EP3705105A1 - Steuerungssystem für ein haptisches exoskelett der unteren extremitäten zur rehabilitation oder zum gehen mit verbesserter gleichgewichtskontrolle, mensch-maschine-schnittstelle - Google Patents

Steuerungssystem für ein haptisches exoskelett der unteren extremitäten zur rehabilitation oder zum gehen mit verbesserter gleichgewichtskontrolle, mensch-maschine-schnittstelle Download PDF

Info

Publication number
EP3705105A1
EP3705105A1 EP19425014.8A EP19425014A EP3705105A1 EP 3705105 A1 EP3705105 A1 EP 3705105A1 EP 19425014 A EP19425014 A EP 19425014A EP 3705105 A1 EP3705105 A1 EP 3705105A1
Authority
EP
European Patent Office
Prior art keywords
postural
control
vector
exoskeleton
bioelectric
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Withdrawn
Application number
EP19425014.8A
Other languages
English (en)
French (fr)
Inventor
Giuseppe Menga
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Syco Di Menga Giuseppe & C Sas
Original Assignee
Syco Di Menga Giuseppe & C Sas
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Syco Di Menga Giuseppe & C Sas filed Critical Syco Di Menga Giuseppe & C Sas
Priority to EP19425014.8A priority Critical patent/EP3705105A1/de
Publication of EP3705105A1 publication Critical patent/EP3705105A1/de
Withdrawn legal-status Critical Current

Links

Images

Classifications

    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61H—PHYSICAL THERAPY APPARATUS, e.g. DEVICES FOR LOCATING OR STIMULATING REFLEX POINTS IN THE BODY; ARTIFICIAL RESPIRATION; MASSAGE; BATHING DEVICES FOR SPECIAL THERAPEUTIC OR HYGIENIC PURPOSES OR SPECIFIC PARTS OF THE BODY
    • A61H3/00—Appliances for aiding patients or disabled persons to walk about
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61H—PHYSICAL THERAPY APPARATUS, e.g. DEVICES FOR LOCATING OR STIMULATING REFLEX POINTS IN THE BODY; ARTIFICIAL RESPIRATION; MASSAGE; BATHING DEVICES FOR SPECIAL THERAPEUTIC OR HYGIENIC PURPOSES OR SPECIFIC PARTS OF THE BODY
    • A61H1/00—Apparatus for passive exercising; Vibrating apparatus ; Chiropractic devices, e.g. body impacting devices, external devices for briefly extending or aligning unbroken bones
    • A61H1/02—Stretching or bending or torsioning apparatus for exercising
    • A61H1/0237—Stretching or bending or torsioning apparatus for exercising for the lower limbs
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61H—PHYSICAL THERAPY APPARATUS, e.g. DEVICES FOR LOCATING OR STIMULATING REFLEX POINTS IN THE BODY; ARTIFICIAL RESPIRATION; MASSAGE; BATHING DEVICES FOR SPECIAL THERAPEUTIC OR HYGIENIC PURPOSES OR SPECIFIC PARTS OF THE BODY
    • A61H2201/00—Characteristics of apparatus not provided for in the preceding codes
    • A61H2201/16—Physical interface with patient
    • A61H2201/1602—Physical interface with patient kind of interface, e.g. head rest, knee support or lumbar support
    • A61H2201/165—Wearable interfaces
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61H—PHYSICAL THERAPY APPARATUS, e.g. DEVICES FOR LOCATING OR STIMULATING REFLEX POINTS IN THE BODY; ARTIFICIAL RESPIRATION; MASSAGE; BATHING DEVICES FOR SPECIAL THERAPEUTIC OR HYGIENIC PURPOSES OR SPECIFIC PARTS OF THE BODY
    • A61H2201/00—Characteristics of apparatus not provided for in the preceding codes
    • A61H2201/50—Control means thereof
    • A61H2201/5058—Sensors or detectors
    • A61H2201/5061—Force sensors
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61H—PHYSICAL THERAPY APPARATUS, e.g. DEVICES FOR LOCATING OR STIMULATING REFLEX POINTS IN THE BODY; ARTIFICIAL RESPIRATION; MASSAGE; BATHING DEVICES FOR SPECIAL THERAPEUTIC OR HYGIENIC PURPOSES OR SPECIFIC PARTS OF THE BODY
    • A61H2201/00—Characteristics of apparatus not provided for in the preceding codes
    • A61H2201/50—Control means thereof
    • A61H2201/5058—Sensors or detectors
    • A61H2201/5069—Angle sensors
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61H—PHYSICAL THERAPY APPARATUS, e.g. DEVICES FOR LOCATING OR STIMULATING REFLEX POINTS IN THE BODY; ARTIFICIAL RESPIRATION; MASSAGE; BATHING DEVICES FOR SPECIAL THERAPEUTIC OR HYGIENIC PURPOSES OR SPECIFIC PARTS OF THE BODY
    • A61H2201/00—Characteristics of apparatus not provided for in the preceding codes
    • A61H2201/50—Control means thereof
    • A61H2201/5058—Sensors or detectors
    • A61H2201/5071—Pressure sensors
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61H—PHYSICAL THERAPY APPARATUS, e.g. DEVICES FOR LOCATING OR STIMULATING REFLEX POINTS IN THE BODY; ARTIFICIAL RESPIRATION; MASSAGE; BATHING DEVICES FOR SPECIAL THERAPEUTIC OR HYGIENIC PURPOSES OR SPECIFIC PARTS OF THE BODY
    • A61H2201/00—Characteristics of apparatus not provided for in the preceding codes
    • A61H2201/50—Control means thereof
    • A61H2201/5058—Sensors or detectors
    • A61H2201/5084—Acceleration sensors
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61H—PHYSICAL THERAPY APPARATUS, e.g. DEVICES FOR LOCATING OR STIMULATING REFLEX POINTS IN THE BODY; ARTIFICIAL RESPIRATION; MASSAGE; BATHING DEVICES FOR SPECIAL THERAPEUTIC OR HYGIENIC PURPOSES OR SPECIFIC PARTS OF THE BODY
    • A61H2230/00—Measuring physical parameters of the user
    • A61H2230/60—Muscle strain, i.e. measured on the user, e.g. Electromyography [EMG]
    • A61H2230/605—Muscle strain, i.e. measured on the user, e.g. Electromyography [EMG] used as a control parameter for the apparatus

