A Neural Network-Based Gait Phase Classification Method Using Sensors Equipped on Lower Limb Exoskeleton Robots

Walking is a fundamental human activity, and analyzing its complexities is essential for understanding gait abnormalities and musculoskeletal disorders. This article delves into the classification of gait phases using advanced machine learning techniques, specifically focusing on dividing these phases into five distinct subphases. The study utilizes data from 100 individuals obtained from an open-access platform and employs two distinct training methodologies. The first approach adopts stratified random sampling, where 80% of the data from each subphase are allocated for training and 20% for testing. The second approach involves participant-based splitting, training on data from 80% of the individuals and testing on the remaining 20%. Preprocessing methods such as Min–Max Scaling (MMS), Standard Scaling (SS), and Principal Component Analysis (PCA) were applied to the dataset to ensure optimal performance of the machine learning models. Several algorithms were implemented, including k-Nearest Neighbors (k-NNs), Logistic Regression (LR), Decision Tree (DT), Random Forest (RF), Support Vector Machine (SVM), Naive Bayes (Gaussian, Bernoulli, and Multinomial) (NB), Linear Discriminant Analysis (LDA), and Quadratic Discriminant Analysis (QDA). The models were rigorously evaluated using performance metrics like cross-validation score, Mean Squared Error (MSE), Root Mean Squared Error (RMSE), accuracy, and R2 score, offering a comprehensive assessment of their effectiveness in classifying gait phases. In the five subphases analysis, RF again performed strongly with a 94.95% accuracy, an RMSE of 0.4461, and an R2 score of 90.09%, demonstrating robust performance across all scaling methods.

The lower limb exoskeleton robot is a new technique of rehabilitation training for lower limb dysfunction, which is paid more and more attention by people. Because the lower limb exoskeleton robot is relatively heavy all over the world, this article designs a new lower limb exoskeleton robot whose hip joint is driven by motor and other joints are unpowered. It effectively reduces the lower limb exoskeleton robot’s weight. The kinematics equation of exoskeleton robot is established by D-H coordinate system, and the correctness of kinematics equation is verified with combining simulation by MATLAB and SolidWorks software. The end coordinate data calculated by kinematics equation is used as a contrast to test, and a good walking effect is achieved.

Stroke is a significant cause of disability worldwide, and stroke survivors often experience severe motor impairments. Lower limb rehabilitation exoskeleton robots provide support and balance for stroke survivors and assist them in performing rehabilitation training tasks, which can effectively improve their quality of life during the later stages of stroke recovery. Lower limb rehabilitation exoskeleton robots have become a hot topic in rehabilitation therapy research. This review introduces traditional rehabilitation assessment methods, explores the possibility of lower limb exoskeleton robots combining sensors and electrophysiological signals to assess stroke survivors' rehabilitation objectively, summarizes standard human-robot coupling models of lower limb rehabilitation exoskeleton robots in recent years, and critically introduces adaptive control models based on motion intent recognition for lower limb exoskeleton robots. This provides new design ideas for the future combination of lower limb rehabilitation exoskeleton robots with rehabilitation assessment, motion assistance, rehabilitation treatment, and adaptive control, making the rehabilitation assessment process more objective and addressing the shortage of rehabilitation therapists to some extent. Finally, the article discusses the current limitations of adaptive control of lower limb rehabilitation exoskeleton robots for stroke survivors and proposes new research directions.

Efficient and high-precision identification of dynamic parameters is the basis of model-based robot control. Firstly, this paper designed the structure and control system of the developed lower extremity exoskeleton robot. The dynamics modeling of the exoskeleton robot is performed. The minimum parameter set of the identified parameters is determined. The dynamic model is linearized based on the parallel axis theory. Based on the beetle antennae search algorithm (BAS) and particle swarm optimization (PSO), the beetle swarm optimization algorithm (BSO) was designed and applied to the identification of dynamic parameters. The update rule of each particle originates from BAS, and there is an individual’s judgment on the environment space in each iteration. This method does not rely on the historical best solution in the PSO and the current global optimal solution of the individual particle, thereby reducing the number of iterations and improving the search speed and accuracy. Four groups of test functions with different characteristics were used to verify the performance of the proposed algorithm. Experimental results show that the BSO algorithm has a good balance between exploration and exploitation capabilities to promote the beetle to move to the global optimum. Besides, the test was carried out on the exoskeleton dynamics model. This method can obtain independent dynamic parameters and achieve ideal identification accuracy. The prediction result of torque based on the identification method is in good agreement with the ideal torque of the robot control.

