re-exhibits active synchronous action consistent with the system output direction, or whether it re-accepts the control right of the exoskeleton through the body natural posture, when the above conditions are met, the system determines that the human-machine trust relationship has been re-established, then gradually removes the frozen state, restores the normal operation of the control strategy, so as to ensure that when the user and the exoskeleton have an uncoordinated conflict, the risk can be reduced in time, and after the trust relationship is rebuilt, it is smoothly transitioned back to the normal interactive control process.

[0028] Combining the drawings Figure 2 Wherein the judgment process of the tension state is completed by multi-source signal fusion analysis, the system collects the user's active force signal in real time through the electromyographic sensor arranged at the user's muscle group position, and records the execution response signal of the exoskeleton joint driving unit at the same time, when the response time difference between the two exceeds the dynamic response threshold set by the system, it is determined that there is a potential abnormal state, in order to avoid misjudgment caused by a single indicator, the system further introduces at least three joint determination indicators for comprehensive analysis: the first is the joint angle deviation index, by comparing the difference between the real-time angle of the exoskeleton execution joint and the user's expected angle or action template, if it exceeds the tolerance range, it is counted as an abnormal factor; the second is the electromyographic activation mode change index, by comparing the changes in amplitude distribution, activation timing and frequency domain characteristics of the current electromyographic signal, whether there is a significant deviation from the normal action mode is identified; the third is the motion response consistency index, the system calculates the degree of coincidence between the user's expected motion trajectory and the actual execution trajectory of the exoskeleton in time and space, if the consistency significantly decreases, it is determined as a potential risk; by comprehensively considering the above multiple indicators, the system adopts a weighted judgment method, when the response time difference exceeds the threshold and two or more of the three indicators meet the abnormal conditions, it is finally confirmed to enter the potential abnormal state, thereby providing a reliable basis for the triggering of the redundant control strategy.

[0029] In the parallel output stage of the main controller and the redundant controller, the system assigns the control output weight of both by dynamically monitoring the trend of user action. In the same control cycle, the main controller generates a normal output signal based on the user's expected action, and the redundant controller generates a compensation or correction signal based on the safety decision. The system uses the user action trend as the basis for weight adjustment. When the user action has clear continuity and directionality, the weight of the main controller is increased to ensure that the user's intention is prioritized. When the user action shows uncertainty or significant deviation from the expected trajectory, the weight of the redundant controller is increased to enhance the safety intervention capability. In the control right attribution judgment process, the system further introduces user action compliance as an auxiliary reference. By collecting the user's electromyographic signal pattern, joint angle change, and the matching degree of the execution trajectory and the main controller output direction within a unit time window, the system determines the compliance level. If the user continuously performs actions consistent with the direction of the main controller output within the time window, the main controller priority remains unchanged, and the redundant controller is in a low weight state, only making slight corrections when necessary. This ensures maximum user dominance while maintaining safety redundancy, achieving a reasonable allocation and dynamic switching of control rights.

[0030] The intervention level of the redundant control strategy is not fixed, but is dynamically adjusted based on the stability index of the user's action trend. The stability index includes at least the action speed change rate, the trajectory repetition rate, and the behavior direction consistency. The action speed change rate is calculated by collecting user joint motion data and calculating the standard deviation of gait or limb motion speed within a preset time window. This index reflects whether the user's action is volatile. The trajectory repetition rate is obtained by comparing the motion trajectories of multiple consecutive action cycles and calculating the overlap area and path difference. This index describes the repeatability and regularity of user motion execution. The behavior direction consistency is calculated by detecting the change in the angle between the expected motion direction and the actual motion direction and taking the average. This index reflects whether the user's control intention and actual execution maintain consistency in direction. When any of the above indices shows a decrease in stability, the system gradually increases the intervention level of the redundant control strategy according to the set grading threshold, ensuring that the system can timely enhance assistance or safety intervention when the user's action trend is unstable. Meanwhile, during the intervention stage of the redundant control strategy, the system also dynamically adjusts the output intervention intensity based on the overlap rate between the user's actual execution trajectory and the target trajectory. When the overlap rate is high, the system maintains low-intensity correction to avoid excessive intervention. When the overlap rate continues to decrease and the deviation expands, the system gradually increases the control output intensity, ensuring a smooth transition from light correction to deep takeover of the redundant control. This is a graded intervention control based on the dual judgment of user behavior stability and trajectory deviation.

