AI-assisted robotic arm gently supporting a human arm, symbolizing recovery and empowerment.

Unlock Effortless Movement: How AI and Robotics are Redefining Assistive Technology

Mira Elwood in Tech & Innovation September 2025 • 5 min read.

"Explore how cutting-edge AI-powered robotic systems are transforming rehabilitation and daily living for individuals with motor impairments."

For decades, understanding how the central nervous system (CNS) controls our arm movements has been a complex challenge. The Bernstein problem highlights this difficulty, questioning how the CNS coordinates a motor system with so many degrees of freedom (DOFs) to efficiently interact with our environment. This challenge is often framed as an inverse kinematics (IK) problem, where we need to determine the joint angles of a kinematic chain to achieve a desired end-effector position.

The problem with IK is that it often presents infinite solutions, especially when dealing with a redundant kinematic chain like the human arm. Researchers have proposed numerous models and methods to explain how the CNS resolves this redundancy and to provide practical solutions for IK. These methods range from using the Jacobian matrix in numerical iterative techniques to virtual spring-damper systems and probabilistic approaches.

One significant application of these models is in the development of assistive robots, designed to aid individuals in daily life, particularly those recovering from motor function loss through rehabilitation. The eNHANCE project, for example, aims to create an interface that combines various inputs—such as user intention, electromyography (EMG), and mechanomyography (MMG) signals—to provide support via an assistive upper limb robot. This robot assists patients who have partially lost the ability to move their upper right limb, supporting only translation and rotation of the forearm.

How Does the AI Model Predict Arm Movements?

AI-assisted robotic arm gently supporting a human arm, symbolizing recovery and empowerment.

To address these challenges, researchers have developed a novel approach using Bayesian networks to model human arm movement in an anthropomorphic manner. This model aims to control an upper limb assistive robot by receiving a desired wrist position as input and generating three key angles as output: the swivel angle (representing the rotation of the plane formed by the upper and lower arm), and two angles corresponding to the sternoclavicular joint's degrees of freedom (elevation/depression and protraction/retraction).

These angles, combined with the wrist position, fully describe the position of the shoulder and elbow. To train the model, data was collected from recording sessions of human motion during four different activities of daily living. The model’s performance was then evaluated based on the end-point errors and Pearson's r values for the elbow and shoulder joints.

Data Collection: Motion data was gathered from healthy subjects using reflective markers placed on key joints.
Task Performance: Subjects performed four tasks selected for their relevance to upper limb rehabilitation and everyday usefulness.
Model Construction: A kinematic model of the upper limb was built, incorporating Denavit-Hartenberg (DH) parameters.
Trajectory Determination: The arm trajectory was determined based on the wrist trajectory.
Performance Evaluation: The model's performance was evaluated using metrics such as mean error and Pearson's r correlation coefficient.

The model demonstrated high accuracy in predicting elbow movement (mean error 0.021 ± 0.020 m, Pearson's r 0.86-0.99) and shoulder movement (mean error 0.014 ± 0.011 m, Pearson's r 0.52-0.99) for wrist trajectories within the training dataset. It also showed promise in generating new motions outside the training set, with improved elbow joint accuracy (mean error 0.042 ± 0.025 m, Pearson's r 0.59-0.99) and average shoulder joint accuracy (mean error 0.026 ± 0.012 m, Pearson's r -0.12–0.99).

The Future of Assistive Movement

This novel approach to solving the inverse kinematics problem for the human upper limb not only offers a method for creating movement outside of training data but also opens new avenues for personalized assistive technology. By continually refining these models and integrating them into real-world robotic systems, we can significantly improve the quality of life for individuals with motor impairments, empowering them to regain independence and control over their movements.

About this Article -

This article was crafted using a human-AI hybrid and collaborative approach. AI assisted our team with initial drafting, research insights, identifying key questions, and image generation. Our human editors guided topic selection, defined the angle, structured the content, ensured factual accuracy and relevance, refined the tone, and conducted thorough editing to deliver helpful, high-quality information.See our About page for more information.

This article is based on research published under:

DOI-LINK: 10.1109/biorob.2018.8487224, Alternate LINK

Title: An Upper Limb Kinematic Graphical Model For The Prediction Of Anthropomorphic Arm Trajectories

Journal: 2018 7th IEEE International Conference on Biomedical Robotics and Biomechatronics (Biorob)

Publisher: IEEE

Authors: B. A. P. S. Noronha, M. Wessels, A. Q. L. Keemink, A. Bergsma, H. F. J. M. Koopman

Published: 2018-08-01

Everything You Need To Know

What is the Bernstein problem and why is it relevant to understanding arm movement?

The Bernstein problem is a fundamental challenge in understanding how the central nervous system (CNS) controls arm movements. It highlights the difficulty in coordinating a motor system with many degrees of freedom (DOFs) to interact efficiently with the environment. The CNS needs to determine the joint angles of a kinematic chain to achieve a desired end-effector position. This is often framed as an inverse kinematics (IK) problem. The complexity arises because IK frequently has infinite solutions, particularly with redundant kinematic chains like the human arm. Resolving this redundancy is crucial for developing effective assistive technology and rehabilitation strategies.

How does the eNHANCE project use AI and robotics to assist individuals with motor impairments?

The eNHANCE project uses an assistive upper limb robot to help patients who have partially lost the ability to move their upper right limb. This robot integrates various inputs such as user intention, electromyography (EMG), and mechanomyography (MMG) signals to provide support. The robot focuses on assisting with the translation and rotation of the forearm, offering a practical application of AI-driven robotic systems in daily life. The goal is to enable individuals to regain mobility and independence by providing support for specific movements.

What are the key components of the AI model used to predict arm movements, and how is it trained?

The AI model utilizes Bayesian networks to predict human arm movements. It takes a desired wrist position as input and generates three key angles: the swivel angle and two angles related to the sternoclavicular joint's degrees of freedom. These angles, along with the wrist position, fully describe the position of the shoulder and elbow. The model is trained using data from recording sessions of human motion during four activities of daily living. The process involves data collection from healthy subjects using reflective markers, performing tasks relevant to rehabilitation, model construction using Denavit-Hartenberg (DH) parameters, trajectory determination, and performance evaluation.

How is the AI model's performance evaluated, and what were the key findings in terms of accuracy?

The model's performance is evaluated based on end-point errors and Pearson's r values for the elbow and shoulder joints. The model demonstrated high accuracy in predicting elbow movement, with a mean error of 0.021 ± 0.020 m and Pearson's r values between 0.86-0.99, and shoulder movement, with a mean error of 0.014 ± 0.011 m and Pearson's r values between 0.52-0.99, for wrist trajectories within the training dataset. It also showed promise in generating new motions outside the training set, with improved elbow joint accuracy and average shoulder joint accuracy. These results suggest the model's potential for creating movement outside of the training data.

What is the significance of using AI and robotics for solving the inverse kinematics problem in assistive technology, and what is the future outlook?

The application of AI and robotics to solve the inverse kinematics (IK) problem for the human upper limb opens new avenues for personalized assistive technology. By continually refining these models and integrating them into real-world robotic systems, the quality of life for individuals with motor impairments can be significantly improved. This approach not only helps in creating movement outside of training data but also empowers individuals to regain independence and control over their movements. The future involves refining these models and integrating them into real-world robotic systems to maximize the benefits of personalized assistive technology.