Definitions

  • the present invention relates to a control system for an exoskeleton configured to be coupled to a body of a subject.
  • it relates to a system for improved control over postural tasks of a lower limb haptic exoskeleton for rehabilitation or walking purposes, with improved postural equilibrium.
  • Haptic exoskeletons devices have been applied for rehabilitation purposes, and in particular for use in lower limb rehabilitation, or for walking.
  • the control system of the exoskeleton is designed to react to biometric stimuli coming from a subject (e.g. a patient) making use of the haptic exoskeleton, thus allowing the subject to contribute to the control of the exoskeleton, and receive as a feedback from the exoskeleton a compliance feeling.
  • Prior art mainly identifies four classes of such existing apparatuses, besides exoskeletons offering pure strength augmentation that are not here considered.
  • a first apparatus class relates to passive systems, where a passive control operates without a direct interaction with the patient (i.e. the patient has no control over the exoskeleton) .
  • Such systems are adapted to sustain the patient and to force him to perform pre-programmed rehabilitation exercises.
  • a second apparatus class relates to reciprocating exoskeletons, which offer a moderate postural support to the patient, by providing coordination to the motion, but not equilibrium. Therefore, the patient must walk with the help of physical supporting means, such as crutches.
  • a third apparatus class relates to exoskeletons that maintain postural equilibrium but move a fully paralyzed patient like an autonomous robot, for example by means of a joystick. The patient is thus again deprived of the faculty of controlling the exoskeleton.
  • a fourth apparatus class relates to active systems, whose hybrid control operates by making a passive control interact with an active control performed by the patient, such as by means of measurements of electromyography (EMG) signals, eye movements, etc.
  • EMG electromyography
  • Such active systems offer coordination and strength augmentation to the efforts of the patient, but do not guarantee posture or equilibrium.
  • the patient wearing the exoskeleton in spite of the compliance offered by the device, has a feeling of the reality that is different with respect to what he would experience in freestanding condition, due to being partially sustained and lacking of gravity.
  • the lack of gravity is detrimental for rehabilitation purposes since, although having successfully terminated the rehabilitation process with the exoskeleton, the patient might still be unable to perform as desired because of the sudden reintroduction of gravity effects.
  • Another challenge in designing the control system of a haptic exoskeleton relies in the real time processing of biometric data, e.g. measured EMG signals, and in their translation into controls of the electrical motors actuating the exoskeleton.
  • Muscle synergy theory is a widely studied subject in motor control research. Although the understanding of signals that the central nervous system produces and sends to the muscles to effect movement is still incomplete, the muscle synergy theory suggests that the central nervous system produces a small number of signals that pass through a network that distributes combinations of these signals to the muscles. Muscle synergies are defined as a coherent activation, in space and time, of a group of muscles. Therefore, according to this theory, the nervous system activates the muscles by combining a number of bundles of activations (known as “muscle synergies" or simply “synergies").
  • the signals sent to such muscles define a set of vectors which are partially linearly dependent, and the synergies are defined as a set of signals linearly independent forming a vector basis for explaining all signals received by the muscles. It is argued that the biological objective is to reduce the number of variables that the nervous system has to control, thus also becoming an attractive solution for controlling an exoskeleton.
  • exoskeleton 1 The exoskeleton disclosed in the above-cited literature is reported in Figure 1 , shown as exoskeleton 1, worn by a patient 2, e.g. for rehabilitation purposes.
  • the exoskeleton 1 integrates an interaction with the patient by measuring and processing biometric data, such as electromyography (EMG) signals, thus making the exoskeleton compliant to the intents of the patient through admittance control (velocity/position) of joints 7, 8, 9 (in the "joint space") of the exoskeleton 1.
  • EMG electromyography
  • the exoskeleton 1 maintains equilibrium or executes postural tasks (such as sit-to-stand, step, walk, etc.), with a postural position control in the three-dimensional Euclidean (Cartesian) space (also named "postural space”).
  • EMG electromyography
  • the exoskeleton 1 is adapted to be worn by a patient 2 undergoing a rehabilitation treatment and comprises a frame structure 3 (shown schematically) to be coupled to the trunk and lower limbs of the patient 2.
  • the frame structure 3 includes: a back support element 6a coupled to the trunk of the patient 2, linked to an upper arm 4, coupled to the thigh of the patient 2 and extending substantially parallel to the femur; a lower arm 5 coupled to the leg of the patient 2 and extending substantially parallel to the tibia.
  • the lower arm 5 is coupled to a foot rest 6b, supporting the foot and sole of the patient 2, via an ankle joint 7 and to the upper arm 4 via a knee joint 8; the upper arm 4 ends at a hip joint 9, connected to the back support of the trunk.
  • the ankle, knee and hip joints 7, 8, 9 allow the relative movement between the foot rest 6b, lower arm 5, upper arm 4 and back support; in particular the ankle and hip joints 7, 9 have two degrees of freedom, for rotations in the sagittal plane (the XZ plane defined by the X axis and the Z axis of Figure 1 ) and in a frontal plane (an YZ plane defined by the Y axis and the Z axis of Figure 1 ), while the knee joint 8 has a single degree of freedom, for rotation in the sagittal plane.
  • Actuators e.g. electric motors 10 are coupled to each one of the ankle, knee and hip joints 7, 8, 9, and are operable to control their movements in the sagittal and frontal planes.
  • a control unit 12 is coupled to the exoskeleton 1 and implements a control algorithm for driving the electric motors 10 at the various joints 7, 8, 9, thus controlling the exoskeleton movement.
  • the control unit 12 receives at input: (i) biometric signals S EMG (more specifically surface EMG signals, i.e. detected on the skin surface), indicative of the patient intention and strength of movement, from bioelectric sensors 14 coupled to the patient lower limbs and trunk and placed on the muscles involved in lower limb movement (in Figure 1 , for reasons of simplicity of depiction only a sensor is schematically shown, placed on the thigh); (ii) position and velocity signals from position/velocity sensors 16, for detection of angles and angular velocities (i.e.
  • postural signals comprising (iii) pressures exerted by the feet of the patient, acquired through reaction force sensors 17 (e.g., pressure sensors), positioned on the exoskeleton foot rests; (iv) movement signals, acquired through inertial sensors 18 (e.g., accelerometers and gyroscopes), positioned on the trunk of the patient 2, to measure the movements (i.e.
