CN116502105A - Exoskeleton gait parameter optimization method for lower limb rehabilitation training - Google Patents

Abstract

The invention discloses an exoskeleton gait parameter optimization method for lower limb rehabilitation training. Firstly, setting an initial gait of a lower limb rehabilitation exoskeleton, and acquiring exoskeleton-gait fitness under the current gait after a user wears the lower limb rehabilitation exoskeleton to perform a walking experiment; then, according to the exoskeleton-gait fitness corresponding to the gait of the current walking experiment, using a Gaussian mixture model to cluster the gait of the walking experiment in a gait library to obtain a gait cluster, and updating the utility degree of the corresponding gait in the gait library; then, the gait in the gait library is sampled to obtain a sampled gait set, then the gait with the maximum utilization degree is selected from the gait set, the gait is continuously optimized, and finally, the gait parameter corresponding to the final gait is taken as the optimal gait parameter. The method can predict the gait parameters of the exoskeleton of the lower limb, which meet the requirements of the user, and has low calculation cost, high convergence speed and higher robustness.

Description

An optimization method of exoskeleton gait parameters for lower limb rehabilitation training

technical field

The invention belongs to the technical field of rehabilitation robots and relates to a method for optimizing gait parameters of exoskeleton, in particular to a method for optimizing gait parameters of exoskeleton for rehabilitation training of lower limbs.

Background technique

Due to damage to the central nervous system, only 3-10% of patients with spinal cord injury and stroke can recover to walk 50m independently in three years after the disease. Long-term, standardized and scientific rehabilitation exercise training can rebuild the relevant functional areas of the nervous system, so that The patient regains voluntary movement. China is rapidly entering a deeply aging society, and the disabled and semi-disabled elderly need rehabilitation equipment and basic rehabilitation services. The lower limb rehabilitation exoskeleton can perform systematic rehabilitation training for patients without relying on physical therapists, and gait design is one of the key technologies of the lower limb rehabilitation exoskeleton. Gait is the trajectory that the lower limb rehabilitation exoskeleton drives the wearer to walk. Personalized gait design for patients plays an important role in improving the efficiency of rehabilitation training and speeding up the recovery of lower limbs.

However, there is a lack of consideration of individualized gait parameters in the current technology. Due to the different structural parameters of the legs and limbs of different patients, there are also individual differences in the degree of motor function impairment. Only effective gait optimization methods for different patients can play a role. Ideal recovery effect. Generating gait only based on the patient's physical characteristic parameters cannot fully reflect the user's preference for gait parameters. It is necessary to use the feedback information of the user's preference for gait to complete the optimization of the user's gait parameters.

Contents of the invention

In order to solve the problems in the background technology, the present invention proposes an exoskeleton gait parameter optimization method for lower limb rehabilitation training. By quantifying the gait of the exoskeleton, the gait is clustered using the mixed Gaussian model according to the fit between the gait and the exoskeleton-gait, and the utility of the gait is calculated by the multi-armed gambling machine algorithm, and the degree of gait meeting the user's needs is predicted. Posterior probability, optimizing gait parameters. The invention can predict the gait parameters of the lower extremity exoskeleton meeting the needs of users, has low calculation cost, fast convergence speed and high robustness.

To achieve the above object, technical solutions of the present invention include:

1) Set the initial gait of the lower limb rehabilitation exoskeleton, and the user wears the lower limb rehabilitation exoskeleton for a walking experiment to obtain the exoskeleton-gait fit under the current gait;

2) According to the exoskeleton-gait fit corresponding to the gait of the current walking experiment, use the mixed Gaussian model to cluster the gait of the walking experiment in the gait library to obtain the gait cluster;

3) Calculate the utility of the gait cluster and update the utility of the corresponding gait in the gait library;

4) Sampling the gaits in the gait library according to the current utility of the gait to obtain a sampled gait set, and then select the gait with the largest utility from the gait set;

5) According to the gait with the highest current utility, the user wears the lower limb rehabilitation exoskeleton for a walking experiment to obtain the exoskeleton-gait fit under the current gait, and repeats 2)-4) until the preset rounds are reached. The gait parameters corresponding to the final gait are taken as the optimal gait parameters.

The 2) is specifically:

2.1) Use Bayesian information criterion to calculate the corresponding Bayesian information criterion operator before and after doubling the number of clusters of the current mixed Gaussian model, and obtain the operator difference after making a difference between the two, and adjust according to the operator difference to obtain The number of clusters in the final mixed Gaussian model;

2.2) Based on the number of clusters of the current mixed Gaussian model and the exoskeleton-gait fit corresponding to the gait of the current walking experiment, use the gait mixed Gaussian model to cluster the gait of the walking experiment in the gait library, and obtain initial gait clustering;

2.3) Adjust the coverage area of the current gait cluster according to the minimum coverage area, so as to obtain the updated gait cluster as the final gait cluster.

