System for controlling a one-legged exoskeleton using computer vision

Abstract

A system (100) for controlling a single leg exoskeleton (200) for walking assistance of a user is disclosed. The exoskeleton (200) comprises a joint unit movable by a first actuator (210) and a second actuator (220). The system (100) comprises a plurality of inertial mass units (302, 304, 306) and a pressure sensor (308) to collect gait information. The system (100) comprises a computer vision unit (110) to collect environmental information associated with a user path and to detect a terrain hazard (122) and its geometric properties. Using additional gait information collected from the actuators (210, 220), a user gait phase (126) is inferred via a processor unit (120) comprised in the system (100). Based on at least the geometric properties of the terrain hazard (122), a user profile (120) with user limitations, and a user intent (128) obtainable from an interface (270, 260), a response with selective instructions for the actuators (210, 220) is generated.

Description

Technical Field

[0001] This invention belongs to the field of orthotic devices and gait assistive devices. In particular, this invention relates to the control of movement of an exoskeleton used only for one leg. Background Technology

[0002] Many health conditions (such as stroke or injury) can result in a form of gait impairment in which full mobility in one of the lower limbs is maintained. For example, many people with severe or mild hemiplegia have this form of gait impairment. They can typically maintain balance while standing without assistance, and some may even walk with great difficulty. Currently, many orthotics for these individuals, targeting only one leg, are largely mechanical and do not include electronic components. More advanced orthotics, such as exoskeletons, sometimes called robotic orthotics or simply robotic, include electromechanical components for performing and / or supervising assistive maneuvers. However, these are mostly designed for both legs. Therefore, the gait impairment needs of this population, where only one leg is affected, remain largely unaddressed. Summary of the Invention

[0003] This invention was conceived in response to the limitations of the prior art. In particular, it is designed to allow users to walk safely and without excessive exertion for extended periods.

[0004] Due to the asymmetric mobility of potential users, there are many factors to consider when operating a single-leg exoskeleton. Among other things, depending on the phase of the gait, certain movements need to be compensated for, the user's gait patterns can vary considerably, and the presence of terrain irregularities during different phases of the gait may require selective responses in a short period of time. For these tasks, inferring the user's movement intentions allows time to be gained and preventative measures taken when terrain hazards (such as obstacles or surfaces that are problematic for the use of the exoskeleton) are identified.

[0005] The object of this invention is a system for controlling a single-leg exoskeleton. The system can cooperate with an exoskeleton comprising joint units movable by a first actuator and a second actuator. The system includes several inertial mass units and pressure sensors to collect gait information. The system also includes a computer vision unit to collect environmental information associated with the user's path and to detect terrain hazards and their geometric properties. Using additional gait information collected from the actuators, the user's gait phase is inferred via a processor unit included in the system. A response with selective commands to the actuators is generated, based at least on the geometric properties of the terrain hazards, a user profile with user restrictions, and user intent that can be obtained from the interface or inferred by processing pressure sensor and IMU information. Attached Figure Description

[0006] Embodiments of the invention are described in the accompanying drawings by way of example only, wherein similar elements are numbered in the same manner in the various figures: Figure 1 : A schematic block diagram of the system according to the present invention.

[0007] Figure 2A : A side view of an example of a single-leg exoskeleton of the present invention. Figure 2B The image shows a side view of an example of a single-legged exoskeleton and its quadrupedal cane that may be used in accordance with the present invention. Figure 2C Body planes and body axes.

[0008] Figure 3A : The phase of the normal gait cycle. Figure 3B : A diagram showing hip flexion / extension during the normal gait cycle. Figure 3C : A diagram showing knee and hip flexion / extension during the normal gait cycle.

[0009] Figure 4 Example table with a list of terrain hazards that can be identified using computer vision units.

[0010] Figure 5 : A schematic block diagram illustrating the implementation of a motion strategy in a system for controlling an exoskeleton according to the present invention.

