Energy expenditure is a key rehabilitation outcome and is starting to be used in robotics-based rehabilitation through human-in-the-loop control to tailor robot assistance towards reducing patients’ energy effort. However, it is usually assessed by indirect calorimetry which entails a certain degree of invasiveness and provides delayed data, which is not suitable for controlling robotic devices. This work proposes a deep learning-based tool for steady-state energy expenditure estimation based on more ergonomic sensors than indirect calorimetry. The study innovates by estimating the energy expenditure in assisted and non-assisted conditions and in slow gait speeds similarly to impaired subjects. This work explores and benchmarks the long short-term memory (LSTM) and convolutional neural network (CNN) as deep learning regressors. As inputs, we fused inertial data, electromyography, and heart rate signals measured by on-body sensors from eight healthy volunteers walking with and without assistance from an ankle-foot exoskeleton at 0.22, 0.33, and 0.44 m/s. LSTM and CNN were compared against indirect calorimetry using a leave-one-subject-out cross-validation technique. Results showed the suitability of this tool, especially CNN, that demonstrated root-mean-squared errors of 0.36 W/kg and high correlation (ρ > 0.85) between target and estimation (${\overline{\mathrm{R}}}^2$ = 0.79). CNN was able to discriminate the energy expenditure between assisted and non-assisted gait, basal, and walking energy expenditure, throughout three slow gait speeds. CNN regressor driven by kinematic and physiological data was shown to be a more ergonomic technique for estimating the energy expenditure, contributing to the clinical assessment in slow and robotic-assisted gait and future research concerning human-in-the-loop control.

On this paper a vision-based contact- and markerless method for gait evaluation is proposed, and validated in different experimental setups against commercial motion capture systems (Vicon) and inertial gait analysis tools (GaitShoes). While the development goal is its integration on the ASBGo Smart Walker platform, only an inexpensive depth camera is required. It is shown to have reasonable results when computing gait metrics in real time, in different experimental setups, from different walker types, vision hardware and walking scenarios. Performance is evaluated through RMSD values for several gait metrics. Results illustrate that the proposed approach can be a valuable non-invasive, contactless and low cost alternative to gait analysis systems used in clinical rehabilitation environments.

The use of robots in rehabilitation attempts an effective, compliant, and time-efficient gait recovery while adapting the assistance to the user’s needs. Assist-as-needed strategies (AAN), such as adaptive impedance control, have been reported as prominent strategies to enable this recovery effects. This study proposes an interaction-based assist-as-needed impedance control strategy for an ankle robotic orthosis that adapts the robotic assistance by changing the Human-Robot interaction stiffness. The adaptability of the interaction stiffness allows the real-time passage from passive assistance to an active one, approaching AAN gait training. The interaction stiffness was successfully estimated by linear regression of the Human-Robot interaction torque vs angle trajectory curve. From the validation with seven able-bodied subjects, we verified the suitability of this adaptive impedance control for a more compliant, natural, and comfortable motion than the trajectory tracking control. Moreover, the proposed strategy considers the users’ motion intention and encourages them to interact closely with the robotic device while guiding their ankle trajectory according to desired trajectories. These achievements contribute to AAN gait training.

Gait disabilities affect the human quality of life. Current directions for time-effective robotic gait rehabilitation require the inclusion of biofeedback systems (BSs) as a complementary robotic tool for efficient motor relearning. This work aims to present the user-centered design and validation of a wearable BS to foster users’ active participation and enable therapists’ effective participation during robotic gait rehabilitation driven by active orthoses. The multimodal BS comprises a development board to manage the activation of the stimuli (vibrotactile through the vibrotactile waist and shank bands, sonorous via single earphone, and visual using RGB LED) according to data tracked by orthosis embedded sensors. The BS’s versatility allows its functioning as a modular and stand-alone system or integrated into the orthotic system. The system’s operability was validated with four healthy subjects walking on a treadmill with the orthotic system and BS at 1 km/h. The results showed an operable system with good usability during robotic gait rehabilitation. This wearable BS has the potential to boost symmetric gait recovery and to effectively augment the user’s active participation during robotic gait therapy; thus, contributing to accelerating the user’s motor recovery.

The movements of sit-to-stand and stand-to-sit are frequently executed on daily life. To develop robotic assistive devices for people with mobility problems, it is important to study how a healthy human performs these tasks. The goal of this study is to present a mathematical model based on acquired kinematic and kinetic data that represents a healthy human body performing these movements. The results revealed that the movements of sit-to-stand and stand-to-sit have symmetric ground reaction force, joint angles and torques. The joint angles and the torques are very similar between each leg. The knee and the hip show more variation of the angle and achieve higher values of torque in comparison to the ankle. Although, ankle has higher torque during standing position. The ground reaction force shows that it is necessary to create an additional force to the weight force to achieve the final position of the movement. The acquired data describe as expected the sit-to-stand and stand-to-sit movements and can be used, in the future, to validate the presented model.