  • Control unit 12 processes the input signals and, based on the implemented control algorithm, outputs suitable driving commands for the electric motors 10, in particular indicative of the velocity that is to be impressed to the ankle, knee and hip joints 7, 8, 9.
  • the exoskeleton 1 can have two configurations:
  • a motion of the joints results in a corresponding motion of all limbs of the exoskeleton.
  • a position or velocity of the joint angles results in a position or velocity (motion) of the coordinates of all points and angles of the limbs of the exoskeleton, measured in the three dimensional Cartesian space.
  • This is called a "posture” of the body, and the motion a "postural task” (or more simply a “task”).
  • a number equal to the DOF of appropriate coordinates of points and/or angles of the limbs is needed. This choice is not unique.
  • each component of the postural position (or postural task vectors) is called “elemental postural component” (or “elemental task component”).
  • elemental task component is relevant when a postural task is performed by moving only one elemental component, while leaving the other components unchanged.
  • the postural components include the two coordinates of the projection on the XY plane (with reference to the three-dimensional Cartesian axis reference shown in Fig. 1 ) of a Center of Mass of the exoskeleton(1)-patient(2) system.
  • This projection point is called Center of Gravity, i.e. COG.
  • the COG coordinates, as components of the posture, are functions of the joint angles.
  • COP Centre of Pressure of reaction ground forces
  • ZMP Zero Moment Point
  • the joint angles control directly the COG, having as objective the ZMP position, either in the still standing, in double stance, or during the gait in single stance, in anticipating the position the swing foot will reach at the touch down.
  • FIG. 2 schematically shows a control loop 50 for the exoskeleton 1, according to the above-cited literature.
  • the control loop 50 includes a velocity feedback loop 11a operating in the joint space and implementing a velocity control over the joints of the exoskeleton, and a position feedback loop 11b operating in the postural space and implementing a control over the postural tasks.
  • the velocity feedback loop 11a includes an admittance-control block 70, implementing a mechanical admittance control over the joints of the exoskeleton 1.
  • Mechanical admittance is the dynamical relationship existing between the force impressed to a mechanical body, as an input, and its velocity, as an output. Mechanical impedance is, vice versa, the inverse relationship between velocity imposed to a body and the force needed to achieve such velocity.
  • the admittance-control block 70 is adapted to implement the transfer function of the reaction that will offer the joints of the exoskeleton to the efforts of the patients.
  • the EMG signal S EMG is a vector, whose components are taken from different muscles, selected to control in pairs each joint.
  • the intensity of the components i.e. their magnitude
  • the intensity of the components is taken as an indication of the effort expressed by the patient to move a corresponding joint.
  • the admittance-control block 70 transfers, independently for each joint, the two components of the EMG signal S EMG taken from antagonistic muscles of the patient into velocity references of the respective joint control loops, indicated, for all joints, as the vector ⁇ p .
  • This direct relationship between components of the EMG signal S EMG related to a joint to its velocity is defined here as "single input-single output admittance" (or SISO admittance).
  • the postural loop signals While the patient tries to move the joints of the exoskeleton, the postural loop signals are superimposed to the ⁇ p derived, through admittance control, from the EMG signal S EMG .
  • the postural loop signals allow maintaining postural equilibrium and/or forcing to execute postural tasks, through a preview block 52 (disclosed in EP 2 497 610 B1 ), a K block 56 (a compensator to guarantee stability of the equilibrium loop, disclosed in EP 2 238 894 B1 ), and an inverse Jacobian block 60.
  • the postural loop operates as follows.
  • the preview block 52 provides reference signals for all posture vector x defined in the postural space (among them the COG and the auxiliary signal ZMP, function of the COG motion), as described in the mentioned patent of the same author ( EP 2 497 610 B1 ).
  • the preview reference signals include preview positions x prev (including preview COG positions COG prev , and preview ZMP positions ZMP prev ) and preview velocities ⁇ prev (including preview COG velocities C ⁇ G prev ) .
  • a vector of position errors e x is obtained as output of a sum block 54, and is calculated as the difference between the preview positions x prev and a vector of estimated positions x est .
  • the estimated positions x est include estimated COG positions COG est and estimated ZMP positions ZMP est , and are generated by a state estimator block 66, as better described in the following.
  • the K block 56 receives as input the position error e x , and is used to calculate a processed error as described in Youngjin Choi, Doik Kim, Yonghwan Oh, Bum-Jae You "Posture/Walking Control for Humanoid Robot Based on Kinematic Resolution of CoM Jacobian With Embedded Motion" IEEE TRANSACTIONS ON ROBOTICS, VOL. 23, NO. 6, DECEMBER 2007 1285 .
  • the vector of summed velocities ⁇ sum represents the desired postural tasks velocities corrected by the tracking error of the feedback.
  • the tracking error goes to zero when perfect tracking is achieved.
  • the vector of summed velocities ⁇ sum is processed by means of the inverse of the Jacobian to transfer signals form the postural to the joint space.
  • a vector of measured joint angular velocities ⁇ meas is then subtracted (sum block 88) in a negative feedback reaction to the vector of reference joint angular velocities ⁇ ref , sum of the contributions of both the admittance control and the postural control, to obtain the vector of angular velocity errors ⁇ ⁇ .
  • the vector of angular velocity errors ⁇ ⁇ is set as input to a compensation block 63, which is used to close the feedback loop of the velocity feedback loop 11a according to per se known techniques (e.g., as described in patent EP 2 497 610 B1 by the same applicant).
  • the compensation block 63 generates as output a system signals vector u s , which is adapted to control the actuators (e.g. electric motors 10) coupled to the exoskeleton.
  • a system block 64 represents the dynamics of the joint exoskeleton 1 and patient 2.
  • the system block 64 receives as input system signals vector u s adapted to control the actuators of the exoskeleton, and outputs a vector of measured joint angular velocities ⁇ meas , a vector of measured joint angles ⁇ meas and a vector of output measures y meas . Said outputs are provided by the position/velocity sensors 16.
  • the vector of output measures y meas includes measured COG positions COG meas , provided by the position/velocity sensors 16, measured ZMP positions ZMP meas , provided directly through the COP by the reaction force sensors 17, measures of position and angular velocity of the centre of mass of the trunk through inertial sensor, and in single stance position of the free foot from the ground from proximity sensors.
