The development of exoskeletons by researchers in order to help all those who have lost their motor skills is increasingly notable. Exoskeletons are basically robots that adapt to the user’s legs to allow them to walk, or to the arms to gain more strength, for example.
Despite the usefulness they have provided so far, current exoskeletons have the disadvantage of relying on manual controls to change position in situations where the person requires to move from one motor state to another (from sitting to standing, or walking, or climbing stairs, etc.).
And it is that resorting to the use of joysticks or special applications installed on a mobile device to direct the movement of the exoskeleton can be uncomfortable and exhausting for the user.
That is why a team of scientists is working on the development of exoskeletons that can act autonomously to effect the change in movement mode. To do this, they have proposed incorporating portable cameras into the structure of the exoskeletons, so that they serve as a means to capture the environment and provide the machines with vision data with which they can act autonomously.
In this sense, the artificial intelligence (AI) software may have the ability to carry out data analysis, so that it can easily recognize structures such as stairs, doors and other elements that may be present in the environment where transit the user.
All of these efforts are part of the ExoNet project, led by University of Waterloo robotics researcher Brokoslaw Laschowski.
It should be noted that ExoNet is the first open source database of high-resolution images of portable cameras applied in scenarios of human locomotion, hosting an estimated of more than 5.6 million images corresponding to walking situations carried out in both environments exteriors and interiors.
This is how the team took advantage of this data to train deep learning algorithms, resulting in a technology capable of recognizing gait environments with greater than 70% accuracy, all this autonomously.
Despite these results, Laschowski and his team will continue to work hard to increase the accuracy of this analysis software until it is as efficient as possible in making real-time onboard operations of robotic exoskeletons successful.