A model-free deep reinforcement learning approach for control of exoskeleton gait patterns

Rose, Lowell; Bazzocchi, Michael C. F.; Nejat, Goldie

doi:10.1017/s0263574721001600

Cited by 13 publications

(10 citation statements)

References 85 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…These trajectories are obtained from healthy subjects as well as from the literature. The drawbacks of this technique are that it does not consider the patient's disability level, and thus it can create discomfort by preventing free motion for the patient [23,87].…”

Section: Control Architecturementioning

confidence: 99%

See 1 more Smart Citation

State of the Art in Wearable Wrist Exoskeletons Part I: Background Needs and Design Requirements

et al. 2023

View full text Add to dashboard Cite

Despite an increase in the use of exoskeletons, particularly for medical and occupational applications, few studies have focused on the wrist, even though it is the fourth most common site of musculoskeletal pain in the upper limb. The first part of this paper will present the key challenges to be addressed to implement wrist exoskeletons as wearable devices for novel rehabilitation practices and tools in the occupational/industrial sector. Since the wrist is one of the most complex joints in the body, an understanding of the bio-mechanics and musculo-skeletal disorders of the wrist is essential to extracting design requirements. Depending on the application, each wrist exoskeleton has certain specific design requirements. These requirements have been categorized into six sections: purpose, kinematics, dynamics, rigidity, ergonomics, and safety. These form the driving factors behind the choice of a design depending on the objectives. Different design architectures are explored, forming the basis for the various technical challenges that relate to: actuation type, power source, power transmission, sensing, and control architecture. This paper summarizes, in a systematic approach, all the current technologies adopted, analyzes their benefits and limitations, and finally proposes future perspectives.

show abstract

Section: Control Architecturementioning

confidence: 99%

“…To overcome the above mentioned limitations, assist-as-needed (AAN) control can be implemented to provide customized assistance [23,87]. By continuously reading sensors signals, AAN can estimate the patient's intent and modify accordingly the device assistance to enhance the patient's participation.…”

Section: Control Architecturementioning

confidence: 99%

State of the Art in Wearable Wrist Exoskeletons Part I: Background Needs and Design Requirements

et al. 2023

View full text Add to dashboard Cite

show abstract

“…The control system is based on state-of-the-art controllers that utilize machine learning strategies to predict hand/wrist movements according to muscle activation. It uses a PID controller to generate the required pulling forces on the cables [16,[79][80][81][82]. The control scheme is structured on two levels, as depicted in Figure 35, to guarantee that the user can move freely as intended and experience assistive forces with appropriate timing and intensity.…”

Section: Design Proposal For a Portable Wrist Exoskeletonmentioning

confidence: 99%

State of the Art in Wearable Wrist Exoskeletons Part II: A Review of Commercial and Research Devices

Pitzalis,

Park,

Caldwell

et al. 2023

Machines

View full text Add to dashboard Cite

Manual handling tasks, both in daily activities and at work, require high dexterity and the ability to move objects of different shapes and sizes. However, musculoskeletal disorders that can arise due to aging, disabilities, overloading, or strenuous work can impact the natural capabilities of the hand with serious repercussions both in working and daily activities. To address this, researchers have been developing and proving the benefits of wrist exoskeletons. This paper, which is Part II of a study on wrist exoskeletons, presents and summarizes wearable wrist exoskeleton devices intended for use in rehabilitation, assistance, and occupational fields. Exoskeletons considered within the study are those available either in a prototyping phase or on the market. These devices can support the human wrist by relieving pain or mitigating fatigue while allowing for at least one movement. Most of them have been designed to be active (80%) for higher force/torque transmission, and soft for better kinematic compliance, ergonomics, and safety (13 devices out of 24, more than 50%). Electric motors and cable transmission (respectively 11 and 9 devices, out of 24, i.e., almost 50% and 40%) are the most common due to their simplicity, controllability, safety, power-to-weight ratio, and the possibility of remote actuation. As sensing technologies, position and force sensors are widely used in all devices (almost 90%). The control strategy depends mainly on the application domain: for rehabilitation, CPM (control passive motion) is preferred (35% of devices), while for assistance and occupational purposes, AAN (assistance-as-needed) is more suitable (38% of the devices). What emerges from this analysis is that, while rehabilitation and training are fields in which exoskeletons have grown more easily and gained some user acceptance (almost 18 devices, of which 4 are available on the market), relatively few devices have been designed for occupational purposes (5, with only 2 available on the market) due to difficulties in meeting the acceptance and needs of users. In this perspective, as a result of the state-of-the-art analysis, the authors propose a conceptual idea for a portable soft wrist exoskeleton for occupational assistance.

show abstract

“…The neural network [19] was also used in designing rehabilitation robots by approximating unknown values of a system. The deep reinforcement learning approach was also applied to perform rehabilitation using lower limb exoskeleton [20]. The musculoskeletal model was developed and 3D simulations were carried out in physics environment on both the healthy gait data and post stroke patient gait data.…”

Section: Related Workmentioning

confidence: 99%

Adaptive Neural & Fuzzy Controller for Exoskeleton Gait Pattern Control Based on Musculoskeletal Modeling

Gupta

Semwal

2022

Preprint

View full text Add to dashboard Cite

The paper deals with the development of 2-link lower limb extremity manipulator for rehabilitation purposes in bio-medical engineering. The musculoskeletal modeling was performed on the collected human gait data to obtain the joint angles and driving force values for hip and knee movement. The results obtained from the proposed controller were verified using Matlab-Simulink. The robust adaptive RBF neural network and fuzzy sliding mode controller were developed to track the desired trajectory. The OpenSim software was utilized to perform inverse kinematics and inverse dynamics on the musculoskeletal model which is scaled according to the physical dimensions of the experimental subject. The simulation results stated that SMC-RBF network has better tracking rate, less chattering problems and handle unknown uncertainties well. Finally, the results obtained were compared with the existing method with the help of performance index parameters to prove their effectiveness.

show abstract

A model-free deep reinforcement learning approach for control of exoskeleton gait patterns

Cited by 13 publications

References 85 publications

State of the Art in Wearable Wrist Exoskeletons Part I: Background Needs and Design Requirements

State of the Art in Wearable Wrist Exoskeletons Part I: Background Needs and Design Requirements

State of the Art in Wearable Wrist Exoskeletons Part II: A Review of Commercial and Research Devices

Adaptive Neural & Fuzzy Controller for Exoskeleton Gait Pattern Control Based on Musculoskeletal Modeling

Contact Info

Product

Resources

About