  • the vector of outputs measures y meas is given as an input to the state estimator block 66.
  • the state estimator block 66 implements a Kalman filter and outputs the vector of estimated positions x est (including a vector of estimated COG positions COG est and a vector of estimated ZMP positions ZMP est ) , in a per se known way already described in the previous patent EP 2 238 894 B1 .
  • the vector of estimated positions x est is set as an input to the sum block 54 and is subtracted to the preview positions x prev , within the negative feedback loop already described with reference to the sum block 54. Moreover, the vector of estimated positions x est is set as an input to an inhibitor block 68.
  • the inhibitor block 68 is adapted to bind the freedom of the patient 2, by setting the gain of the admittance control module to zero (inhibiting the action of the patient), and generating in the preview some automatic recovering action, when equilibrium limits are reached.
  • the inhibitor block 68 receives as an input the vector of estimated ZMP positions ZMP est included in the vector of estimated positions x est , and compare it with a convex hull of the feet of the patient 2.
  • the convex hull is defined as a polygonal area that includes the soles of the feet of the patient 2.
  • the convex hull is the smallest geometrical convex set including a set of points of the XY plane of Figure 1 , such set of points being defined by the soles of the patient's feet on the ground.
  • the equilibrium is guaranteed.
  • the aim of the present invention is to provide a haptic exoskeleton configured to be coupled to a body of a patient, that meets the above requirements.
  • a haptic exoskeleton is provided, as defined in the annexed claims.
  • the present invention is disclosed with initial reference to Figure 3 , being aimed at integrating the compliance with the patient 2 to the automatic control of posture, according to a first aspect, and, according to a further aspect, at using muscle synergies and artificial intelligence to use all biometrical data as a whole to control all joints together instead of pairs of data for each single joint in isolation.
  • an exoskeleton operating in a fixed position with motion on a sagittal plane (i.e. an XZ plane-sagittal plane-defined by the X axis and Z axis shown in Figure 1 ) with coordinate control of the pairs, right and left, of ankles, knees and hips, thus being able to perform tasks such as a sit-to-stand exercise (e.g. sit and raise from a chair) .
  • a sagittal plane i.e. an XZ plane-sagittal plane-defined by the X axis and Z axis shown in Figure 1
  • this exoskeleton is provided with six joints, only three of them are independent between each other, hence the kinematics has only three DOF: the jointly pairs of ankles, of knees and of hips.
  • three exemplary elemental postural tasks in the Cartesian posture space
  • the vector x the COG moving, along the X axis from heel to toe of the feet soles; raising or lowering the trunk, along the Z axis; bending forward or backward the trunk in the sagittal plane.
  • consistent posture can be guaranteed partially (some elemental postural components or tasks) automatically by the postural loop and partially (the complementary postural components or tasks) by the admittance control of the joints from the patient, or jointly by both.
  • the components of the posture controlled either by the automatic postural loop or by the patient
  • the components of the joints operated by the patient
  • the two actions can operate independently over complementary elemental postural tasks, or can partially/totally overlap.
  • the exoskeleton when overlapping, must offer complete postural support at least in the earlier rehabilitation phases (being stiff, firmly holding the patient), and be more and more compliant in the terminal phases, when the patient regains postural functionalities.
  • Tutoring means that all elemental postural tasks are under the control of the external automatic postural loop. This is equivalent to WBC.
  • Coordination means that independently some elemental postural tasks are under automatic control and the complementary tasks are controlled by the patient, that has to coordinate his motion with the previous ones.
  • Cooperation means that some or all postural tasks are jointly controlled by either loops, and the intensity of contribution of the patient can be modulated, i.e. the two control actions have to cooperate.
  • R u ,N u ,R p ,N p are square selection matrices of dimension n ⁇ n with only ones or zeros on the diagonal, spanning ranges and null-spaces, respectively, of the postures and of the joint spaces
  • J is the Jacobian matrix
  • the modulation coefficient ⁇ is a coefficient such that the modulation coefficient ⁇ is equal to or comprised between zero and one (i.e., 0 ⁇ ⁇ ⁇ 1) .
  • the "posture range matrix” R u defines the subspace spanned by the automatic task loop;
  • the "posture null-space matrix” N u defines a subspace not spanned by the automatic task loop (thus complementary to that spanned by the automatic task loop), but covered, vice versa, by the action of the patient.
  • the "joint range matrix” R p defines the subspace of the joint components controlled by the patient, as a consequence, for a consistent posture N p defines the subspace of the joint components controlled automatically by the postural loop.
  • the objective of the scheme of Fig. 3 is to limit the action of the automatic postural task loop to the subspace of postural tasks spanned by R u , and the direct action of the patient, through the admittance control operating on the joints belonging to the subspace spanned by R p , to not interfere with the tasks in the range of R u , but to generate task motion only in its null-space N u .
  • the value of ⁇ affects both the gain of the admittance control and the admittance felt by the patient.
  • the control of the exoskeleton is in cooperation mode, where tasks in the range of N u are jointly controlled by the external postural loop 48 and by the patient, that must cooperate.
  • the value of the coefficient ⁇ modulates the intensity of the control contribution of the patient, and consequently, the stiffness that the patient feels as response from the exoskeleton.
  • each elemental task and elemental joint can be trained firstly through coordination, and then through cooperation, by changing the range subspaces of postural and joint spaces, with a proper choice of the R u and R p matrices.
  • R u and R p the dimension of the subspaces, always modifying R u and R p , to increase the number of joints and tasks under the patient control (e.g., from one at the beginning of the rehabilitation until the full joint space is covered).
  • the two equilibrium tasks i.e.
  • the control system 75 is shown with reference to Fig. 3 .
  • the signals ⁇ sum are multiplied by the logic function implemented by the block M1 78 to obtain ⁇ M 1 , so as to limit the action of the automatic postural task loop to the subspace spanned by R u , in coordination mode, or to modulate the control action in cooperation mode.
  • EMG signals ⁇ P provided at the output of admittance control module 91 originated from the EMG signal S EMG are multiplied by the logic function implemented by the block M2 82, so as to limit the direct action of the patient to the subspace spanned by R p , and balance its contribution.