In the above 2.1), the number of clusters of the final mixed Gaussian model is adjusted and obtained according to the operator difference, specifically:

The operator difference is obtained by subtracting the doubled Bayesian information criterion operator from the doubled Bayesian information criterion operator. When the operator difference is less than 0, the current mixed Gaussian model clustering number is clustered After the number is doubled, it is used as the final number of clusters of the mixed Gaussian model, otherwise the current number of clusters of the mixed Gaussian model remains unchanged.

The formula of the gait mixture Gaussian model is as follows:

G={(α ₁ , θ ₁ ) _{(L, H, W, T, f)} , (α ₂ , θ ₃ ) _{(L, H, W, T, f)} ,..., (α _K , θ _K ) _{(L, H, W, T, f)} }

Among them, G represents the gait mixture Gaussian model, (α _K , θ _K ) _{(L, H, W, T, f)} represents the Kth Gaussian distribution cluster in the gait mixture Gaussian model, L is the step size, H is the step height, W is the step width, T is the gait period, f is the exoskeleton-gait fit, α _K is the probability of the K-th Gaussian distribution, θ _K is the probability of the K-th Gaussian distribution Density function, K is the number of clusters of the mixed Gaussian model.

In said 2.3), for each initial gait cluster, the following formula is used to calculate the coverage area of each initial gait cluster, the formula is as follows:

where Cov denotes the coverage area of each initial gait cluster, _σi denotes the standard deviation of the i-th dimension of the gait cluster, and N denotes the dimension of the gait parameters.

If the coverage area of the current gait cluster is smaller than the minimum coverage domain, the current minimum coverage domain is used as the coverage area of the current gait cluster, and then the current gait cluster is updated; otherwise, the current gait cluster The area covered remains unchanged.

The calculation formula of the minimum coverage area is as follows:

minTh=Th×d ^t

Among them, minTh is the minimum coverage area, Th is the current minimum coverage area, d is the convergence rate, and t is the number of iterations.

The gait parameters include step length, step width, step height, and gait cycle.

In the above 3), the gait clusters are used as each arm of the multi-armed bandit algorithm, and the multi-armed bandit algorithm is used to calculate the utility degree corresponding to the gait clusters.

In the multi-armed bandit algorithm of said 3), the utility degree calculation formula of the gait clustering cluster is as follows:

Among them, P(f|D, GMM) represents the utility of the gait clustering cluster, D represents the set of gaits that have been walked, GMM represents the Gaussian mixture model of the gait, and K is the number of clusters of the Gaussian mixture model, Indicates the probability of the k-th Gaussian distribution of the t-th round, /> Indicates the gait mixture Gaussian model of round t, /> Indicates the k-th Gaussian distribution of the t-th round, Indicates the covariance matrix of the Gaussian distribution of the k-th gait in the t-th round, φ(f _i ^t , σ) represents the cumulative distribution function of the Gaussian distribution composed of the gait data of the walking experiment, and σ is the standard deviation of the noise value of the Gaussian distribution .

The beneficial effects of the present invention are:

The invention can predict the gait parameters of the lower extremity exoskeleton meeting the needs of users, has low calculation cost, fast convergence speed and high robustness.

Description of drawings

Figure 1 is a detailed flowchart of the optimization method of exoskeleton gait parameters for lower limb rehabilitation training.

Figure 2 is a schematic diagram of the step length and step width parameters of the gait.

Fig. 3 is a schematic diagram of the step height and gait cycle parameters of the gait.

Fig. 4 is an overall flowchart of an exoskeleton gait parameter optimization method for lower limb rehabilitation training.

Detailed ways

In order to make the object, technical solution and advantages of the present invention clearer, the present invention will be further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described here are only used to explain the present invention, not to limit the present invention. In addition, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not constitute a conflict with each other.

As shown in Figure 1 and Figure 4, the present invention comprises the following steps:

1) Set the initial gait of the lower limb rehabilitation exoskeleton. After the user wears the lower limb rehabilitation exoskeleton for a walking experiment, the exoskeleton-gait fit under the current gait is obtained. The subjective evaluation of is taken as exoskeleton-gait fit; exoskeleton-gait fit is a numerical type.

