Table 2 Summary of the articles included in the review

Kyrkjebø et al., 2018 [31]

UR5

Force sensor, virtual boundary

Feasibility of cobot for rehabilitation

Simulations

The UR5 cobot is capable of functioning in both assistive and resistive modes. With the implementation of appropriate safety boundaries and control strategies, the UR5 may support rehabilitation training.

Gherman et al., 2019 [68]

ABB IRB 14000 YuMi

3D-printed custom gravity- compensation support

Application of cobot for upper limb rehabilitation

N = not specified

With an appropriate arm support and predefined trajectory, the ABB IRB 14000 YuMi is feasible fo targeted rehabilitation exercises for the upper extremity.

Fortini et al., 2019 [63]

Franka Emika Panda

Eye-tracker, sEMG sensor

To develop a robotic assistive-reaching system that uses eye- tracking technology to guide users in reaching various targets

N = 10 healthy participants

The system enables users to intuitively and accurately perform reaching tasks without prior training or physical skills.

Majidirad et al., 2018 [42]

UR5

Force sensor, gripper, customized knob, gripper, sEMG

To assess the effectiveness of cobot intervention in a task- based upper extremity rehabilitation exercise

N = 5 healthy participants

There is a correlation between muscle activation patterns in the upper extremity based on force data from the robot and the signal recorded by the sEMG sensors.

Sørensen et al., 2014 [40]

UR5

Force sensor and customized handle

To mimic real-world rehabilitation exercises

Simulations

The UR5 cobot can replicate the functionality of different training devices like ‘curl’ and ‘rope.’

Majidirad et al., 2020 [45]

UR5

Ergonomic knob, Force sensor, gripper, sEMG sensor, IMU

To develop a novel and controlled intervention procedure for upper extremity task-based rehabilitation

N = 5 healthy participants

A strong correlation exists between force exertion and muscle activities in the upper extremity during task- based exercises.

Kato et al., 2020 [62]

Franka Emika Panda

Custom elbow support, sEMG sensor, virtual reality display with haptic feedback

To investigate how target difficulty and haptic feedback influence muscle activities of the upper extremity during assisted reaching exercises

N = 1 healthy participant

Different distances to the target, target size, and haptic feedback influenced the intensity of muscle activities during reaching tasks.

Scotto di Luzio et al., 2018 [51]

Kuka LWR 4+

Motorized arm- weight support, ergonomic arm brace, sEMG sensor, M-IMU

To develop a novel 3D bio-cooperative robotic strategy for upper extremity rehabilitation that includes the patient within the control loop

N = 10 healthy participants

The system is capable of reducing muscle fatigue by providing assistance based on user’s biomechanical and physiological measurements.

Nielsen et

al., 2017 [39]

UR5

Force sensor,

wrist support,

customized

handle

To develop a novel approach for training upper extremities after a stroke using an industrial robotic arm and dynamic movement primitives (DMPs) with force feedback

Simulations

The novel approach is feasible for personalized and adaptive training of the upper extremities for individuals with different levels of impairments.

Papaleo et al., 2013 [50]

Kuka LWR 4+

3D printed custom wrist support, M-IMU sensors

To develop and validate a patient-tailored adaptive robotic system for upper extremity rehabilitation

N = 2 healthy participants

The adaptive system can safely adjust to individual needs, enhance user engagement through performance indicators, and may improve therapy outcomes through analysis of the user’s biomechanical data.

Zhang et al., 2021 [61]

Franka Emika Panda

3D-printed

customized

handle

To propose an online reference path generation method for upper extremity rehabilitation robots that considers the initial motions of the participants

N = 10 healthy

participants

This adaptive model could provide assistance as needed and effectively generate accurate reference paths for point-to-point tasks based on the initial motions of the participants.

de Azevedo Fernandes et al., 2020 [41]

UR3

3D printed tool, force sensor

To demonstrate the effectiveness of upper extremity rehabilitationusing an intelligentsystem that learns andadapts its behaviourbased on the patient’s performance during therapy sessions.

N = 1 healthy participant

The system is able to learn and dynamically adapt to varying user forces.