  • the output of the block M2 82 is indicated as ⁇ M 2 , which both enters the sum node 86 and is feedforwarded before the inverse Jacobian block, after multiplying ⁇ M 2 by the logic function implemented by the block F 84 to generate as output of the block F 84 the vector ⁇ F .
  • ⁇ F is then subtracted to ⁇ M 1 (result of the postural loop) in the sum node 80, and the output of the sum node 80 is called u comb .
  • u comb is then processed by the inverse Jacobian 60.
  • the output of the inverse Jacobian is called u j , and is summed to ⁇ M 2 in the sum node 86 to generate the reference ⁇ ref of the joint velocity control loops.
  • control system 75 The following logic blocks of the control system 75 are unchanged with respect to what shown in Fig. 2 : the joint velocity loops, the compensator, and the system outputs that, through the Kalman filter, generate the postural position estimates to close the postural loop.
  • ⁇ ⁇ ref J ⁇ 1 ⁇ R u + N u 1 ⁇ ⁇ ⁇ x ⁇ sum + N u ⁇ J ⁇ ⁇ ⁇ R p ⁇ ⁇ ⁇ p
  • the inhibitor block 68 is adapted to bind the freedom of the patient 2 if equilibrium limits are reached, and to perform automatic recovery tasks according to per se known techniques when this is not already done by the control of the postural tasks (i.e. when the patient has full control of the exoskeleton).
  • the inhibitor block 68 receives as an input the vector of estimated ZMP positions ZMP est included in the vector of estimates from the output measures y meas , and compares it with the convex hull of the feet of the patient 2.
  • the inhibitor forces ⁇ to zero in the blocks M1, M2, F, thus setting the exoskeleton in tutoring mode. Then, the inhibitor forces the preview to perform some recovery tasks.
  • the recovery tasks can be of two types, according to some decision criteria: either acting on the joints to move the ZMP in the centre of the feet sole support, or, if such a task is not enough to recover equilibrium, acting on the joints of the exoskeleton to move one foot into a new position to intercept the ZMP.
  • the internal joint loop 49 which comprises the compensation block 63, the system block 64 and in particular the M2 block 82, implements a velocity control (i.e. a control on angular velocities of joints), and is adapted to guarantee compliance to the intentions of the patient expressed through the acquired bioelectric (e.g., EMG) signals S EMG .
  • the external task loop 48 which comprises the preview block 52, the K block 56the state estimator block 66 and in particular the M1 block 78, implements postural position control in the postural space, and is adapted to guarantee postural equilibrium.
  • admittance control module 91 configured to generate the signals ⁇ P from EMG signals S EMG , with reference to Figure 4 .
  • An aspect of the present solution instead of exploiting single input-single output admittance (SISO admittance) for each joint in isolation, introduces a generalized multi input-multi output admittance (MIMO admittance). Therefore, multiple EMG signals are processed together to control multiple joints, irrespective of the original separation of pairs of antagonist muscles for each joint.
  • SISO admittance single input-single output admittance
  • MIMO admittance generalized multi input-multi output admittance
  • the number of biometric data measured is greater than what strictly needed for controlling each joint. Therefore, an aspect of the present solution exploits muscle synergies and artificial intelligence to generate from such multivariable signal patterns the control of the joints, to achieve the desired postural task.
  • Both the synergies extractor and the neural network are trained for each specific patient in a series of preliminary exercises, before the rehabilitation starts. Then the outputs of the training are used in real time for admittance control during the rehabilitation. Since the physiological behaviour of the patient changes during the course of the rehabilitation process, the training is repeated in time accordingly.
  • the training is performed by collecting and recording time samples of EMG signals S EMG , along with joint positions and velocities, from a series of elemental postural tasks. Since the patient needs support at the beginning of the rehabilitation, these exercises exploit the same functionalities of decomposition of the postural tasks offered by the exoskeleton, as described before.
  • a old fashion "SISO" admittance control is temporarily implemented on one joint, separately from the other joints of the exoskeleton.
  • the exercise is performed on one elemental postural task, while the others are maintained at constant value by the automatic postural loop.
  • the patient is asked to repeat for that elemental postural task and on that joint the exercise for a number of times.
  • the tutoring or cooperation mode is used to help the patient.
  • time samples of all available EMG signals are collected.
  • Matrix A has dimension N ⁇ M, where N is the number of time samples and M is the number of measures, i.e. the number of EMG sensors.
  • N is the number of time samples
  • M is the number of measures, i.e. the number of EMG sensors.
  • P and V have both dimension N ⁇ h, where h is the number of independent joints of the exoskeleton.
  • the matrix A of the samples of EMG signals is used to extract the synergies according to known techniques, such as matrix factorization, Principal Component Analysis (PCA), Singular Value Decomposition (SVD), Independent Component Analysis (ICA), Non-Negative Matrix Factorization (NNMF), etc.
  • PCA Principal Component Analysis
  • Singular Value Decomposition SVD
  • ICA Independent Component Analysis
  • NMF Non-Negative Matrix Factorization
  • the EMG matrix A is factorized in the form of A ⁇ W ⁇ H, where W is the matrix of synergy samples with dimensions N ⁇ K with the number of synergies K ⁇ M and H a full rank coefficient matrix of dimensions K ⁇ M.
  • the synergy matrix W stores the information of the whole (M) measured EMG signals S EMG .
  • the multiplication matrix H calculated during the first training procedure, is used in real time to generate the synergies from the measured EMG signals S EMG .
  • H H' ⁇ ( H ⁇ H' ) -1 .
  • This matrix constitutes the factorization block 92 of the admittance control module 91 in Figure 4 .
  • the present solution exploits artificial neural network techniques (neural network block 94 of the admittance control module 91 in Figure 4 ) to process the extracted muscle synergies and generate signals ⁇ ⁇ .
  • a Nonlinear autoregressive Exogenous Model (NARX), receiving the synergies and the joint positions as inputs, is used here to generate the desired vector of processed EMG signals ⁇ p to control the exoskeleton 1, as shown in the block diagram of Figure 3 .
  • NARX Nonlinear autoregressive Exogenous Model
  • the external "exogenous" inputs of the NARX (considered as time series) are here represented by the synergies and the joint positions.
  • each synergy component enters the NARX as one input, while the joint positions (which play only an ancillary role) enter as input in a product form with the synergies following a Recursive Neural Network scheme.
  • each position component will generate two groups of inputs, since each position component at a given time sample i-2 is multiplied with all synergy components taken at a first time sample i-1 and at a second time sample i.