In this embodiment, there are 8 groups of initial experimental gaits, and the results are shown in Table 1:

Table 1 is the initial gait parameter table

serial number step size step height Step width gait cycle Evaluation value 1 1.24 0.09 0.18 3.90 31 2 1.26 0.09 0.19 2.58 39 3 1.29 0.09 0.20 1.99 75 4 1.29 0.09 0.18 1.75 74 5 1.36 0.13 0.21 1.27 80 6 1.35 0.10 0.18 1.03 76 7 1.46 0.12 0.18 0.91 60 8 1.50 0.13 0.20 0.79 41

2) According to the exoskeleton-gait fit corresponding to the gait of the current walking experiment, use the mixed Gaussian model to cluster the gait of the walking experiment in the gait library to obtain gait clusters; each group of gait The corresponding gait parameters include step length, step width, step height, and gait cycle, as shown in Figure 2 and Figure 3.

2) Specifically:

2.1) Use Bayesian information criterion to calculate the corresponding Bayesian information criterion operator before and after doubling the number of clusters of the current mixed Gaussian model, and obtain the operator difference after making a difference between the two, and adjust according to the operator difference to obtain The number of clusters in the final mixed Gaussian model;

2.1), adjust and obtain the final number of clusters of the mixed Gaussian model according to the operator difference, specifically:

The operator difference is obtained by subtracting the Bayesian Information Criterion operator after doubling from the Bayesian Information Criterion operator before doubling, that is, ΔBIC=BIC _K -BIC _2K , where BIC _K represents the Bayesian Information Criterion before doubling Operator, BIC _2K represents the doubled Bayesian Information Criterion operator. When the operator difference ΔBIC is less than 0, it means that the existing number of clusters is not enough to accurately fit the model, then the current mixed Gaussian model cluster number is doubled as the final mixed Gaussian model cluster number, otherwise The number of clusters in the current mixed Gaussian model remains unchanged, that is, remains K.

In this embodiment, the calculation formula of the corresponding Bayesian information criterion operator before and after doubling the number of clusters of the current mixed Gaussian model is as follows:

ΔBIC=BIC2-BIC4=-74.89-(-124.20)=47.31

Since ΔBIC>0, it means that the existing number of clusters is enough to accurately fit the data, and there is no need to increase the number of clusters.

2.2) Based on the number of clusters of the current mixed Gaussian model and the exoskeleton-gait fit corresponding to the gait of the current walking experiment, use the gait mixed Gaussian model to cluster the gait of the walking experiment in the gait library, and obtain initial gait clustering;

The formula for the gait mixture Gaussian model is as follows:

G={(α ₁ , θ ₁ ) _{(L, H, W, T, f)} , (α ₂ , θ ₃ ) _{(L, H, W, T, f)} ,..., (α _K , θ _K ) _{(L, H, W, T, f)} }

Among them, G represents the gait mixture Gaussian model, (α _K , θ _K ) _{(L, H, W, T, f)} represents the Kth Gaussian distribution cluster in the gait mixture Gaussian model, g={L, H , W, T, f}, g represents a set of gaits, L is the step length, H is the step height, W is the step width, T is the gait cycle, f is the exoskeleton-gait fit, α _K is the first The probability of K Gaussian distributions of gaits, θ _K represents the probability density function of the Kth Gaussian distributions, satisfying θ _K = (μ _K , ∑ _K ), μ _K is the mean value of the Kth Gaussian distributions, ∑ _K is the covariance matrix of the Kth gait Gaussian distribution. K is the number of clusters in the mixed Gaussian model. Initialize K=2, and the parameters are random.

Solve for the gait mixture Gaussian model parameters using the expectation-maximization algorithm:

α = [0.625 0.375]

2.3) Adjust the coverage area of the current gait cluster according to the minimum coverage area, so as to obtain the updated gait cluster as the final gait cluster.

For each initial gait cluster, the coverage area of each initial gait cluster was calculated using the following formula:

where Cov denotes the coverage area of each initial gait cluster, _σi denotes the standard deviation of the i-th dimension of the gait cluster, and N denotes the dimension of the gait parameters.

In the early stage of the gait parameter optimization process, due to the small number of sampling points, the model fitting is not accurate. If the coverage area of the current gait cluster is smaller than the minimum coverage area, it is likely that the unknown area cannot be explored. At this time, the coverage of the Gaussian model should be limited. If the area is the minimum value, the current minimum coverage domain is used as the coverage area of the current gait cluster, and then the current gait cluster is updated; otherwise, the coverage area of the current gait cluster remains unchanged.

The formula for calculating the minimum coverage area is as follows:

minTh=Th×d ^t

Among them, minTh is the minimum coverage area, Th is the current minimum coverage area, d is the convergence rate, and t is the number of iterations.

