Artificial Intelligence (AI) is a branch of computer science, abbreviated as AI, which attempts to understand the essence of intelligence and produce a new intelligent machine capable of responding in a manner similar to that of human intelligence. One of the main goals of its research is to make machines capable of performing complex tasks that would normally require human intelligence to accomplish.
Artificial Intelligence can be categorized according to what a human does: machine learning (brain), computer vision (the eyes), natural language processing (ears and mouth), and robotics (limbs).
The specifics are as follows:
Computer Vision: Face recognition, fingerprint recognition, map search, image semantic understanding, target recognition, etc.
Natural Language Processing: Question and answer systems, machine translation, etc.
Knowledge Engineering: Application of knowledge graphs in personalized recommendations, Q&A systems, semantic search, and other scenarios.
Speech Recognition: AI speakers.
Mobile Robotics: SLAM (Simultaneous Localization and Mapping), path planning.
Industrial Robots: Motion planning, 3D vision.
Rehabilitation exoskeleton robots are wearable mechanical devices that apply ergonomics, bionics, and other related knowledge to the field of robotics, realizing the perfect combination of human intelligence and robot physical strength. In the field of rehabilitation and medical treatment, the uses of exoskeletons can be broadly categorized into rehabilitation training and sports replacement.
For the loss of human locomotor ability due to different injuries and diseases, they can be categorized as:
Partial Loss and Recoverable
Partial Loss Not Recoverable
Complete Loss and Recoverable
Total Loss Not Recoverable
In the case of recoverable conditions, the patient’s general wish is to regain the ability to play sports. However, depending on the degree of loss of movement, the patient’s active participation in movement will vary.
A patient with necrotic calf muscles but intact thigh muscles may need only a leg exoskeleton below the knee to assist with calf movement, whereas a patient with paraplegia in the median position will often need a whole-body exoskeleton in lieu of movement.
For the former, the control task may simply be to “predict gait” and “replace the wearer’s calf and knee joints to support body movement” while the patient can control the movement of the thighs on their own. In contrast, the control task of exoskeletons for paraplegic use is to “replicate lower limb walking in its entirety” and “maintain balance.”
However, due to cost constraints and safety considerations, the function of “maintaining balance” has been replaced by crutches and weight-reducing self-propelled vehicles in most of the rehabilitation exoskeletons currently available for patients with total lower limb paralysis.
In the case of rehabilitation, it is generally hoped that the patient can regain the ability to move under training. Therefore, the exoskeleton for rehabilitation training not only needs to be able to “reproduce the movement” but also has the ability to “recognize the degree of active participation of the patient.”
For example, hand rehabilitation training exoskeletons usually sense the wearer’s intention to perform some kind of hand movement, such as grasping, and then control the actuator to realize the grasping action. Through repeated muscle movement and nerve stimulation, the wearer’s muscle movement-neural circuits will gradually build up, achieving the restoration of hand movement ability.
Imagine a simple classic exoskeleton task: weight-bearing. The position of the exoskeleton in the model varies depending on its function.
Class A Situation: If the load is put on the exoskeleton, the exoskeleton plays the role of taking the place of the person to load the weight. The person just adjusts the posture of the exoskeleton to make it better balanced.
Class B Situation: If the load is placed on the person and the boost provided by the exoskeleton is transferred to the person, it would be the equivalent of the person pushing the object directly and the exoskeleton pushing the person.
Class C Situation: If the exoskeleton pushes an object together with the human, the human and the exoskeleton each take part of the load. The exoskeleton receives information about the load directly, and the human partially receives information about the load.
In reality, an actual exoskeleton may have more than one type. For example:
A lower-limb weight-bearing exoskeleton has a tray at the back, through which the load is transferred to the exoskeleton structure. However, the weight may be carried in a backpack, and the person will carry the shoulder straps of the backpack, which is a composite of Type A and Type C scenarios.
A walking-assisted exoskeleton, where the exoskeleton is driven by the person most of the time, but the exoskeleton joints can apply torque to the person during the support phase of the gait, reducing the load on the person’s muscles. This type of exoskeleton is a composite of Class A and Class B.
A walking-assisted exoskeleton, where the load is carried on the person’s back, applies joints that assist the person during the support phase of gait, and at the same time provides support to the load itself to reduce the burden on the person’s shoulders. This kind is a composite of Type B and Type C.
In the field of rehabilitation exoskeletons, which are dominated by class C situations (human-machine collaboration), there is still difficulty in distinguishing the difference between the operator’s force and the external force, and in distinguishing the body’s biosignals. Nowadays, the rapid development of artificial intelligence has solved many problems in the process of connecting and interacting with rehabilitation exoskeletons, including the recognition of biosignals and human-machine interaction.
AI Image Recognition: Can provide the exoskeleton with human eye-like vision to determine obstacles ahead and perform path and gait planning.
Biosignal Processing: AI can allow the exoskeleton to directly feel the intention of human movement, realizing better human-machine interaction.
Customization: AI can customize the most appropriate gait according to different people through intelligent learning, for better human-machine interaction and gait training.
In the rehabilitation process, many rehabilitation results with the subjective component of the doctor are not necessarily accurate. However, with patient data generated by machine learning, the exoskeleton robot can be scored objectively, instead of the doctor, to conduct real-time training guidance, improving the training effect.
Researchers at the University of Waterloo in Canada have updated the first open-source high-resolution wearable camera image database of human motion scenes. Based on this, AI and wearable cameras can be used to enable exoskeleton robots to walk autonomously. When exoskeleton robots are first utilized, they usually rely on manual control by the operator to switch between movement modes.
With AI and wearable cameras, the development and landing of rehabilitation exoskeletons and biologically-assisted robots will accelerate, expanding their commercial and residential markets. Rehabilitation exoskeletons inherently have the ability to influence human movement, and with the ability to sense and make decisions, these exoskeleton systems can become an interdependent intelligence.
English