  • the neural network adopted is called NARX-RNN, since it receives inputs according to the RNN scheme.
  • NARX and RNN are well known topics in the Artificial intelligence literature, described, in particular, in a recent paper for estimating torques from EMG signals ( Z. Li, M. Hayashibe, Q. Zhang, and D. Guiraud, "FES-Induced Muscular Torque Prediction with Evoked EMG Synthesized by NARX-Type Recurrent Neural Network", 2012 IEEE/RSJ International Conference on Intelligent Robots and Systems October 7-12, 2012. Vilusine, Algarve, Portugal ).
  • the NARX-RNN neural network block 94 in Figure 4 is trained during a second step of the training procedure from the same W, P and V matrices obtained during the postural exercises performed by the patient 2 as already described, by means of per se known deep-learning and neural networks techniques.
  • the NARX-RNN receives as inputs the synergy matrix W, the position matrix P in the product form with W, and as output the velocity matrix V. Training will be performed with a reinforcement learning procedure until the error between the matrix output of the network and the velocity matrix V is lower than a predefined error threshold.
  • the input scheme is applied to the trained neural network and the corresponding vector of processed EMG signals ⁇ pRT , output of the NARX-RNN neural network block 94, is thus provided to the control system 75 of Fig. 3 (as the real time processed EMG signals ⁇ p ).
  • the exoskeleton 1 can have applications both for rehabilitation and/or for walking.
  • the training is advantageously repeated several times during the rehabilitation phases, to track the progresses of the patient 2.
  • a co-training of both the artificial neural network applied to the exoskeleton 1 and of the physical, natural, neural network of the patient 2 will therefore be achieved.
  • the characteristics of the resulting artificial neural network can be used to monitor the progresses.
  • the training may be performed only once (the first time that the exoskeleton 1 is worn), in order to assure the best wearability of the exoskeleton 1 by the subject.
  • the co-training encompasses, at the beginning, a limited number of tasks with a limited number of joints controlled by the patient 2; then, as the rehabilitation process proceeds, the number of involved tasks and joints can be increased, based on the progresses of the patient 2.
  • Complete control of the equilibrium of the exoskeleton 1 by the patient 2 is left to the last steps of the rehabilitation process. If the progress of the patient so requires, the equilibrium control (i.e. the control of the elemental postural tasks represented by the two coordinate of the COG), through a proper choice of the matrices R u , R p may always be left to the automatic task loop 48.
  • the exoskeleton 1 could be either used in a fixed position, mounted on a platform to perform rehabilitation exercises, free from the platform (to perform walking on a treadmill, with electronics on a fixed position), or for free walking with an on-board electronics.
  • different joints 7, 8, 9 of the patient 2 can be trained selectively according to the disability of the patient 2, with a progression during the rehabilitation process.
  • cases of hemiparesis can also be treated, by training only some joints and by exploiting the only able joints for the control implemented in the control system 75.
  • a virtual reality condition can be provided by the exoskeleton 1 to the patient 2.
  • the patient 2 wearing the exoskeleton 1 in coordination condition has a different feeling of the reality with respect to what he would experience in freestanding condition, due to being partially sustained and the lacking of gravity.
  • a virtual reality management is offered by the exoskeleton 1 to the patient.
  • Virtual reality management means that the patient 2 can perceive gravity effects by means of mechanical stimuli, artificially introducing torques on the joints.
  • FIG 5 shows the block diagram of the control scheme according to this further embodiment.
  • the logic blocks shown in Figure 5 and already discussed with reference to Figure 3 are not further described; same reference numerals are used.
  • the gravity perception is artificially implemented for the patient by introducing torques (T block 90 in the control system 75) on each joint through an admittance scheme in the velocity loop of the actuators controlling the joints 7, 8, 9 of the exoskeleton 1.
  • the T block 90 calculates these torques and, by means of per se known filtering and amplification techniques, outputs a vector of torque velocities ⁇ T , which is set as an additional input to the sum block 88. Therefore, this vector of torque velocities ⁇ T acts similarly to an additional signal to the vector of reference joint angular velocities ⁇ ref , and causes the artificial perception of gravity for the patient 2 through a virtual admittance.
  • the control system 75 allows to realize an exoskeleton 1 able to perform postural tasks and to offer compliance to an impaired patient 2, by means of integration of two competing control loops. Moreover, it allows to realize a processing interface using biometric human signals (EMG signals S EMG ) and to guarantee compliance and wearability to the exoskeleton 1.
  • EMG signals S EMG biometric human signals
  • the exoskeleton 1 is designed to implement admittance control and to process in real time the measured EMG signals S EMG though the use of a neural network.
  • Admittance control is performed by the patient 2 on the exoskeleton 1 by means of the internal joint loop 49, and the progression of the patient's control over the exoskeleton 1 is achieved by:
  • exoskeleton tutoring mode no contribution from the patient 2
  • coordination mode full contribution of the patient 2
  • cooperation mode partial contribution of the patient 2).
  • neural networks to process the EMG signals S EMG according to muscle synergy theory allows to learn the patterns of the most significant postural tasks, and allows the patient 2 to perform multiple and complex tasks simultaneously, resulting in complex and coordinated motion.
  • the exoskeleton 1 can have applications both for rehabilitation and for walking, thanks to the co-training procedure that can either be repeated several times during the rehabilitation (to follow the progresses of the patient 2) or only once (the first time that the exoskeleton 1 is worn, to set the matrix factorization and NARX model in order to assure the best wearability of the exoskeleton 1).
  • a virtual reality condition can be provided by the exoskeleton 1 to the patient 2, to avoid the feel of lack of gravity.
  • the present solution allows to flexibly manage the interaction between the two loops.
  • control system could in principle be used (with suitable modifications) for controlling the movement also of the upper limbs or other parts of the body.
  • the intention of movement of the patient may also be determined from other biological measures; for example, the intention of movement may be determined directly or indirectly based on measures of joint angles and forces exchanged between the patient and the exoskeleton frame structure.
EP19425014.8A 2019-03-08 2019-03-08 Steuerungssystem für ein haptisches exoskelett der unteren extremitäten zur rehabilitation oder zum gehen mit verbesserter gleichgewichtskontrolle, mensch-maschine-schnittstelle Withdrawn EP3705105A1 (de)