[0011] Figure 6 Example table containing user profile information.

[0012] Figure 7 : A practical example of using motion strategies for terrain detection and response in a system for controlling an exoskeleton according to the present invention.

[0013] Glossary of Numerical Reference Symbols: 100 system.

[0014] 102 users.

[0015] 110 Computer Vision Units.

[0016] 112 AI computing units.

[0017] 118 camera.

[0018] 120 processor units.

[0019] 190 movement strategy.

[0020] 200 exoskeletons.

[0021] 202 Hip Unit.

[0022] 204 femoral units.

[0023] 204a Femoral extension.

[0024] 206 Tibial and Fibular Units.

[0025] 206a Tibia and fibula extension.

[0026] 208 foot units.

[0027] 210 First Actuator.

[0028] 220 Second Actuator.

[0029] 230 Adjustable belt.

[0030] 240 back unit.

[0031] 242 battery.

[0032] 250 waist unit.

[0033] 260 Exoskeleton Embedded Interface.

[0034] 270 User Interface.

[0035] 272. Four-legged walking stick.

[0036] 302 Back IMU.

[0037] 304 femoral IMU.

[0038] 306 Tibia and fibula IMU.

[0039] 308 Foot Pressure Sensor. Detailed Implementation

[0040] Referring to the figures above, various embodiments of the present invention are presented without limitation for better understanding.

[0041] Figure 1 A system 100 for controlling the movement of the exoskeleton 200 to provide gait assistance to the user wearing it is illustrated schematically.

[0042] Some types of exoskeletons 200 have joints, driven via actuators, contained within the exoskeleton legs. In this example, there is a first actuator 210 for moving the hip unit 202 relative to the femoral unit 204, and a second actuator 220 for moving the femoral unit 204 relative to the tibia and fibula unit 206. The proposed system 100 can be coupled with these existing actuators 210, 220 for gait assistance. If the exoskeleton 200 does not include them or cannot be coupled to the system 100, the exoskeleton should be modified to accommodate controllable actuators 210, 220. Without limitations, Figure 1Assume that the existing actuators 210 and 220 of the exoskeleton 200 are compatible with the system 100.

[0043] To provide appropriate assistance to the user, system 100 detects the presence of hazardous terrain based on environmental information and determines its response based on the current user gait stage, the user's movement intention, and a user profile previously stored in memory 130. A key aspect of achieving this is the computer vision unit 110, which collects environmental information for system 100 to avoid or mitigate risks and takes initial action by accordingly modifying the operation of the exoskeleton 200. The computer vision unit 110 may include an artificial intelligence (AI) computing unit 112 and one or more cameras 118 equipped with depth sensing technology to allow depth data to be captured, typically using infrared spectroscopy (e.g., Intel RealSense). The cameras 118 can provide terrain geometry attributes of paths or areas where the user might walk. Terrain geometry attributes are typically the slope or irregularity of the terrain surface and the depth, width, and height of objects on the terrain.

[0044] The computer vision unit 110 includes algorithms that can be trained by machine learning and can classify images from the nearby environment.

[0045] In the embodiments, nine different terrain hazards are defined (see Figure 4 Some terrain irregularities and small obstacles along the path followed by the exoskeleton 200, such as ramps, curbs, bumps, potholes, rocks, branches, or wet surfaces, may represent sources of safety hazards or discomfort for the user. In such cases, system 100 can activate selective responses to actuators 210 and 220. Tests performed have shown that the computer vision unit 110 is highly reliable and can classify these nine terrain hazards with high accuracy (typically greater than 95%) in near real-time (typically less than 200 ms).