Active power-assist lower limb exoskeleton robots aim to enhance wearer assistance while ensuring wearer comfort and simplifying the exoskeleton’s design and control. This study proposes an active assistance method known as Equivalent Force on Connection (EFOC). The EFOC method effectively addresses the limitations encountered in conventional Joint Torque Proportional Compensation (JTPC) approaches. These limitations include the necessity for exoskeleton robot configurations to align with human limb structures for parallel assistance at each lower limb joint, as well as the exoskeleton’s inability to contribute a greater proportion of assistance due to the excessive load on specific skeletal and muscular structures, resulting in wearer discomfort. Furthermore, the effectiveness of the EFOC method is evaluated and validated for assistance during both the stance and swing phases of single-leg movements. Finally, the proposed EFOC method is implemented on a hydraulic-driven lower limb exoskeleton robot to assist wearers in squatting, stepping, and jumping locomotion. The experimental results demonstrate that the proposed EFOC method can effectively achieve the desired assistance effect.

Lower limb exoskeleton robot is a robot that functions as a walking tool of movement for the lower of the human body (usually made for patients with stroke and paraplegic). However, as time goes by lower limb exoskeleton robot is no longer reserved for patients with stroke and paraplegic, some of them are made to military requirements or the need to lift heavy stuff. In this case, the lower limb exoskeleton robot is used for rehabilitation aids for people with paraplegic who suffered an accident on the lower part of the human body (lower limb). This paper presents the design and simulation of the lower limb exoskeleton robot using CAD software to design, and Matlab/Simulink for simulation. The first step after the design of the lower limb exoskeleton robot is completed, we export the file of robot design to the SimMechanics (using first generation). After we obtain the results of the export file, we give a Proportional-Integral-Derivative (PID) controller at each revolute joint. Then, we use a Matlab/Simulink to analyze the response system of PID controller. For the stability analysis of controller, we plot the transfer function of system into the root locus method. Based on the results of simulation, the PID controller has been able to making each revolute joint of the lower limb exoskeleton robot to become stable.

Lower limb exoskeleton robots help with walking movements through mechanical force, by identifying the wearer’s walking intention. When the exoskeleton robot is lightweight and comfortable to wear, the stability of walking increases, and energy can be used efficiently. However, because it is difficult to implement the complex anatomical movements of the human body, most are designed simply. Due to this, misalignment between the human and robot movement causes the wearer to feel uncomfortable, and the stability of walking is reduced. In this paper, we developed a two degrees of freedom (2DoF) ankle exoskeleton robot with a subtalar joint and a talocrural joint, applying a four-bar linkage to realize the anatomical movement of a simple 1DoF structure mainly used for ankles. However, bidirectional tendon-driven actuators (BTDAs) do not consider the difference in a length change of both cables due to dorsiflexion (DF) and plantar flexion (PF) during walking, causing misalignment. To solve this problem, a BTDA was developed by considering the length change of both cables. Cable-driven actuators and exoskeleton robot systems create uncertainty. Accordingly, adaptive control was performed with a proportional-integral-differential neural network (PIDNN) controller to minimize system uncertainty.

Accurate detection of human motion states and prediction of intentions are crucial for the natural walking assistance provided by lower-limb exoskeletons. Several methods for gait task recognition, phase estimation, and motion trajectory prediction have been proposed, but few studies have considered all of these simultaneously. This article proposes a novel scheme for gait recognition and trajectory prediction, capable of detecting gait events, recognizing multiple gait tasks, estimating gait phases, and predicting trajectories. First, this scheme proposes a gait task recognition method based on gait events and a multilayer perceptron neural network (MLPNN), which reduces computational cost and improves recognition stability. Next, a gait phase estimation method based on resettable symmetric adaptive oscillators (RSAOs) is introduced, which offers faster convergence speed and smaller error compared to traditional adaptive oscillators (AOs). Subsequently, a trajectory prediction method with adaptive dynamic motion primitives (ADMP) and feedback regulation is proposed. Finally, gait task recognition, phase estimation, trajectory prediction, and comprehensive online experiments are carried out respectively. The online experimental results show that the recognition accuracy of gait tasks can reach 99.2%, the phase estimation error is less than 0.3652 rad, and the trajectory estimation error is less than 1.1814°.