Priority Applications (1)

Application Number Priority Date Filing Date Title
EP19425014.8A EP3705105A1 (de) 2019-03-08 2019-03-08 Steuerungssystem für ein haptisches exoskelett der unteren extremitäten zur rehabilitation oder zum gehen mit verbesserter gleichgewichtskontrolle, mensch-maschine-schnittstelle

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
EP19425014.8A EP3705105A1 (de) 2019-03-08 2019-03-08 Steuerungssystem für ein haptisches exoskelett der unteren extremitäten zur rehabilitation oder zum gehen mit verbesserter gleichgewichtskontrolle, mensch-maschine-schnittstelle

Publications (1)

Publication Number Publication Date
EP3705105A1 true EP3705105A1 (de) 2020-09-09

Family

ID=66049146

Family Applications (1)

Application Number Title Priority Date Filing Date
EP19425014.8A Withdrawn EP3705105A1 (de) 2019-03-08 2019-03-08 Steuerungssystem für ein haptisches exoskelett der unteren extremitäten zur rehabilitation oder zum gehen mit verbesserter gleichgewichtskontrolle, mensch-maschine-schnittstelle

Country Status (1)

Country Link
EP (1) EP3705105A1 (de)

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113995629A (zh) * 2021-11-03 2022-02-01 中国科学技术大学先进技术研究院 基于镜像力场的上肢双臂康复机器人导纳控制方法及系统
CN114028156A (zh) * 2021-10-28 2022-02-11 深圳华鹊景医疗科技有限公司 康复训练方法、装置及康复机器人
CN114392137A (zh) * 2022-01-13 2022-04-26 上海理工大学 一种穿戴式柔性下肢助力外骨骼控制系统

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
EP2238894A1 (de) 2009-04-07 2010-10-13 Syco Di Hedvig Haberl & C. S.A.S. System zur Steuerung einer haptischen Exoskelettvorrichtung zu Rehabilitationszwecken und entsprechende haptische Exoskelettvorrichtung
EP2497610A1 (de) 2011-03-09 2012-09-12 Syco Di Hedvig Haberl & C. S.A.S. System zur Steuerung einer Robotervorrichtung während des Gehens, insbesondere zu Rehabilitationszwecken, sowie zugehörige Robotervorrichtung
CN105213153A (zh) * 2015-09-14 2016-01-06 西安交通大学 基于脑肌信息变阻抗的下肢康复机器人控制方法
CN106214427A (zh) * 2016-08-11 2016-12-14 上海交通大学 受试者主导的下肢康复外骨骼控制系统及方法
CN107397649A (zh) * 2017-08-10 2017-11-28 燕山大学 一种基于径向基神经网络的上肢外骨骼康复机器人控制方法

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
EP2238894A1 (de) 2009-04-07 2010-10-13 Syco Di Hedvig Haberl & C. S.A.S. System zur Steuerung einer haptischen Exoskelettvorrichtung zu Rehabilitationszwecken und entsprechende haptische Exoskelettvorrichtung
EP2497610A1 (de) 2011-03-09 2012-09-12 Syco Di Hedvig Haberl & C. S.A.S. System zur Steuerung einer Robotervorrichtung während des Gehens, insbesondere zu Rehabilitationszwecken, sowie zugehörige Robotervorrichtung
CN105213153A (zh) * 2015-09-14 2016-01-06 西安交通大学 基于脑肌信息变阻抗的下肢康复机器人控制方法
CN106214427A (zh) * 2016-08-11 2016-12-14 上海交通大学 受试者主导的下肢康复外骨骼控制系统及方法
CN107397649A (zh) * 2017-08-10 2017-11-28 燕山大学 一种基于径向基神经网络的上肢外骨骼康复机器人控制方法

Non-Patent Citations (6)

* Cited by examiner, † Cited by third party
Title
G. MENGA ET AL.: "Lower limb exoskeleton for rehabilitation with improved postural equilibrium", ROBOTICS, 2018
L. SENTIS; O. KHATIB: "Synthesis of whole-body behaviors through hierarchical control of behavioral primitives", INT. J. HUMAN. ROB., vol. 2, no. 4, December 2005 (2005-12-01), pages 505 - 518
YOUNGJIN CHOI ET AL: "Posture/Walking Control for Humanoid Robot Based on Kinematic Resolution of CoM Jacobian With Embedded Motion", IEEE TRANSACTIONS ON ROBOTICS, IEEE SERVICE CENTER, PISCATAWAY, NJ, US, vol. 23, no. 6, 1 December 2007 (2007-12-01), pages 1285 - 1293, XP011332702, ISSN: 1552-3098, DOI: 10.1109/TRO.2007.904907 *
YOUNGJIN CHOI; DOIK KIM; YONGHWAN OH; BUM-JAE YOU: "Posture/Walking Control for Humanoid Robot Based on Kinematic Resolution of CoM Jacobian With Embedded Motion", IEEE TRANSACTIONS ON ROBOTICS, vol. 23, no. 6, December 2007 (2007-12-01), pages 1285, XP011332702, DOI: doi:10.1109/TRO.2007.904907
Z. LI; M. HAYASHIBE; Q. ZHANG; D. GUIRAUD: "FES-Induced Muscular Torque Prediction with Evoked EMG Synthesized by NARX-Type Recurrent Neural Network", 2012 IEEE/RSJ INTERNATIONAL CONFERENCE ON INTELLIGENT ROBOTS AND SYSTEMS, 7 October 2012 (2012-10-07)
Z. LI; M. HAYASHIBE; Q. ZHANG; D. GUIRAUD: "FES-Induced Muscular Torque Prediction with Evoked EMG Synthesized by NARX-Type Recurrent Neural Network", 2012 IEEE/RSJ INTERNATIONAL CONFERENCE ON INTELLIGENT ROBOTS AND SYSTEMS, 7 October 2012 (2012-10-07), XP032287448 *