[0046] System 100 can operate according to a hierarchical scheme with different layers and their corresponding functions. For example, system 100 can have three layers. The higher-level control layer determines the motion strategy by means of different information (including information from the surrounding environment) (generally, terrain information collected by computer vision unit 110 covers the 3D spatial area surrounding the user, typically with a field of view of 60° vertically, 90° horizontally, and a depth between 0.2 cm and 4 m). Each motion strategy involves determining the biomechanics and corresponding kinematics of movement necessary to respond to the considered terrain irregularities, barriers, and obstacles; that is, defining the movement patterns, positions, velocities, accelerations, and trajectories of body segments and joints required to execute such a strategy.

[0047] Once the strategy is determined, system 100 uses intermediate and low-level control layers responsible for calculating the dynamic equations that the actuator should follow and the corresponding joint positions, speeds, and torques, and controlling its execution.

[0048] Terrain hazards such as stairs, walls, doors, and large moving objects are also included in the training process of algorithms typically performed by the AI ​​computing unit 112 of the computer vision unit 110 (such as the NVIDIA Nano Jetson Orin). Sometimes, specific terrain geometry properties such as excessive height, depth, length, or slope may mean generating an immediate stop response for the exoskeleton 200 (e.g., walls on a path).

[0049] To maintain safety and comfort, the kinematics of joint units 202, 204, and 206 are influenced by the user profile (e.g., age, anthropometrics, health status, gait pattern settings, fit, etc.). Key parameters such as toe and heel clearance on inclined surfaces or balance are significantly affected by age and physical condition and can greatly influence movement strategies (generally, older and poorer individuals tend to have lower toe and heel clearances and poorer balance). Other parameters, such as walking speed (defined by time of step in the gait pattern settings of the user profile), can also strongly influence movement strategies due to delays in achieving certain responses that may be too long for safe and / or comfortable execution.

[0050] Another important aspect of System 100 is the execution of algorithms for inferring gait phases and algorithms for inferring the user's subsequent movement intentions. The inputs used to infer gait phases are sensor data from the exoskeleton leg related to the kinematics of the disabled side and pelvis, as well as data from the actuators. Details of the algorithms for inferring user intentions are beyond the scope of this disclosure.

[0051] More specifically, sensor data is primarily obtained via the femoral IMU 304 located on the femoral unit, the tibia and fibula IMU 306 located on the tibia and fibula unit, and the dorsal IMU 302 located in the dorsal unit 240, and also from the pressure sensor 308 located on the foot unit 208 (on the sole of the shoe). The algorithm output is categorized into seven classes corresponding to seven normal gait phases: loading response, mid-standing, late-standing, pre-swing, early-swing, mid-swing, and late-swing, as follows: Figure 3A As shown in the image.

[0052] In addition, the algorithm can provide two additional categories related to the user's movement intention, depending on whether he / she wants to continue walking or stop: walk, stop. The back IMU 302, located on the back unit 240 near the pelvic region, is a major contributor to this.

[0053] System 100 also considers input signals received from one or more interfaces 260, 270 to follow predefined commands, which, for example, may be: walk, stop, sit down, or stand up, and—unless unsafe—take precedence over other automatic responses calculated by system 100.

[0054] Figure 2A A side view of a single-leg exoskeleton 200 worn by user 102 is shown, which can be fitted with an embodiment of a system for controlling it according to the present invention. Figure 2B A side view of the same exoskeleton 200 is shown. These two figures will be together with... Figure 2D Describe them together. Figure 2D The same exoskeleton is shown without a user present.

[0055] Exoskeleton 200 is an orthopedic robotic structure used to move a user's leg in a gait-like motion that mimics human gait. Exoskeleton 200 is anatomically conceived and designed to be attached or secured to the user's body above clothing, typically also used on the legs, waist, and preferably on the torso and / or shoulders of the user 102 via adjustable straps 230. Exoskeleton 200 comprises several articulated and non-articulated units, some of which are driven by motors via actuators. The user is responsible for changing direction, as the exoskeleton is preferably designed to assist in forward movement (sagittal axis).

