Interaction Engineering

Project lead: Prof. Dr. Robert Riener, Sensory-Motor Systems Lab, ETH Zurich, and Balgrist University Hospital, Zurich

In general, we aim to investigate the physiological and functional effect of novel technical features on individuals with specific patho-physiological characteristics. The long-term goal is to obtain optimized individual treatment programs that are characterized by a dedicated choice of modalities and technical features. We hypothesize that patient-tailored technology-aided therapy supported by a comprehensive assessment (both before and online during therapy) will result in improved therapeutic outcomes.

Currently, we check for clusters of responders to conventional & robot-assisted therapy based on data gained in a multicenter study (described here: Klamroth-Marganska, Curt, Luft, Riener et al. , Lancet Neurology 13: 159-166, 2014). Therefore, we apply hierarchical population models to represent the continuous time course of clinical scores prior to, during, and after therapy. Covariates are taken into account to improve the estimation of the individual progress the theoretical maximal effectiveness of the therapy.

In addition, we have developed interfacing technologies to accurately assess workload in robot-assisted training by using EEG, eye tracking and autonomic nervous system responses. Data fusion methods have also been applied to detect movement intentions in robot-assisted training. Based EMG, joint torques, and eye tracking, we got an excellent intention estimation accuracy and demonstrated online applicability. Together with our clinical partners, we will further exploit the possibilities to minimize patient discomfort and time investment for sensor attachment while accuracy of recognition/estimation remains reasonable.

Recently, we have also obtained valuable information on how healthy humans interact with intelligent robots online and on the benefit of individualized feedback. Compared to healthy subjects, skills and needs in patients vary even more. Thus, patients will even more benefit from individualization than healthy subjects. However, we are also aware that we cannot simple transfer our concept from healthy subjects to patients as, for instance, training behaviour and training goals are quite different. Therefore, we continue our close collaboration with clinical CRPP partners to elaborate meaningful training strategies in robot-assisted therapy.
ARMin arm rehabilitation robot