[0031] In which the action speed change rate and trajectory repetition rate are used as important basis for determining the dynamic adjustment of the intervention level of the redundant control strategy. During the operation of the system, firstly, the user's gait speed data is statistically processed within a preset time window, and the standard deviation of the speed value within the time window is calculated to obtain the action speed change rate. When the standard deviation continuously exceeds the system set fluctuation threshold in continuous multiple time windows, it indicates that the user's gait speed has large fluctuation and instability, and the system will increase the intervention level of the redundant control strategy by one level to enhance the auxiliary control strength, so as to intervene in time during the unstable stage of user's action to ensure safety. At the same time, the system also calculates the trajectory repetition rate by comparing the motion trajectories of two consecutive action cycles of the user. The repetition rate is composed of the trajectory overlap area and the path difference. If the trajectory repetition rate shows a downward trend in continuous multiple action cycles, it is determined that the user's action execution regularity decreases and stability weakens. The system accordingly increases the redundant strategy level, so that the exoskeleton control strategy gradually transitions from light correction to higher level deep intervention, so as to provide stronger motion assistance and safety redundancy guarantee in the case of insufficient action repetition, and dynamically adjusts the degree of redundant intervention according to the actual motion stability of the user.

[0032] In which the behavior direction consistency is used as an important index for determining the reliability of the user's control intention. The angle formed between the user's expected moving direction and the actual moving direction of the exoskeleton system is detected cycle by cycle within each action cycle, and the average value of the angle in continuous cycles is calculated as the reference basis for judging the direction consistency. When the average angle value is small, it indicates that the user's intention is highly consistent with the system execution, and the system maintains a low intervention control state. When the angle continuously increases, it indicates that there is a deviation trend between the user's control intention and the system response, and the system accordingly increases the intervention level of the redundant strategy to enhance the assistance or correction ability. In addition, in order to further improve the adaptability and individualization ability of the system in the initialization stage, the initial level of the redundant control strategy is determined by the user's historical action trend stability score. The score is obtained by weighted average calculation of multiple action stability indicators recorded by the user in the past multiple control cycles, including but not limited to action speed change rate, trajectory repetition rate and behavior direction consistency, etc. The system presets the redundant intervention level based on the stability score. If the score is high, the initial intervention level of the system is low to maintain user dominance. If the score is low, the system initially increases the proportion of redundant control to enhance safety protection.

[0033] In calculating the overlap rate between the user's execution trajectory and the system target trajectory, to ensure the consistency of the comparison process in the time dimension, the system sets a time synchronization window, projects the user's actual trajectory within the window according to the time index into the corresponding time period of the target trajectory for matching comparison. Through this time alignment method, trajectory mismatch caused by differences in action rhythm can be avoided, thus more accurately reflecting the degree of coincidence between the user's action and the target expectation. On this basis, the system takes the overlap rate as the reference signal of the output intervention intensity, and uses proportional regulation method to linearly interpolate the output value. Specifically, when the overlap rate is high, the interpolation calculation result corresponds to a lower output intensity, so that the redundant control remains low intervention. When the overlap rate decreases, the interpolation result gradually increases, and the output intensity increases accordingly, thus ensuring the smooth transition of the output of the controller in continuous multiple cycles, avoiding sudden mechanical jumps that cause discomfort or danger to the user. At the same time, the system jointly determines the abnormal state. When the trajectory overlap rate continuously falls below the set threshold, and the user's behavior direction and the system's control output direction are always in opposite states in more than two consecutive action cycles, the system immediately identifies it as a high-risk state and triggers the highest level of redundant control intervention. At this level, the redundant controller completely takes over the main output, and the main controller only retains the monitoring function, to ensure that when there is a serious antagonistic trend between the user and the system, safety protection can be quickly established and the risk can be minimized.