[0056] Figure 2A-2B and Figure 2DThe example exoskeleton 200 includes a lumbar unit 250, a back unit 240, a hip unit 202, a femoral unit 204, a tibia and fibula unit 206, and a foot unit 208. The exoskeleton 200 includes a first actuator 210 for moving the femoral unit 204 relative to the hip unit 202. The exoskeleton 200 includes a second actuator 220 for moving the tibia and fibula unit 206 relative to the femoral unit 204. In other embodiments, the foot unit 208 can be moved relative to the tibia and fibula unit 206 via the actuator. To save weight, this example exoskeleton 200 does not include a third actuator for the user's ankle. However, the length between the foot unit 208 and the tibia and fibula unit 206 can be adjusted and set according to the anatomy and needs of the user 102. Optionally, the foot unit 208 can be modified to include a durable but resilient material (e.g., carbon fiber) to act as a spring that accumulates energy during the standing phase and releases the accumulated energy at the start of a swing. The user's foot can move within the foot unit 206 according to gait movements. Some units may be extendable to adjust for different user heights. In this embodiment, there are femoral extensions 204a and tibia / fibula extensions 206a, which are typically implemented as telescopic mechanisms, as well as foot extensions.

[0057] The exoskeleton 200 allows the femoral unit 204 and the tibia and fibula unit 206 to rotate in the sagittal plane. The sagittal plane is the anatomical plane that divides the body into left and right segments. Figure 2C As shown, the sagittal plane is perpendicular to the transverse and coronal planes along the anatomical axis.

[0058] The system controls the exoskeleton to allow movement between joint units to be electronically and / or mechanically restricted within a range, thereby ensuring the safety of user 102. Typically, the first range is between 15 degrees of extension and 120 degrees of flexion for the hip, and between 0 degrees of extension and 90 degrees of flexion for the knee. These ranges can be selected individually as part of the settings.

[0059] Based on information acquired by several sensors (e.g., IMU, pressure sensors for the sole of the shoe) and a computer vision unit, as well as by gait phase and user intent modules and interfaces, joint movement is electronically controlled by system 100. Gait patterns can be predefined and customized for each user based on their needs and preferences (e.g., health status). Gait patterns can also adapt in real time to the user's movement intentions, user commands, or alternative movement strategies (e.g., in the event of detected terrain hazards).

[0060] In this embodiment, three inertial mass units (IMUs) are present: a dorsal IMU 302, a femoral IMU 304 located on the femoral unit 204, and a tibia / fibula IMU 306 located on the tibia / fibula unit 206. Several pressure sensors 308 are also attached to the sole of the foot unit 208. These are considered basic sensors and computer vision units that collaborate with the processor unit to allow interaction with the environment, thereby monitoring and / or modifying the exoskeleton's operation during the gait cycle as needed.

[0061] In this embodiment, each actuator 210, 220 of the exoskeleton 200 includes a DC motor, a gearbox, control electronics, and one or more encoders. The actuators 210, 220 are connected to both a power source (e.g., a DC battery 242) and a processor unit via signal and power wiring. The power source is typically located on the back unit 240 of the exoskeleton 200. The processor unit typically includes a microprocessor and a communication card (Bluetooth, WiFi, etc.). An embedded interface 260 located on the back unit 240 can be used for operation of the exoskeleton 200 by an assistive human.

[0062] A user interface 270 may also be provided to the user to operate the exoskeleton 200. For example, in some embodiments, the user interface may be an application installed on a mobile phone and communicate via Bluetooth or other wireless protocols. Additionally or alternatively, in some embodiments, the exoskeleton 200 may be associated with a quadrupedal cane 272 for the user to hold when standing or walking. The quadrupedal cane may include a user interface in the form of a keypad and display in the handle and communicate via Bluetooth or other wireless protocols.

