The physical act of walking may seem simple, but each step involves intricate processes that researchers are just starting to understand fully.
A recent study by Madonna’s Research Institute used an artificial neural network (ANN) to predict how much energy people use while walking and compared those findings to the traditional Multiple Linear Regression (MRL) model.
“Artificial Neural Networks are a type of artificial intelligence designed to recognize patterns and make predictions based on complex data,” Arash Gonabadi, MS, Ph.D., assistant research director at Madonna’s Research Institute and the co-author of the study, said. “Unlike traditional methods, which often assume relationships are linear, ANNs can handle the nonlinear reality of how we move.”
Gonabadi’s team found that ANNs significantly outperformed MLR in predicting metabolic costs—or how much energy a person uses while walking.
“The ANN model identified patterns that reveal how the body maintains stability and adjusts to disruptions during walking,” Gonabadi said. “For instance, one metric highlighted how sensitive a person’s walking is to small disturbances, like a slight trip or wobble. Another showed how steps are coordinated over time, which directly influences walking efficiency. These insights would be challenging to uncover with traditional methods.”
Indeed, these new insights into gait analysis have practical implications for patients recovering from a stroke, a brain injury or other serious medical conditions.
“Our findings can directly inform the development of smarter, more adaptive assistive devices,” Gonabadi said. “By understanding which aspects of walking require the most energy, we can design exoskeletons that provide targeted support during those moments. This could reduce fatigue, improve stability, and make walking more efficient and comfortable for users.”
This latest study builds on Madonna’s earlier research which also used ANNs to predict how much energy we use based on two things: ground reaction forces (the forces we create when our feet hit the ground) and joint moments (how our joints move). Gonabadi envisions a bright future for AI in rehabilitation care and biomechanics.
“AI has the potential to revolutionize gait analysis, rehabilitation, and the design of assistive technologies,” Gonabadi said. “As AI models become more refined, they will be able to provide even more personalized insights into human movement, leading to the development of highly adaptive devices that respond to an individual’s specific needs and improve overall mobility.”