BackgroundEvidence is emerging for the use of overground lower limb robotic exoskeletons in the rehabilitation of people with spinal cord injury (SCI), with suggested benefits for gait speed, bladder and bowel function, pain management and spasticity. To date, research has focused on devices that require the user to support themselves with a walking aid. This often precludes use by those with severe trunk, postural or upper limb deficits and places the user in a suboptimal, flexed standing position. Free-standing exoskeletons enable people with higher level injuries to exercise in an upright position. This study aimed to evaluate the feasibility of therapy with a free-standing exoskeleton for those with SCI, and to determine the potential health-related benefits of this intervention.MethodsThis 12-week intervention study with 12-week waitlist control and 12-week follow up, provided people with SCI scoring < 5 on the mobility section of the spinal cord independence measure (SCIM-III) twice weekly therapy in the REX (Rex Bionics, Auckland, NZ), a free-standing lower limb robotic exoskeleton. The primary outcome measure of interest was function, as measured on the SCIM-III. A battery of secondary outcomes was included. Participants also completed a survey on their perceptions of this treatment modality, to determine acceptability.ResultsForty-one potential participants were screened for eligibility. Two females (one ASIA A, one ASIA C) and one male (ASIA B) completed all 24 intervention sessions, and the follow up assessment. One participant showed positive trends in function, fatigue, quality of life and mood during the intervention phase. Grip and quadriceps strength, and lower limb motor function improved in another. Two improved their percentage of lean body mass during the intervention phase. Remaining results were varied across patients, time points and outcomes. The intervention was highly acceptable to all participants.ConclusionWith three of 41 potential participants being eligible and completing this study, our results show that there are potential benefits of exercise in a free-standing exoskeleton for people with severe mobility impairment due to SCI, for a small subset of patients. Further research is warranted to determine those most likely to benefit, and the type of benefit depending on the patient characteristics.Trial registration The trial was registered prospectively on 20 April 2018 at www.anzctr.org.au/ (ACTRN12618000626268)

A wearable lightweight lower-limb exoskeleton (LLEX) robot that can help paraplegic patients perform daily activities such as standing/sitting, walking, stair ascending/descending is proposed in this paper firstly. In order to ensure comfort between the exoskeleton and the wearer, a novel configuration of the exoskeleton robot that combines active and passive joints, adaptive structure size, and high compatibility with human physiological parameters is proposed. Secondly, this paper proposes a dynamic gait planning algorithm for stairs ascending and descending based on hip and ankle motion position planning, which is implemented to LLEX. It is intended to provide paraplegic patients with personalized gaits for independently ascending/descending stairs of diverse sizes with the help of LLEX. The system employs the inertial measurement unit (IMU) module located on the back to identify the wearer's movement intention while using the depth camera located on the front of the LLEX to detect heights and depths of different stairs for a dynamic gait. Experiments have been carried out to prove the feasibility and pervasiveness of the gait planning algorithm. The application of LLEX can help patients with various physiological characteristics to complete the task of ascending and descending stairs.

Research on robotic exoskeletons has rapidly expanded over the previous decade. Advances in robotic hardware and energy supplies have enabled viable prototypes for human testing. This review paper describes current lower limb robotic exoskeletons, with specific regard to common trends in the field. The preponderance of published literature lacks rigorous quantitative evaluations of exoskeleton performance, making it difficult to determine the disadvantages and drawbacks of many of the devices. We analyzed common approaches in exoskeleton design and the convergence, or lack thereof, with certain technologies. We focused on actuators, sensors, energy sources, materials, and control strategies. One of the largest hurdles to be overcome in exoskeleton research is the user interface and control. More intuitive and flexible user interfaces are needed to increase the success of robotic exoskeletons. In the last section, we discuss promising future solutions to the major hurdles in exoskeleton control. A number of emerging technologies could deliver substantial advantages to existing and future exoskeleton designs. We conclude with a listing of the advantages and disadvantages of the emerging technologies and discuss possible futures for the field.

This paper proposes a gait phase classifier using a Recurrent Neural Network (RNN). Walking is a type of dynamic system, and as such it seems that the classifier made by using a general feed forward neural network structure is not appropriate. It is known that an RNN is suitable to model a dynamic system. Because the proposed RNN is simple, we use a back propagation algorithm to train the weights of the network. The input data of the RNN is the lower body's joint angles and angular velocities which are acquired by using the lower limb exoskeleton robot, ROBIN-H1. The classifier categorizes a gait cycle as two phases, swing and stance. In the experiment for performance verification, we compared the proposed method and general feed forward neural network based method and showed that the proposed method is superior.