[0063] User interface 270 and embedded interface 260 allow for basic movement or general functional operations, such as stopping, walking, sitting, standing, and system status such as battery level or malfunction. The user interface may include an on / off switch, an emergency stop button, LEDs, a buzzer, and haptic indicators. Two-way communication between the exoskeleton 200 and the user is important to enable collaborative work. The exoskeleton 200 can communicate anticipated responses to the user to avoid accidents and improve cooperation. An example is when the computer vision unit 110 detects an obstacle and the exoskeleton 200 initiates an obstacle avoidance response, but the obstacle is not noticed by the user. Therefore, it is important to provide information in a non-intrusive manner so that the system 100 can be updated to abort unnecessary obstacle avoidance responses. Ideally, communication can use various channels to ensure information reception and to make it as non-intrusive as possible. Preferably, a combined haptic-audio and / or haptic-visual strategy can be employed.

[0064] Another example scenario is a user facing a curb. Once the camera detects the curb, vibration occurs (so the user is aware of a possible change in kinematics), and the user can then decide to look to confirm the curb detection. Depending on user preference, system 100 can be configured to require or not require user confirmation before performing some response within the exoskeleton.

[0065] Figure 3A The phases of a typical gait cycle according to a conventional gait model are shown. The gait cycle is measured from the moment one foot touches the ground until the next touch of the same foot. The gait cycle can be broken down into a stance phase and a swing phase, which alternate for each leg. During the left stance phase, the left foot remains on the ground. During the right swing phase, the right foot remains in the air. Similarly, for the opposite foot, there is a right stance phase and a left swing phase. Therefore, during the right stance phase, the right foot remains on the ground, and during the left swing phase, the left foot remains in the air.

[0066] Furthermore, the stance phase includes several sub-phases: the double-support sub-phase, the right single-support sub-phase, the double-support sub-phase, and the left single-support sub-phase. The gait cycle has several periods: the weight-bearing reaction period, the middle of the stance phase, the end of the stance phase, the pre-swing phase, the early swing phase, the middle of the swing phase, and the end of the swing phase.

[0067] Figure 3B A diagram showing hip flexion / extension during a normal gait cycle is shown. Figure 3C A diagram representing knee flexion / extension during a normal gait cycle is shown. The system's processor unit is configured to implement a movement strategy that simulates these patterns, making appropriate adjustments based on user profiles (e.g., anthropometry, health limitations, etc.) during the gait cycle under normal conditions (e.g., walking on safe terrain).

[0068] When a terrain hazard is detected, or when a user generates commands for the exoskeleton via the interface, the system responds based on several factors. Specifically, the system response can vary depending on the specific attributes of the terrain hazard, the current gait stage, and / or the user's intent. The processor unit manages this response according to a motion strategy.

[0069] Figure 4A table is shown with a list of examples of terrain hazards in the first column, which can be identified by the system based on environmental information collected by a computer vision unit. The second column of the table defines the attributes of the first column, and the third column provides annotations about several objects that might belong to a certain category. These attributes indicate whether additional variables need to be considered. For example, for bumps or potholes, it is important to consider not only the height (positive for bumps and negative for potholes) but also the depth (in the walking direction) to assess the best strategy for a user facing it, because if the depth of the obstacle is greater than the step length a user can take, then it may be dangerous. Typically, the computer vision unit is calibrated to detect when an obstacle meets a specific criterion.

[0070] In the case of a surface, there may be a positive or negative tilt in the direction of travel (pitch: up / down) or in its lateral direction (yaw), and it can also be positive or negative (clockwise / counterclockwise).

[0071] Figure 5 A block diagram of an embodiment of a system 100 for controlling an exoskeleton 200 is shown. Processor unit 120 generates a set of instructions for actuators 210, 220 of exoskeleton 200 based on each terrain hazard 122 and other inputs such as user profile 124, gait phase 126, and user intent 128.