[0034] In combination with the accompanying Figure 3 In the matching process of the execution trajectory and the target trajectory, an asymmetric time synchronization window is used. This time window is designed to extend the user's trajectory data forward by multiple frames and backtrack by a small number of frames. That is, based on the current matching time point, the system preferentially introduces predictive trajectory data of the user's future frames, while retaining a small amount of historical frames for correction reference. In this way, an advance capture mechanism for user behavior trends is established. This structure makes the trajectory projection matching more biased towards the user's advance response capability, thus more effectively identifying potential intention deviations or action abnormalities. In the process of continuously monitoring the trajectory overlap rate, the system further calculates the first derivative of the overlap rate change, and captures the trend of the change rate in real time. When the system detects a sudden change in the first derivative value, which is manifested as a reversal of the downward trend of the overlap rate in a short time (from falling to rising) or a sharp steepening (rapid falling in a short time), it is determined that there is a potential unstable fluctuation risk in the trajectory matching. The system regards this as a precursor to the action mismatch state, and immediately triggers the redundant control strategy to advance the intervention level, so that the system can complete the control right transfer and auxiliary mechanism upgrade before the mismatch develops into a serious abnormality.

[0035] Before the actual trajectory is projected onto the target trajectory, the system detects the user's movement rhythm characteristics. It obtains the user's current step frequency through periodic analysis of cadence sensors or joint angular velocities and uses this as an adaptive adjustment parameter to dynamically compress or expand the time scale of the target trajectory. When the user is in a high-frequency movement state, the target trajectory is appropriately compressed to maintain consistency with the user's movement rhythm, thereby ensuring improved time alignment accuracy between the two parties during trajectory matching and avoiding matching deviations caused by frequency differences. On this basis, in the overlap rate monitoring stage, the system not only judges whether the current overlap rate is lower than the set threshold, but also introduces the historical average value and the rate of decline as joint judgment conditions. When the overlap rate has not reached the minimum threshold but is lower than the historical average value and its rate of decline exceeds the system's preset threshold, the system considers that the deviation between the user trajectory and the target trajectory has a developing trend. Therefore, it activates the redundancy level preparation mode in advance. In this mode, the redundant controller gradually increases the output ratio and enters the takeover state, thereby completing the defensive deployment before a real serious mismatch occurs.

[0036] When the user's behavior direction conflicts with the system's control output direction, the system calculates the unit rate of change of the spatial direction difference vector. If this rate exceeds a preset threshold, it is identified as a high-risk mismatch segment, triggering the early intervention of redundancy strategies. Simultaneously, in trajectory overlap rate monitoring, when the overlap rate oscillates rapidly over multiple consecutive cycles, the system adds a rate of change threshold to the original linear interpolation output value to prevent unexpected large jumps in the output control quantity within a short period. Based on the above design, the system constructs a discriminant function that integrates direction vector difference rate analysis and overlap rate change sensitivity identification. This function uniformly measures whether strong intervention control should be triggered or the jump in output interpolation should be limited in the current state. The specific calculation formula is as follows:

[0037] in, For the current moment The intervention-sensitive function value is used to determine whether redundant control transitions or interpolation suppression need to be triggered. The unit time interval of the system control cycle; It represents the magnitude of the spatial unit difference vector between the user's intended direction and the system's control direction, i.e., the degree of direction mismatch; The first derivative of the difference vector in time reflects the rate of directional mismatch. This is the normalized standard deviation of the fluctuation in trajectory overlap rate, used to characterize trajectory stability; Trajectory overlap rate The second order derivative in time, representing the acceleration feature of the trajectory oscillation tendency. The derivation process includes: first, normalize both the user intention direction vector and the system output direction vector and calculate the difference vector, take the modulus to get , and combine its time derivative , get the first part of the risk term , which is used to reflect the aggravation trend of direction mismatch in time dimension; second, in the trajectory matching dimension, define as the real-time overlap rate of the executed trajectory and the target trajectory, and reflect the oscillation acceleration of the overlap rate change through its second order derivative , at the same time, take the normalized standard deviation of trajectory fluctuation as the weight correction term, form the second part of the risk measure ; finally, superimpose the above two risk measures according to the time step, form the complete intervention sensitivity function , when exceeds the set safe intervention threshold , the system will determine that the current is a high risk segment, trigger the highest level of redundant control strategy to take over or impose a speed suppression mechanism on the interpolated output, so as to realize the dual monitoring and safety control of direction mismatch and trajectory fluctuation.