Robotic lower limb exoskeletons have been built for augmenting human performance, assisting with disabilities, studying human physiology, and re-training motor deficiencies. At the University of Michigan Human Neuromechanics Laboratory, we have built pneumatically-powered lower limb exoskeletons for the last two purposes. Most of our prior research has focused on ankle joint exoskeletons because of the large contribution from plantar flexors to the mechanical work performed during gait. One way we control the exoskeletons is with proportional myoelectric control, effectively increasing the strength of the wearer with a physiological mode of control. Healthy human subjects quickly adapt to walking with the robotic ankle exoskeletons, reducing their overall energy expenditure. Individuals with incomplete spinal cord injury have demonstrated rapid modification of muscle recruitment patterns with practice walking with the ankle exoskeletons. Evidence suggests that proportional myoelectric control may have distinct advantages over other types of control for robotic exoskeletons in basic science and rehabilitation.

Gait phase is an important state variable that helps exoskeletons provide assistance to users. Exoskeleton systems need to determine the assistance time and calculate the reference assistance force based on the gait phase. A large number of studies have used Finite State Machines (FSMs) or Adaptive Oscillators (AOs) to estimate gait phase. However, FSMs require manual calibration of gait events, and AOs can only estimate the phase of stable gait in real time. Both are difficult to adapt to irregular changes in gait. Deep learning algorithms can learn the nonlinear characteristics of kinematic data to estimate joint torque, gait phase, etc. The control algorithm based on Long Short-Term Memory (LSTM) Neural Network can automatically detect gait events according to the joint angle collected by Inertial Measurement Unit (IMU), effectively responding to dynamic changes in gait. This paper designed a Light-Weight Long Short-Term Memory (LWLSTM) Neural Network to estimate gait phase precisely and reduce training time of LSTM model. The experimental results showed that the Root Mean Square Error (RMSE) of LWLSTM was 8. 0 0 3 3%, quite lower than Convolutional Neural Network (CNN) and AO.

Context: Robot-assisted gait rehabilitation is a rapidly evolving field that aims to enhance therapeutic outcomes through the integration of advanced robotic systems. These systems reduce the physical burden on therapists while enabling intensive and repetitive patient training. However, current methods often require manual adjustments tailored to individual gait patterns, limiting their generalizability and efficiency. Objective: This paper provides a comprehensive review of control approaches for robot-based gait rehabilitation and locomotion assistance, focusing on improving adaptability and human-robot synchronization. The goal is to identify effective techniques and methods that optimize exoskeleton performance and improve patient outcomes. Methodology: A state-of-the-art review was conducted through a structured literature search across multiple scientific databases to identify research on lower-limb exoskeletons, centering on control strategies for robot-assisted gait rehabilitation. The search initially employed broad keywords such as “lower limb exoskeletons,” “extremity lower exoskeletons,” and “lower limb robotic exoskeletons,” among others. It was then refined using more specific terms including “lower limb exoskeleton control,” “assisted gait control,” “assistance control strategy,” “wearable robots control,” and “robotic rehabilitation,” among others. Well-defined inclusion criteria ensured thematic relevance, methodological rigor, and comprehensive coverage of recent advancements. A total of 77 studies meeting these criteria were analyzed to identify and categorize control approaches, and to discuss their scope, advantages, and limitations rather than directly assessing their clinical effectiveness or patient outcome improvements. Future research should focus on data-driven control methods to address individual variability and improve exoskeleton adaptability. Results: The review highlights a taxonomy of essential control strategies for robot-assisted gait, such as trajectory tracking based on position, force-impedance control, control driven by bio-signals, and adaptive control techniques. Real-time adaptive strategies, such as Velocity Field Control and Assist-As-Needed approaches, demonstrate strong potential for improving synchronization and reducing manual tuning requirements. Conclusions: Adaptive control strategies, particularly Assist-As-Needed controllers, are effective in encouraging active patient participation and optimizing rehabilitation outcomes. Financing: This research work received funding from Universidad del Valle under the project "Plataforma tecnológica modular para la valoración objetiva de la marcha humana" (C.I. 21259) and from the Ministry of Science, Technology, and Innovation (Minciencias - Open call 647).