Individualized reference gait patterns for lower limb rehabilitation robots can greatly improve the effectiveness of rehabilitation. However, previous methods can only generate customized gait patterns at several fixed discrete walking speeds and generating gaits at continuously varying speeds and stride lengths remains unsolved. This work proposes an individualized gait pattern generation method based on a recurrent neural network (RNN), which is proficient in series modeling. We collected the largest gait data set of this kind, which consists of 4,425 gait patterns from 137 subjects. Using this data set, we trained an RNN to create a function mapping from body parameters and gait parameters to a gait pattern. The experimental results indicate that our model is able to generate gait patterns at continuously varying walking speeds and stride lengths while also reducing the errors in the ankle, knee, and hip measurements by 12.83%, 20.95%, and 28.25%, respectively, compared to previous state-of-the-art method.

No abstract

Lower limb assistive robots have a wide range of applications in medical rehabilitation, hiking, and the military. The purpose of this work is to investigate the efficiency of wearable assistive devices under different weight-bearing walking conditions. We designed an experimental platform, with a lightweight ankle-assisted robot weighing 5.2 kg and carried mainly on the back. Eight subjects were tested in three experimental conditions: free walk with load (FWL), power-off with load (POFL), and power-on with load (PONF) for different levels of force at a walking speed of 3.6 km/h. We recorded the metabolic expenditure and kinematics of the subjects under three levels of weight-bearing (equal to 10%, 20%, and 30% of body mass). The critical forces from the fit of the assistive force and metabolic depletion curves were 130 N, 160 N and 215 N at three different load levels. The intrinsic weight of our device increases mechanical work at the ankle as the load weight rises, with 2.08 J, 2.43 J, 2.73 J for one leg during a gait cycle. The ratio of the mechanical work input by the robot to the mechanical work output by the weight of the device decreases (0.904, 0.717, and 0.513 with different load carriages), verifying that the walking assistance efficiency of such devices decreases as the weight rises. In terms of mechanical work in the ankle joint, our results confirm that the efficiency of the ankle-assisted walking robot decreases as weight bearing increases, which provides important guidance for the lightweight design of portable weight-bearing walking robots.

Background: Obtaining appropriate assistance timings for individual users of active lower limb assistant robots (ALLARs) is one of the major challenges that limit the practical application of robots since very small assistance timing errors greatly affect the robot's assistance effect. However, neither theoretical nor experimental methods can currently generate appropriate assistance timings due to their respective availability or accuracy limitations. Method: In this paper, we proposed a new method to generate appropriate assistance timings for individual users of ALLARs via machine learning. The method has the accuracy of theoretical methods and the availability of experimental methods. We established a database of ten static physiological parameters, three dynamic parameters, and theoretical appropriate assistance timings, and mapped the static physiological parameters and the dynamic parameters to the theoretical assistance timings using multilayer neuron networks. Fold-cross validation and determination efficient were used to test the fit of the model. The root mean square error between generated values and true values of each subject was compared to that between the mean of the sample and the true values of each subject to evaluate the data accuracy of our method. We also set ±2% error as the boundary of the practical accuracy and compared the practical accuracy when using our method to that when using the mean generally. Result: The model achieved a small standard deviation of the square root error in the 10-fold cross-validation experiment and a large determination coefficient. We reduced the data error of starting and ending assistance timing from 0.0265 and 0.0172 to 0.014±0.000429 and 0.0079±0.000875, respectively, and improved the practical accuracy of starting and ending assistance timing from 54.93% and 75.49% to 89.54% and 99.95%, respectively.Conclusion: The proposed method can generate an appropriate assistance timings for different users of ALLARs walking at different speeds. Moreover, a new reference for ending assistance timings is provided and the database can be used as a reference for futer research. The practical effect of the method will be tested in future work.