[0038] Embodiment 1: combined with the attached Figure 4In this embodiment, the system real-time collects the electromyography signals of quadriceps femoris and hamstrings of a rehabilitation patient during the lower extremity exoskeleton-assisted walking training, and monitors the lower extremity movement through joint angle sensor and inertial measurement unit. When the patient actively generates a force signal during the stepping process, the system detects that the electromyography amplitude of quadriceps femoris rapidly increases within 200 ms, showing obvious starting movement intention, while the driving response of the exoskeleton knee joint produces corresponding angle change only after 450 ms, resulting in a response time difference of 250 ms, which exceeds the preset dynamic response threshold of 200 ms of the system, at this time the system determines that there is a potential abnormality. Further, the system combines three indicators for joint determination: first, in terms of joint angle deviation, the patient expects the knee joint flexion angle to be 30°, while the exoskeleton actually executes only 22°, with a deviation of 8°, which exceeds the set tolerance range of 5°; second, in terms of electromyography activation mode, the patient's hamstrings signal appears abnormal early peak value during the flexion and extension coordination period, resulting in a time sequence offset of 18% compared with the normal gait database, indicating that the activation mode is abnormal; third, in terms of movement response consistency, the patient expects the stepping direction to be 10° forward, while the system actually executes the direction to be 2° forward, with an average deviation of 8° between the two, and the consistency is lower than the standard lower limit of 85%. Two of the three indicators exceed the threshold, so the system finally confirms that it enters an abnormal state. At this time, the system enters the parallel output stage of the main controller and the redundant controller, and sets the main controller to generate a joint output torque of 12 Nm according to the patient's intention, while the redundant controller generates a conservative output torque of 16 Nm according to the abnormality determination, and the system dynamically allocates weights according to the user's action execution trend, with the main controller weight set to 0.6 and the redundant controller weight set to 0.4 in the initial stage, and the synthesized output is 13.6 Nm. However, in the subsequent 3 control cycles, the system observes that the user's electromyography signal and actual joint movement direction remain consistent, for example, within a unit time window of 1 s, the patient's expected knee joint flexion direction and the output direction of the main controller maintain a consistency of above 95%, and the electromyography activation mode restores to a normal gait database matching degree of 92%, and the joint angle deviation gradually reduces to 3°, indicating that the user's action compliance performs well. In this case, the system maintains the priority of the main controller, increases its weight to 0.8, and reduces the weight of the redundant controller to 0.2, and finally the synthesized output value returns to 12.8 Nm, to ensure the dominance of the user's intention and reduce unnecessary intervention of the redundant control.

[0039] As the patient gradually enters the continuous walking phase, the exoskeleton system begins to monitor the user motion trend stability indicators in real time in order to dynamically adjust the intervention level of the redundant control strategy. In terms of motion speed change rate, the system records the knee flexion and extension speed data of the patient with a preset time window of 2s, and calculates the gait speed standard deviation to be 0.42m / s, while the set fluctuation threshold is 0.35m / s. The standard deviation exceeds the threshold for three consecutive time windows, indicating that the patient's pace rhythm fluctuates significantly, and the system therefore increases the redundant strategy level from level 1 to level 2, increasing the proportion of the redundant controller in the output weight from 20% to 40%, enhancing the intervention for gait instability. Secondly, in terms of trajectory repetition rate, the system matches the hip-knee-ankle joint trajectory of the patient for two consecutive gait cycles, calculates the overlap area to be 142cm², the path difference to be 36cm, and the overall repetition rate to be 74%, while the repetition rate of the previous cycle is 82%, and the repetition rate shows a downward trend in the next three cycles, with a cumulative decrease of 12%. The system therefore increases the redundant strategy level from level 2 to level 3, and further adjusts the output weight proportion to 50% for the main controller and 50% for the redundant controller, in order to significantly increase the protective effect of redundant control.

[0040] Finally, in terms of behavior direction consistency, the system detects the change in the angle between the expected walking direction and the actual execution direction within a single step cycle, calculates the average angle to be 11°, which is lower than the set consistency reference threshold of 15°, and determines that the direction consistency is acceptable, so it does not trigger additional level increase. During the intervention of the redundant control strategy, the system also dynamically adjusts the output intervention intensity by referring to the overlap rate of the actual execution trajectory and the target trajectory, for example, the target trajectory planning step is 0.65m, while the patient's actual step is 0.58m, the overlap rate is 89%, and the output intervention intensity is adjusted to 0.3 times the redundant correction. When the next overlap rate drops to 76%, the output intervention intensity is increased to 0.6 times the redundant correction, thereby achieving smooth incremental intervention intensity.