In this embodiment, the calculation formulas of the coverage area of the gait cluster and the minimum coverage domain are as follows:

The initial minimum coverage domain Th is 1.25e-5, and the convergence rate is 0.8.

minTh = Th × d ^t = 1e-5

Since Cov<minTh, the variance of the model is corrected, the minimum variance is log ₅ 1e-5=0.1, and the variance after correction is [0.1, 0.1, 0.1, 0.1, 46.4].

3) Calculate the utility of the gait cluster and update the utility of the corresponding gait in the gait library;

The gait clusters are used as each arm of the multi-armed bandit algorithm, and the multi-armed bandit algorithm is used to calculate the utility corresponding to the gait clusters. The utility calculation formula of gait clustering cluster is as follows:

Among them, P(f|D, GMM) represents the utility of the gait clustering cluster, D represents the set of gaits that have been walked, GMM represents the Gaussian mixture model of the gait, and K is the number of clusters of the Gaussian mixture model, Indicates the probability of the k-th Gaussian distribution of the t-th round, /> Indicates the gait mixture Gaussian model of round t, /> Indicates the k-th Gaussian distribution of the t-th round, Indicates the covariance matrix of the Gaussian distribution of the k-th gait in the t-th round, φ(f _i ^t , σ) represents the cumulative distribution function of the Gaussian distribution composed of the gait data of the walking experiment, and σ is the standard deviation of the noise value of the Gaussian distribution .

4) Continuously sample the gaits in the gait library, and eliminate the gaits that obviously cannot meet the user's satisfaction and expectations through the pruning method, thereby gradually reducing the scope of exploration and speeding up the convergence progress. According to the current utility of the gait, the gaits in the gait library are sampled to obtain a sampled gait set, and then the gait with the highest utility is selected from the gait set; the specific formula is as follows:

Among them, g ^t+1 represents the gait parameter selected in the t+1 round, argmax represents the maximum value operation, and P( ^ft |D ^t , GMM ^t ) represents the utility of the t-th round of gait clustering.

5) According to the gait with the highest current utility, the user wears the lower limb rehabilitation exoskeleton for a walking experiment to obtain the exoskeleton-gait fit under the current gait, and repeats 2)-4) until the preset rounds are reached. The gait parameters corresponding to the final gait are taken as the optimal gait parameters.

In this embodiment, the sampling gait set is as shown in Table 2:

Table 2 is the sampling gait set table

serial number step size step height Step width gait cycle gait utility 1 1.29 0.08 0.19 1.85 75.08 2 1.37 0.12 0.20 1.08 79.56 3 1.30 0.09 0.17 1.51 74.91 4 1.32 0.10 0.20 1.66 77.60 5 1.32 0.09 0.17 1.31 75.43 6 1.32 0.09 0.18 1.29 75.06 7 1.30 0.11 0.20 1.83 77.13 8 1.50 0.13 0.19 0.79 40.99

According to the calculation results, the gait of the second group is more suitable for the current user, so the step length is 1.37m, the step height is 0.12m, the step width is 0.20m, and the gait cycle is 1.08s as the gait parameters for the next round.

The above specific embodiments are only illustrative to illustrate the principles and effects of the present invention, and are not intended to limit the present invention. Anyone skilled in the art can modify or change the above-mentioned embodiments without departing from the spirit and scope of the present invention. Therefore, all equivalent modifications or changes made without departing from the spirit and technical ideas disclosed in the present invention should still be covered by the claims of the present invention.

Claims (9)

1. An exoskeleton gait parameter optimization method for lower limb rehabilitation training is characterized by comprising the following steps of:

1) Setting an initial gait of a lower limb rehabilitation exoskeleton, and acquiring exoskeleton-gait fitness under the current gait after a user wears the lower limb rehabilitation exoskeleton to perform a walking experiment;

2) According to the exoskeleton-gait fitness corresponding to the gait of the current walking experiment, using a Gaussian mixture model to cluster the gait of the walking experiment in a gait library to obtain a gait cluster;

3) Calculating the effectiveness degree of the gait cluster and updating the effectiveness degree of the corresponding gait in the gait library;

4) Sampling the gait in the gait library according to the current utilization degree of the gait to obtain a sampled gait set, and selecting the gait with the maximum utilization degree from the gait set;

5) And (2) according to the gait with the maximum current utilization, after the user wears the lower limb rehabilitation exoskeleton to carry out a walking experiment, acquiring the exoskeleton-gait fitness under the current gait, repeating the steps 2-4) until a preset round is reached, and taking the gait parameter corresponding to the final gait as the optimal gait parameter.