[0072] A common source of safety and comfort issues is terrain, and therefore it will be discussed in more detail. The computer vision unit 110 collects environmental information near the user's path, and it is programmed to identify whether this information matches terrain hazards 122, such as barriers, bumps, potholes, uneven surfaces, slippery surfaces, obstacles, curbs, slope pitch, and surface roll. These terrain hazards can be included in, for example... Figure 4 The list shown can be updated.

[0073] When a terrain hazard is detected, the motion strategy algorithm 190 in processor unit 120 analyzes it, taking into account gait phase, user movement intent, and user profile input. Therefore, Figure 3B-3C The normal gait cycle movement represented in the diagram may be altered. For example, the actuators may be instructed to modify these gait patterns and follow other patterns to prevent the user from falling or dropping.

[0074] Each movement strategy involves determining the biomechanics and corresponding kinematics of the movement necessary to respond to the terrain irregularities, barriers, and obstacles under consideration.

[0075] Figure 6A table showing an example of information from a user profile is provided. Several parameters affect the appropriate response the system can provide to the user in the event of a specific terrain hazard. A particular key parameter is step time, which can limit the possible response the system can provide to the user to overcome a detected terrain hazard. This step time is typically on the order of 700-900 ms.

[0076] As described above, other specific implementations and different realizations of the system are possible.

[0077] Figure 7 This illustrates a practical example of how system 100 operates when a specific obstacle is detected and a corresponding response is provided.

[0078] System 100 is coupled to the single-leg exoskeleton that the user is wearing. The single-leg exoskeleton can have... Figure 2A-2B The system 100 (or the single-leg exoskeleton itself) includes a first actuator 210 responsible for moving the hip unit relative to the femoral unit, and a second actuator 220 responsible for moving the femoral unit relative to the tibia and fibula units. The processor unit 110 controls the actuators 210 and 220.

[0079] System 100 includes a computer vision unit 110 that collects environmental information near the user. In this case, a terrain hazard 122 is detected within a configurable safe range (e.g., 0.6 meters from the user). This is assumed to be an obstacle with geometric properties of 6 cm height and 6 cm depth. This information is used by the processor unit 110 to generate a selective response based on other considerations.

[0080] User intent 128 is considered by system 100 via, for example, a user interface 270 present in the cane (e.g., a “walk” button is pressed).

[0081] System 100 includes three IMUs 302, 304, and 306 associated with the exoskeleton unit and with a pressure sensor (308). They collect gait information generated by user / exoskeleton motion, typically acceleration α for each IMU along three axes. xyz angular velocity w xyz and position O xyz The processor unit 120 processes gait information from actuators 210, 220, IMUs 302, 304, 306, and pressure sensor 308, and infers the gait phase 126 of the gait cycle that the user is currently performing (e.g., he is in the early swing phase of the gait cycle).

[0082] To generate a selective response effective for a particular user, processor unit 110 processes user profile 120 to obtain anatomical constraints and / or physical information recorded during system 100 setup. This information may be general, such as age, degree of disability, anthropometrics, health status, injured side, foot size, or more specific, such as gait pattern setting steps (e.g., statistics obtained from tests conducted on the user). Note that system 100 typically uses user age to adjust the applicability of potential responses by setting safety margins (e.g., responses that are close to user profile values ​​may fail more frequently in older individuals).

[0083] The processor unit 110 infers the maximum toe gap (e.g., 10 cm) and the time delay (e.g., 100 ms) for the initial stage of the swing by combining the user profile.

[0084] The inputs, including terrain hazard geometry, user intent, gait phase, and user profile, are processed according to a motion strategy, which in this embodiment may be in the form of a database. That is, these inputs define the output of the processor 110, which is transformed into a response with instructions for the actuators 210 and 220.

[0085] In this scenario, the motion strategy has a time delay (e.g., 150 ms) and implies an increased toe gap (e.g., up to 8 cm). Processor unit 110 determines that the terrain hazard 122 can be overcome. Therefore, it activates the motion strategy. Consequently, a response is generated with instructions for actuators 210 and 220 (e.g., a first actuator torque of 30 N and a second actuator torque of 60 N).