[0041] The exoskeleton system further combines the behavior direction consistency and the history stability score to optimize the dynamic response of the redundant control strategy. In the initial training stage, the system statistically analyzes the motion stability of the patient according to the control cycle data of the past 5 minutes, calculates the average value of the motion speed change rate as 0.28 m / s (threshold value 0.35 m / s), the average value of the trajectory repetition rate as 81% (threshold value 75%), and the average included angle of the behavior direction consistency as 9° (threshold value 15°), and sets the speed weight as 0.4, the trajectory weight as 0.35, and the direction weight as 0.25 according to the weighted average formula. Finally, the stability score is 0.86 (full score 1.0), and the system sets the initial level of the redundant control strategy as level one, that is, the output proportion of the redundant controller is only 20%, and the main controller remains dominant to ensure the user's autonomy. In the subsequent single-step motion cycle, the system real-time detects the included angle change between the user's expected moving direction and the actual executing direction, for example, in a 1.2s gait cycle, the expected direction is forward 12°, the actual direction fluctuates in the range of 14°~17°, the average included angle deviation is calculated as 4.2°, which is less than the reference threshold value 15°, and the direction consistency is good, so the current redundant level is kept unchanged. However, in the next cycle, the patient's expected direction is forward 10° due to fatigue, and the actual direction deviates to forward -5°, the included angle change reaches 15°, which is greater than the threshold value for two consecutive cycles, and the direction consistency decreases, so the system triggers the level promotion, and the redundant control level is upgraded to level two, with the proportion adjusted to 40%. At the same time, the system enters the overlap rate calculation process of the execution trajectory and the target trajectory, and in a 3s time synchronization window, the actual trajectory points of the patient are projected to the target trajectory corresponding to the time period for matching and comparison, and the current overlap rate is 72%, which is lower than the set threshold value 75%. According to the proportional adjustment method, the output intervention intensity is modified from the original 0.3 times to 0.6 times. In order to ensure continuity, the system uses the linear interpolation method to gradually increase the intervention intensity in the next three cycles instead of jumping at once, for example, the first cycle is increased to 0.4 times, the second cycle is 0.5 times, and the third cycle is 0.6 times, so as to avoid the output torque from producing an unexpected large jump in a short time. If the user's behavior direction and the system's control output direction are still opposite in the next two cycles, for example, the user continuously intends to move backward while the exoskeleton control output is still forward, the system determines that it is in a serious confrontation state, and under the condition that the trajectory overlap rate is insufficient and the direction is opposite, the redundant control strategy is switched to the highest level, the redundant controller output weight is increased to 100%, and the motion control is forcibly taken over, and the exoskeleton is locked in the safe and conservative mode to ensure that the patient will not fall or be injured due to the conflict between man and machine.

[0042] Example 2: Combined with the Figure 5When the patient's rehabilitation training gradually enters the complex motion stage, the exoskeleton system introduces the asymmetric structure of the time synchronization window and the step frequency self-adaptive adjustment mechanism to realize more forward-looking and more sensitive redundant control. When the patient is walking at a fast pace, the average step frequency is increased to 1.9 Hz, while the original rhythm of the system target trajectory is 1.5 Hz. At this time, the system first adaptively compresses the time scale of the target trajectory, and the scaling coefficient is calculated as 1.9 ÷ 1.5 = 1.27, that is, the time axis of the target trajectory is compressed by 27%, so that the rhythm planned by the system remains aligned with the high-frequency motion of the user, avoiding system lag in the trajectory matching process. Subsequently, in the trajectory comparison link, the system uses an asymmetric time synchronization window, with 5 frames of future data as forward extension and only 2 frames of data as correction, so that the trajectory projection is more biased towards the advance trend of predicting the patient's action. For example, in a certain period, the actual step trajectory of the patient appears 150 ms ahead of the target trajectory, and the system effectively captures this advance trend through forward extension, and the trajectory overlap rate after matching is 84%, which is about 9% higher than the traditional symmetric window, indicating that the capture of the advance trend improves the matching accuracy. In the continuous monitoring process, the system calculates the first derivative of the change of the trajectory overlap rate, when the overlap rate continuously decreases from 88% to 73% in 3 cycles, the first derivative is -5% / cycle, and then suddenly rebounds to 79% in the 4th cycle, the derivative changes from negative to positive and the change amplitude is more than 7% / cycle, the system determines that it is the reversal phenomenon of the downward trend, immediately triggers the redundant intervention level to increase from level 2 to level 3, so that the weight of the redundant controller increases from 40% to 60%, to prevent the action from being enlarged in advance. On the other hand, in the high-frequency fast walking state, although the overlap rate is not lower than the set threshold of 70%, but it sharply decreases from the historical average of 85% to 74%, with a decrease rate of 11% / cycle, which exceeds the set threshold of 8% / cycle, the system immediately starts the redundant level preparation mode, and increases the redundant output ratio from 30% to 50% in the next cycle, and gradually increases through linear interpolation in two control cycles, avoiding sudden intervention. When the patient appears obvious mismatch trend due to fatigue, the system has entered the high-level preparation in advance, ensuring the smooth connection and safe takeover of the exoskeleton action.