Miao et al., 2018 [44]

UR5 and UR10

Force sensors and customized handlebars

To propose a three- stage trajectory generation method for robot-assisted bilateral upper limb training with subject-specific adaptation

N = 7 healthy participants

The proposed method effectively provided safe and individualized bilateral upper extremity training exercises.

Sheng et al., 2019 [43]

UR5 and UR10

Force sensors and customized handlebars

To develop and evaluate a bilateral upper extremity rehabilitation system based on modern industrial robots

N = 10 healthy participants

The cobots used are feasible in providing bilateral upper extremity rehabilitation exercises.

Becker et al., 2018 [53]

Kuka LBR iiwa 14

Customize hand brace, impedance control

To implement the Assist-as-Needed (AAN) principle for adaptive upper extremity therapy by adjusting support based on position, velocity, and force.

N = 10 healthy participants

The cobot provided movement quality and consistent comfort, while smoothness improved with increased effort.

Zhang, Guo, & Sun 2020 [54]

Kuka LBR iiwa 14

Custom handles, display screens, cameras, and a body structure module

To enhance upper extremity rehabilitation by providing adaptive assistance through a virtual stiffness gradient

N = 1 healthy participant

The system demonstrated good trajectory adherence and controlled interaction forces with AAN.

Lim et al., 2023 [47]

UR robot

Gripper, custom hand brace

To facilitate in-home therapy for individuals with motor disabilities, targeting both gross and fine motor skills training

N = 5 healthy participants

The system showed significant muscle activation during robot-guided exercises compared to therapist- assisted training. Participants reported high satisfaction regarding training efficacy and system safety.

Behidj et al., 2023 [67]

Kinova MICO Gen 2

Custom- designed hand support

To develop and implement a position-based impedance control algorithm for a 6-DOF upper limb rehabilitation robot, thereby eliminating the necessity for external force sensors

N = Not specified

The system demonstrated the capability to adjust robot compliance and support dynamically, enhancing access to upper extremity rehabilitation and potentially reducing caregiver demands.

Pezeshki et al., 2023 [64]

Franka Emika Panda

Customized hand brace, display screen

To encourage patient participation and enhance the effectiveness of training sessions by minimizing robot intervention while following a predefined path

N = 4 healthy participants

The AAN strategy enhances user engagement, supports active and passive modes, and provides resistance-based challenges.

Liu et al., 2023 [66]

Agile cobot

Hand straps, custom handle

To enable stroke patients achieve accurate, real-time, and stable rehabilitation outcomes through a human–machine interaction system based on a 7-DOF cobot arm

N = Not specified

The system effectively identified force magnitude and direction and halted the arm’s movement when force exceeded a predefined threshold. This demonstrates the potential for enhancing user engagement and trajectory control in rehabilitation.

Rodrigues et al., 2023 [49]

UR cobot

Augmented reality, HoloLens headset, Unity game engine

To enhance upper extremity rehabilitation through gamification and interactive experiences

N = 31 healthy participants

The system was user-friendly, significantly enhanced participant motivation, and received high therapist satisfaction due to its customizable features.

Shi & Luo 2024 [65]

Sawyer cobot

A haptic device, sEMG sensor

To create a teleoperation framework that assists patients with upper limb hemiplegia controlling a slave robotic arm to perform specific passive and active rehabilitation training

N =Not specified

The system achieved smooth trajectory reproduction without abrupt stops, effectively generated guiding virtual forces, and highlighted the potential benefits of cobot-assisted rehabilitation.

Chiriatti et al., 2023 [46]

UR5e

Specialized handle

To present a rehabilitation framework for upper limbs of neurological patients, utilizing a collaborative robot to assist users in performing a given three-dimensional trajectory

N = 1 healthy participant

Preliminary tests with healthy participants demonstrated the system’s intuitiveness, user-friendliness, and effective safety measures.

Zhang, Guo, & Sun 2020 [55]

Kuka LBR iiwa 14

Custom-made handle

To develop and validate an assist-as-needed (AAN) controller for robot-aided rehabilitation training of the upper extremity, which can adapt to different rehabilitation modes such as passive, assistive, active, and resistive training.

N = 1 healthy participant

The AAN controller effectively facilitated task completion, promoted active engagement, enhanced motion performance, and demonstrated potential for improving motor recovery in individuals with upper extremity impairments.

