This contribution is concerned with joint angle calculation based on inertial measurement data in the context of human motion analysis. Unlike most robotic devices, the human body lacks even surfaces and right angles. Therefore, we focus on methods that avoid assuming certain orientations in which the sensors are mounted with respect to the body segments. After a review of available methods that may cope with this challenge, we present a set of new methods for: (1) joint axis and position identification; and (2) flexion/extension joint angle measurement. In particular, we propose methods that use only gyroscopes and accelerometers and, therefore, do not rely on a homogeneous magnetic field. We provide results from gait trials of a transfemoral amputee in which we compare the inertial measurement unit (IMU)-based methods to an optical 3D motion capture system. Unlike most authors, we place the optical markers on anatomical landmarks instead of attaching them to the IMUs. Root mean square errors of the knee flexion/extension angles are found to be less than 1° on the prosthesis and about 3° on the human leg. For the plantar/dorsiflexion of the ankle, both deviations are about 1°.

The orientation of a magneto and inertial measurement unit (MIMU) is estimated by means of sensor fusion algorithms (SFAs) thus enabling human motion tracking. However, despite several SFAs implementations proposed over the last decades, there is still a lack of consensus about the best performing SFAs and their accuracy. As suggested by recent literature, the filter parameters play a central role in determining the orientation errors. The aim of this work is to analyze the accuracy of ten SFAs while running under the best possible conditions (i.e., their parameter values are set using the orientation reference) in nine experimental scenarios including three rotation rates and three commercial products. The main finding is that parameter values must be specific for each SFA according to the experimental scenario to avoid errors comparable to those obtained when the default parameter values are used. Overall, when optimally tuned, no statistically significant differences are observed among the different SFAs in all tested experimental scenarios and the absolute errors are included between 3.8 deg and 7.1 deg. Increasing the rotation rate generally leads to a significant performance worsening. Errors are also influenced by the MIMU commercial model. SFA MATLAB implementations have been made available online.

A growing number of researchers consider iterative learning control (ILC) a promising tool for numerous control problems in biomedical application systems. We will briefly discuss why classical ILC theory is technically too restrictive for some of these applications. Subsequently, we will extend the classical ILC design in the lifted systems framework to the class of repetitive trajectory tracking tasks with variable pass length. We will analyse the closed-loop dynamics for two standard learning laws, and we will discuss in which sense the tracking error can be reduced by which controller design strategies. Necessary and sufficient conditions for monotonic convergence will be derived. We then summarise all results in a set of practical controller design guidelines. Finally, a simulation study is presented, which demonstrates the usefulness of these guidelines and illustrates the special dynamics that occur in variable pass length learning.

Many stroke patients suffer from the drop foot syndrome, which is characterized by a limited ability to lift (the lateral and/or medial edge of) the foot and leads to a pathological gait. In this contribution, we consider the treatment of this syndrome via functional electrical stimulation (FES) of the peroneal nerve during the swing phase of the paretic foot. A novel three-electrodes setup allows us to manipulate the recruitment of m. tibialis anterior and m. fibularis longus via two independent FES channels without violating the zero-net-current requirement of FES. We characterize the domain of admissible stimulation intensities that results from the nonlinearities in patients' stimulation intensity tolerance. To compensate most of the cross-couplings between the FES intensities and the foot motion, we apply a nonlinear controller output mapping. Gait phase transitions as well as foot pitch and roll angles are assessed in realtime by means of an Inertial Measurement Unit (IMU). A decentralized Iterative Learning Control (ILC) scheme is used to adjust the stimulation to the current needs of the individual patient. We evaluate the effectiveness of this approach in experimental trials with drop foot patients walking on a treadmill and on level ground. Starting from conventional stimulation parameters, the controller automatically determines individual stimulation parameters and thus achieves physiological foot pitch and roll angle trajectories within at most two strides.

Inertial-sensor-based attitude estimation is a crucial technology in various applications, from human motion tracking to autonomous aerial and ground vehicles. Application scenarios differ in characteristics of the performed motion, presence of disturbances, and environmental conditions. Since state-of-the-art attitude estimators do not generalize well over these characteristics, their parameters must be tuned for the individual motion characteristics and circumstances. We propose RIANN, a ready-to-use, neural network-based, parameter-free, real-time-capable inertial attitude estimator, which generalizes well across different motion dynamics, environments, and sampling rates, without the need for application-specific adaptations. We gather six publicly available datasets of which we exploit two datasets for the method development and the training, and we use four datasets for evaluation of the trained estimator in three different test scenarios with varying practical relevance. Results show that RIANN outperforms state-of-the-art attitude estimation filters in the sense that it generalizes much better across a variety of motions and conditions in different applications, with different sensor hardware and different sampling frequencies. This is true even if the filters are tuned on each individual test dataset, whereas RIANN was trained on completely separate data and has never seen any of these test datasets. RIANN can be applied directly without adaptations or training and is therefore expected to enable plug-and-play solutions in numerous applications, especially when accuracy is crucial but no ground-truth data is available for tuning or when motion and disturbance characteristics are uncertain. We made RIANN publicly available.