[0043] The exoskeleton system enters the high-dynamic monitoring stage, at this time the patient tries to speed up the step frequency to 2.1 Hz, but due to insufficient lower limb muscle strength, the motion stability decreases, resulting in a continuous deviation between the user's intended direction and the system's control direction. Set the user's expected direction vector at time , , the system's output direction vector is , the difference vector between the two is , and the modulus is . In the next 3 control cycles, the modulus of the direction difference vector increases from 0.45 to 0.63, and then to 0.88, and the first derivative is approximately calculated as , set , then . Thus the first part of the risk term is .

[0044] On the other hand, the system monitors the trajectory overlap rate , whose values in the last 5 cycles are 0.88, 0.81, 0.69, 0.76 and 0.62 respectively, indicating that there is obvious oscillation. The second derivative approximation is calculated as

[0045] For example, at , substituting , we get

[0046] At the same time, the normalized standard deviation of the recent window is calculated, assuming that the mean of the trajectory overlap rate is 0.752 and the standard deviation is 0.095, and after normalization, we take . Then the second part of the risk term is

[0047] In summary, the two parts of the metric are substituted into the intervention sensitivity function:

[0048] Set the safety intervention threshold of the system setting to , then the current calculation value , the system immediately determines that it is a high mismatch risk segment, triggering the highest level of redundant control intervention. At this time, the weight of the redundant controller is increased from 50% to 100%, forcibly taking over the motion output, and at the same time starting the interpolation output rate threshold to limit the output rate to within 10% / cycle, avoiding unexpected torque jumps caused by oscillation of the overlap rate. Under this protection, although the patient has obvious directional resistance and trajectory fluctuations, the system realizes sensitive capture and rapid response to sudden risks through formula determination and real-time intervention, ensuring the safety and stability of the patient in high-frequency fast-step rehabilitation training.

[0049] The above shows and describes the basic principles, main features and advantages of the present application. Those skilled in the art should understand that the present application is not limited to the above examples, and the above examples and descriptions in the specification are only to illustrate the principles of the present application. Without departing from the spirit and scope of the present application, various changes and improvements can be made to the present application, and these changes and improvements all fall within the scope of the claimed present application. The scope of protection of the present application is defined by the appended claims and their equivalents.

Claims

1. An exoskeleton safety redundancy strategy and switching method based on anomaly detection, characterized in that The method comprises: Collecting the tension state between the user's active control intention and the response execution of the exoskeleton system, and defining it as a potential anomaly when detecting that there is persistent asynchronous tension between the user's expected control direction and the system's inertial behavior, and serving as a trigger precondition for the redundant strategy; After the anomaly is triggered, the system allows the main controller and the redundant controller to output control suggestions at the same time, and combines the user's reverse force, posture offset trend behavior to judge whether the control right should be taken over by the redundant strategy, continue to maintain the main control, or enter the forced frozen state; According to the explicitness of the user's current behavior and the response degree of the system intervention, a coupling relationship between behavior trend and control intervention degree is constructed to dynamically adjust the intervention level of the redundant control strategy, and continuous control evolution from low invasion to high takeover is realized; If it is detected that there is an irreconcilable confrontation trend between the user's continuous reverse intention input and the system control output, and the system intervention imbalance continues to expand, the control retreat is actively executed, the system enters the soft frozen or conservative posture mode, and the further control instruction output is terminated; In the frozen state, the system judges whether the human-machine trust relationship is rebuilt by identifying whether the user's behavior re-exhibits active synchrony or posture yielding action.