Ai et al., 2023 [52]

KUKA LBR iiwa R700

Custom handle, motion capture system

To develop and validate the Uncertainty Compensated High- Order Adaptive Iteration Learning Control (UCHAILC) for improving trajectory tracking in robot- assisted upper limb rehabilitation, particularly for stroke patients.

N = Not specified

The UCHAILC method enhanced tracking performance, enabling more accurate and personalized assistance. This approach may improve rehabilitation outcomes by offering task-specific, repetitive training aligned with individual patient needs.

Tucan et al.

2021 [32]

KUKA LBR iiwa

3D printed leg constraining and sole plates

To develop a robotic system that uses a cobot equipped with a specially designed device for ankle rehabilitation

N =1 healthy participant

The cobot used is feasible for delivering ankle rehabilitation exercises.

Peterson et al.,2020 [38]

ROBERT\(\circledR\) (Kuka)

Leg brace, sEMG electrodes, FES electrode

To implement and evaluate a novel sEMG- triggered functional electrical stimulation (FES) hybrid robotic rehabilitation system for enhancing lower limb rehabilitation in stroke patients

N = 10 healthy

participants

The system achieved high success in exercise repetitions and force generation, indicating that sEMG- triggered FES combined with robotic assistance may enhance rehabilitation and functional recovery.

Leerskov et al., 2022 [59]

ROBERT\(\circledR\) (Kuka)

Leg brace, sEMG electrodes, FES electrode

To systematically investigate the extent of both potentiation and fatigue in velocity and interaction force during repetitive ROBERT\(\circledR\)-FES

exercising

N = 8 healthy participants

The results demonstrated varied responses: 50% of participants showed potentiation (increased velocity and force), while 50% experienced fatigue, emphasizing the need for adaptive rehabilitation systems.

Leerskov et al., 2024 [58]

ROBERT\(\circledR\) (Kuka)

FES electrodes, EMG electrodes, foot brace

To evaluate the technical performance and clinical feasibility of integrating the ROBERT\(\circledR\) with FES within an AAN strategy for the rehabilitation of lower extremity function in stroke patients

N = 10 healthy participants and 2 individuals with stroke

The system achieved over 96% accuracy in behavior detection and assistance modulation, enhanced engagement, voluntary effort, and motor learning in post-stroke participants.

Rikhof et al., 2024 [57]

ROBERT\(\circledR\) (Kuka)

FES electrodes, EMG electrodes, foot brace

To assess the feasibility of combining robotics and FES with an AAN approach to support actively-initiated leg movements in (sub-) acute stroke patients

N = 9 individuals with subacute stroke

Assistance was needed in 44% of ankle dorsiflexion and 5% of knee extension repetitions, with mild to moderate fatigue, demonstrating the feasibility of integrating robotics and FES for stroke leg rehabilitation.

Sórensen et al., 2024 [56]

ROBERT\(\circledR\) (Kuka)

Leg brace

To determine the feasibility of conducting a large trial designed to determine whether the ROBERT\(\circledR\) can be used to increase strength in the hip flexor muscles after SCI.

N = 4 individuals ith SCI

The study reported a 92% training adherence, no adverse events, and positive feedback, suggesting ROBERT\(\circledR\) is acceptable and potentially effective for enhancing hip flexor strength.

Wolański et al., 2023 [48]

UR10

Leg brace, Noraxon My- oMotion system, IMU sesnors

To evaluate the feasibility of adapting the UR10e industrial robot to assist in the rehabilitation process, specifically focusing on its ability to perform Proprioceptive Neuromuscular Facilitation (PNF) movements and comparing its performance to that of a physiotherapist

N = Not specified

The cobot executed repetitive exercises, with movement accuracy dependent on trajectory programming, and demonstrated potential efficiency advantages in cobot- assisted rehabilitation.

Chan 2022 [60]

ROBERT\(\circledR\) (Kuka)

Not specified

To evaluate the cost- effectiveness and clinical benefits of the ROBERT\(\circledR\) robotic rehabilitation device for lower limb rehabilitation for stroke patients

N = Not specified

ROBERT\(\circledR\) may reduce hospital stays and readmissions, incur only one-tenth the operating costs of outpatient sessions, and provides a cost- effective rehabilitation solution in clinical settings.