2. The method for optimizing exoskeleton gait parameters for rehabilitation training of lower limbs according to claim 1, wherein the 2) specifically comprises:

2.1 Calculating corresponding Bayesian information criterion operators before and after doubling the current mixed Gaussian model cluster number by adopting a Bayesian information criterion, obtaining operator differences after difference between the current mixed Gaussian model cluster number and the doubled current mixed Gaussian model cluster number, and adjusting and obtaining the final mixed Gaussian model cluster number according to the operator differences;

2.2 Based on the current mixed Gaussian model clustering number and exoskeleton-gait fitness corresponding to the gait of the current walking experiment, using a gait mixed Gaussian model to cluster the gait of the walking experiment in a gait library to obtain an initial gait cluster;

2.3 According to the minimum coverage area, the coverage area of the current gait cluster is adjusted, so that the updated gait cluster is obtained and is used as the final gait cluster.

3. The method for optimizing exoskeleton gait parameters for rehabilitation training of lower limbs according to claim 2, wherein in the step 2.1), the number of final mixed gaussian model clusters is adjusted and obtained according to operator differences, specifically:

the operator difference is obtained by subtracting the doubled Bayesian information criterion operator from the pre-doubled Bayesian information criterion operator, when the operator difference is smaller than 0, the number of clusters of the current Gaussian mixture model cluster is doubled and then used as the final Gaussian mixture model cluster number, and otherwise, the current Gaussian mixture model cluster number is unchanged.

4. The exoskeleton gait parameter optimization method for lower limb rehabilitation training according to claim 2, wherein the gait gaussian mixture model has the following formula:

G＝{(α ₁ ，θ ₁ ) _{(L，H，W，T，f)} ，(α ₂ ，θ ₃ ) _{(L，H，W，T，f)} ，…，(α _K ，θ _K ) _{(L，H，W，T，f)} }

wherein G represents a gait Gaussian mixture model (alpha) _K ，θ _K ) _{(L，H，W，T，f)} The K-th Gaussian distribution cluster in the gait Gaussian mixture model is represented, L is the step length, H is the step height, W is the step width, T is the gait cycle, f is the exoskeleton-gait fitness, and alpha _K Probability of Gaussian distribution for the Kth gait, θ _K Probability density function representing Kth gait Gaussian distribution, K being mixed Gaussian model clusterNumber of the same.

5. The method for optimizing exoskeleton gait parameters for rehabilitation training of lower limbs according to claim 2, wherein in 2.3), for each initial gait cluster, the coverage area of each initial gait cluster is calculated by the following formula:

wherein Cov represents the coverage area, σ, of each initial gait cluster _i The standard deviation of the ith dimension of the gait cluster is represented, and N represents the dimension of the gait parameter.

If the coverage area of the current gait cluster is smaller than the minimum coverage area, taking the current minimum coverage area as the coverage area of the current gait cluster, and updating the current gait cluster; otherwise, the coverage area of the current gait cluster is unchanged.

6. The exoskeleton gait parameter optimization method for lower limb rehabilitation training according to claim 5, wherein the calculation formula of the minimum coverage area is as follows:

minTh＝Th×d ^t

wherein minTh is the minimum coverage area, th is the current minimum coverage area, d is the convergence rate, and t is the iteration number.

7. The method of exoskeleton gait parameter optimization for rehabilitation training of a lower limb according to claim 1, wherein the gait parameters comprise a step size, a step width, a step height and a gait cycle.

8. The method for optimizing exoskeleton gait parameters for rehabilitation training of lower limbs according to claim 1, wherein in 3), gait clusters are used as arms of a multi-arm gambling machine algorithm, and the effectiveness degree corresponding to the gait clusters is calculated by using the multi-arm gambling machine algorithm.

9. The exoskeleton gait parameter optimization method for lower limb rehabilitation training of claim 8, wherein in the multi-arm gambling machine algorithm of 3), the utility degree calculation formula of the gait cluster is as follows:

wherein P (f|D, GMM) represents the utility degree of gait clusters, D represents the set of gait of the walking experiment, GMM represents the gait Gaussian mixture model, K is the number of Gaussian mixture model clusters,representing the probability of the kth gait gaussian distribution of the t-th wheel,a gait Gaussian mixture model representing the t-th round, < >>Indicates the kth gait Gaussian distribution of the t-th wheel,>covariance matrix representing kth gait gaussian distribution of t-th round, +.>A cumulative distribution function of a gaussian distribution composed of walking experimental gait data is shown, and σ is the standard deviation of the noise value of the gaussian distribution.