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114028156A (zh) * 2021-10-28 2022-02-11 深圳华鹊景医疗科技有限公司 康复训练方法、装置及康复机器人
CN113995629A (zh) * 2021-11-03 2022-02-01 中国科学技术大学先进技术研究院 基于镜像力场的上肢双臂康复机器人导纳控制方法及系统
CN113995629B (zh) * 2021-11-03 2023-07-11 中国科学技术大学先进技术研究院 基于镜像力场的上肢双臂康复机器人导纳控制方法及系统
CN114392137A (zh) * 2022-01-13 2022-04-26 上海理工大学 一种穿戴式柔性下肢助力外骨骼控制系统
CN114392137B (zh) * 2022-01-13 2023-05-23 上海理工大学 一种穿戴式柔性下肢助力外骨骼控制系统

Similar Documents

Publication Publication Date Title
Ferrati et al. Virtual modelling of a real exoskeleton constrained to a human musculoskeletal model
Cifuentes et al. Multimodal human–robot interaction for walker-assisted gait
EP3705105A1 (de) Steuerungssystem für ein haptisches exoskelett der unteren extremitäten zur rehabilitation oder zum gehen mit verbesserter gleichgewichtskontrolle, mensch-maschine-schnittstelle
Loconsole et al. An EMG-based approach for on-line predicted torque control in robotic-assisted rehabilitation
Prentice et al. Artificial neural network model for the generation of muscle activation patterns for human locomotion
Ruiz-Olaya et al. Upper and lower extremity exoskeletons
Emmens et al. Improving the standing balance of paraplegics through the use of a wearable exoskeleton
Luksch Human-like control of dynamically walking bipedal robots
Camardella et al. Gait phases blended control for enhancing transparency on lower-limb exoskeletons
Analia et al. Design of assistive torque for a lower limb exoskeleton based on motion prediction
Kamnik et al. Rehabilitation robot cell for multimodal standing-up motion augmentation
Yang et al. A centaur system for assisting human walking with load carriage
Kamnik et al. Robot assistive device for augmenting standing-up capabilities in impaired people
Kamnik et al. Human voluntary activity integration in the control of a standing-up rehabilitation robot: A simulation study
Jović et al. Upper and lower body coordination in FES-assisted sit-to-stand transfers in paraplegic subjects—A case study
Li et al. Lower limb exoskeleton hybrid phase control based on fuzzy gain sliding mode controller
Chen et al. Hybrid control for human-powered augmentation exoskeleton
Li et al. Electric stimulation and cooperative control for paraplegic patient wearing an exoskeleton
Lippi et al. Identification of gait phases with neural networks for smooth transparent control of a lower limb exoskeleton
Moreno et al. Introduction to the special section on wearable robots
Hu et al. Research on Power-Assisted Strategy and Device Based on Muscle Synergy
Higgins Characterization, estimation, and realization of human intent for exoskeleton-assisted walking
Zhang et al. Gait Prediction and Assist Control of Lower Limb Exoskeleton Based on Inertia Measurement Unit
Miura et al. Analyzing compensation strategy in impaired walking using a humanoid robot
He et al. Auto-LEE: A novel autonomous lower extremity exoskeleton for walking assistance

Legal Events

Date Code Title Description
STAA Information on the status of an ep patent application or granted ep patent

Free format text: STATUS: UNKNOWN

PUAI Public reference made under article 153(3) epc to a published international application that has entered the european phase

Free format text: ORIGINAL CODE: 0009012

STAA Information on the status of an ep patent application or granted ep patent

Free format text: STATUS: THE APPLICATION HAS BEEN PUBLISHED

AK Designated contracting states

Kind code of ref document: A1

Designated state(s): AL AT BE BG CH CY CZ DE DK EE ES FI FR GB GR HR HU IE IS IT LI LT LU LV MC MK MT NL NO PL PT RO RS SE SI SK SM TR

AX Request for extension of the european patent

Extension state: BA ME

STAA Information on the status of an ep patent application or granted ep patent

Free format text: STATUS: REQUEST FOR EXAMINATION WAS MADE

17P Request for examination filed

Effective date: 20210216

RBV Designated contracting states (corrected)

Designated state(s): AL AT BE BG CH CY CZ DE DK EE ES FI FR GB GR HR HU IE IS IT LI LT LU LV MC MK MT NL NO PL PT RO RS SE SI SK SM TR

GRAP Despatch of communication of intention to grant a patent

Free format text: ORIGINAL CODE: EPIDOSNIGR1

STAA Information on the status of an ep patent application or granted ep patent

Free format text: STATUS: GRANT OF PATENT IS INTENDED

RIC1 Information provided on ipc code assigned before grant

Ipc: A61H 3/00 20060101AFI20210607BHEP

Ipc: B25J 9/00 20060101ALI20210607BHEP

Ipc: A61H 1/02 20060101ALN20210607BHEP

INTG Intention to grant announced

Effective date: 20210629

RIN1 Information on inventor provided before grant (corrected)

Inventor name: MENGA, GIUSEPPE

STAA Information on the status of an ep patent application or granted ep patent

Free format text: STATUS: THE APPLICATION IS DEEMED TO BE WITHDRAWN

18D Application deemed to be withdrawn

Effective date: 20211110