Claims

1. A system (100) for controlling a single-leg exoskeleton (200) to assist a user (102) in walking, wherein the exoskeleton (200) comprises: The unit comprises a lumbar unit (250), a hip unit (202), a femoral unit (204), a tibia and fibula unit (206), and a foot unit (208), wherein the hip unit (202) is movable relative to the back unit (240), the femoral unit (204) is movable relative to the hip unit (202), the tibia and fibula unit (206) is movable relative to the femoral unit (204), and the foot unit (208) is movable relative to the tibia and fibula unit (206); The exoskeleton (200) or system (100) includes: A first actuator (210) for moving the femoral unit (204) relative to the hip unit (202), and a second actuator (220) for moving the tibia and fibula unit (206) relative to the femoral unit (204). The system (100) includes: Multiple inertial mass units (302, 304, 306) and a pressure sensor (308) are configured to collect a first set of gait information. The computer vision unit (110) is configured to collect environmental information associated with the user's path, detect terrain hazards (122) within a safe range, and provide multiple geometric attributes of the terrain hazards (122); The processor unit (120) is configured to collect a second set of gait information from the first actuator (210) and the second actuator (220), and to infer a user gait phase (126) using the first set of gait information and the second set of gait information, wherein the processor unit (120) is further configured to generate a response, wherein the response includes selective instructions for the first actuator (210) and the second actuator (220) based at least on geometric properties of terrain hazards (122), a user profile (120) including user anatomical limitations, and user intent (128).

2. The system for controlling a single-leg exoskeleton (200) according to claim 1, wherein the geometric properties of the terrain hazard (122) obtained by the computer vision unit (110) include a combination of one or more of the following: positive or negative height, depth, positive or negative dip angle in the sagittal plane, upward or downward tilt, and lateral clockwise or counterclockwise rotation.

3. The system for controlling a single-leg exoskeleton (200) according to claim 1 or 2, wherein the user profile (120) includes at least time information of the gait phases of the user's gait cycle, and the maximum range to which the actuators (210, 220) can be applied when moving the associated unit.

4. The system for controlling a single-leg exoskeleton (200) according to any one of claims 1 to 3, further comprising an interface (270, 260) for inputting user intent (128), said user intent (128) including stopping walking and walking.

5. The system for controlling a single-leg exoskeleton (200) according to any one of claims 1 to 4, wherein the user intent is obtained from the interface (270, 260).

6. The system for controlling a single-leg exoskeleton (200) according to any one of claims 1 to 5, wherein the user intent is obtained by processing information from the IMU (302, 304, 306), the pressure sensor (308), and the actuator (210, 220).

7. A single-leg exoskeleton (200) comprising the system according to any one of claims 1 to 5, wherein a computer vision unit (110) is previously trained by a machine learning algorithm to classify detected terrain hazards as barriers, bumps, potholes, uneven surfaces, slippery surfaces, obstacles, curbs, pitching slopes or surface rolls.

8. A single-legged exoskeleton (200), comprising: The unit comprises a lumbar unit (250), a hip unit (202), a femoral unit (204), a tibia and fibula unit (206), and a foot unit (208), wherein the hip unit (202) is movable relative to the back unit (240), the femoral unit (204) is movable relative to the hip unit (202), the tibia and fibula unit (206) is movable relative to the femoral unit (204), and the foot unit (208) is movable relative to the tibia and fibula unit (206); A first actuator (210) for moving the femoral unit (204) relative to the hip unit (202) in the sagittal plane, and a second actuator (220) for moving the tibia and fibula unit (206) relative to the femoral unit (204) in the sagittal plane. The system according to any one of claims 1 to 4, wherein the foot unit (208) includes an elastic portion for receiving the user's foot and accumulating and releasing energy during the gait cycle.

Images