2. The abnormality detection based exoskeleton safety redundancy strategy and switching method according to claim 1, characterized in that The judgment of the tension state includes: identifying the response time difference between the user's active force signal and the exoskeleton system execution response, and regarding the time difference exceeding the preset dynamic response threshold as a potential abnormal state; When judging the abnormal state, at least three index joint determination results are referred to, including: joint angle deviation, electromyographic activation mode change and motion response consistency.

3. The abnormality detection based exoskeleton safety redundancy strategy and switching method of claim 1, wherein In the parallel output stage of the main controller and the redundant controller, the control output weight of the two is dynamically allocated based on the execution trend of the user's action; When judging the control right attribution, the action compliance performance of the user is referred to, if the user continuously executes the action consistent with the output direction of the main controller in a unit time window, the priority of the main controller is maintained.

4. The abnormality detection based exoskeleton safety redundancy strategy and switching method of claim 1, wherein The intervention level of the redundant control strategy is adjusted according to the user action trend stability index, which includes the action speed change rate, the trajectory repetition rate and the behavior direction consistency; When the redundant control strategy intervenes, the output intervention intensity is adjusted according to the overlap rate between the actual execution trajectory and the target trajectory of the user.

5. The abnormality detection based exoskeleton safety redundancy strategy and switching method according to claim 4, characterized in that The action speed change rate is the standard deviation of the user's gait speed in a preset time window, and the intervention level of the redundant control strategy is increased by one level when the standard deviation continuously exceeds the set fluctuation threshold; The trajectory repetition rate is calculated by comparing the trajectory overlap area and the path difference of two consecutive action periods, if the repetition rate shows a downward trend in consecutive periods, the level of the redundant strategy is increased.

6. The abnormality detection based exoskeleton safety redundancy strategy and switching method according to claim 4, characterized in that The behavior direction consistency is calculated by detecting the change of the included angle between the expected moving direction and the actual moving direction in each action period, and the average of the included angle change is taken as the consistency reference index; The initial level of the redundant control strategy is determined by the user's historical action trend stability score, which is the weighted average of multiple action stability indexes in the past control period.

7. The abnormality detection based exoskeleton safety redundancy strategy and switching method according to claim 4, characterized in that In the calculation of the overlap rate between the execution trajectory and the target trajectory, the actual trajectory is projected into the target trajectory corresponding time period for matching comparison based on a preset time synchronization window; The output intervention intensity adjustment takes the overlap rate as a reference signal, adopts a proportional adjustment method to perform linear interpolation on the output value, and makes the controller output intensity change smoothly transition within continuous multiple periods; when the trajectory overlap rate is lower than a set threshold and the user behavior direction and the system control output direction are continuously opposite for more than two periods, the redundant control strategy automatically enters the highest intervention level.

8. The abnormality detection based exoskeleton safety redundancy strategy and switching method according to claim 7, characterized in that The time synchronization window is of an asymmetric structure, the user trajectory is extended by multiple frames forward and backtracked by few frames backward, the projection matching focuses on the early trend of user behavior; the first-order derivative change rate is monitored in the continuous change of the trajectory overlap rate, and when the downward trend of the overlap rate suddenly reverses or sharply steepens, the redundant intervention level is triggered to be raised in advance.

9. The abnormality detection based exoskeleton safety redundancy strategy and switching method according to claim 7, characterized in that Before the actual trajectory is projected to the target trajectory, the target trajectory time scale is adaptively adjusted based on the current step frequency of the user, so that the rhythm is aligned in the high-frequency motion state; when the overlap rate is lower than the historical average and the descending rate exceeds the set threshold, even if the set threshold is not reached, the redundant level preparation mode is started in advance.

10. The abnormality detection based exoskeleton safety redundancy strategy and switching method according to claim 7, characterized in that The opposite state of the user behavior direction and the control output direction is calculated by the unit change rate of the spatial direction difference vector, and if the rate exceeds the threshold, it is considered as a high mismatch risk segment; if the overlap rate rapidly oscillates, the system increases the change rate threshold of the linear interpolation output value to avoid unexpected jumps in the output within continuous periods.

Images