Research in Wearable and Mobility Assistance Robots
Our research in human-centered robotics addresses wearable robots and mobility assistance devices in addition to Research in humanoid robotics.
Assistive devices and wearable robots can improve mobility in persons with particular needs, such as spinal cord injury patients or older adults, and have a high potential impact in society. In addition, they can play an important role in injury prevention. Here the interaction with the human is generally very close, often with permanent large surface or multi-point contact between humans and robots posing particular challenges on robot design and control. Specifically, we investigate:
Lower limb exoskeletons for rehabilitation and motion assistance – Our research covers medical lower limb exoskeletons for different impairments, ranging from single joint exoskeletons to those covering the entire lower body. We aim to develop a profound understanding of the effect of exoskeletons on human movement and the human-exoskeleton co-adaption process. In that context, we have evaluated different measures and measurement devices for adaptation and familiarization, and we have developed procedures to support the familiarization process with a new exoskeleton. Using model-based design optimization and innovative mechanical design ideas formulated as parameterized exoskeleton models in combination with subject-specific human models and particular attention to interface modeling, our goal is to develop personalized exoskeleton technology that supports individual needs. We also use optimization and learning approaches to develop better exoskeleton controllers, particularly improving stability control in the absence of crutches.
Spinal exoskeletons for lower back pain prevention – We are also interested in developing spinal exoskeletons for the prevention of back pain and vocational reintegration after injuries. Such exoskeletons are useful for all professions that require handling heavy loads or remaining in flexed positions for extended periods, such as logistics workers, mechanics, gardeners, nurses, or dentists. Our research in this area includes the design and component optimization and personalization of passive and active spinal exoskeletons. For this, subject-specific human models in combination with parameterized exoskeleton models including realistic interface models are crucial. Our computational studies also include the comparison of exoskeleton support vs the ergonomic improvement of lifting techniques and the identification of behavior models of persons with and without back pain. Currently, we are performing detailed biomechanical evaluation studies of human-exoskeleton co-adaptation which will allow us to learn better models and include model-based optimization as a reliable component in the design cycle of personalized exoskeletons. In addition, we are including optimization-based model predictive control in exoskeleton controllers.
Mobility assistance devices for the elderly generation – We are developing intelligent assistive devices for the elderly (robotic rollators and other actively engaging devices) to maintain and enhance their mobility – a key indicator for quality of life. Such devices should be able to perceive the environment, understand the situation, understand the intention of the user, determine the best possible action, and be versatile and maneuverable. They should operate in various terrains and environments, both outdoor and indoor. We perform detailed experimental and computational studies to improve the design and control of the motions of the devices and their interactions with the user in order to optimize support. Our particular interest is in active sit-to-stand and stand-to-sit assistance since these motions pose a particular challenge for frail older adults and have a high risk of falls. For this purpose, we have developed a robust lab device that allows the evaluation of arbitrary STS controllers for the actuated handles – a knowledge later to be exploited for the development of lightweight mobile devices.
Lower limb prostheses – In our past research, we have performed different studies on lower limb prostheses, ranging from passive running-specific prostheses (see below) to general-purpose trans-tibial prostheses. We want to understand the effect of prostheses on whole-body motions and in particular on the stability of gait. Our long-term goal is to develop better personalized exoskeleton technology employing digital twins and in particular consider tailored prostheses for elderly users.
Prostheses in sports – In the context of prostheses in sports, in particular in athletics, there is a controversial discussion about whether the passive spring-like transtibial prostheses might be advantageous when compared to a human ankle due to their light weight and capacity to store energy, leading already to the ban of amputee athletes from regular competitions.
Our research aims to develop a simulator that addresses these questions using efficient multibody system models and optimal control techniques. Its tasks are to reconstruct the dynamics of amputee and non-amputee athletes from motion capture recordings and to explain advantages and disadvantages in amputee motions compared to non-amputee motions based on predictive simulations. Within the framework of this virtual environment, two virtual twins are generated, and the subject-specific model of an amputee athlete can be compared to his virtual non-amputee counterpart. In this research, we consider both running and long jumping motions.
Functional electrical stimulation (FES) – We are interested in stimulating muscles artificially directly or indirectly, contributing to projects on FES of patients with hemiplegia (drop foot syndrome) and on developing grasps for quadriplegic patients. For FES, model-based approaches (and in some cases their combination with model- free methods) can help to find suitable stimulation patterns.
All these assistive technologies will be developed in a design cycle involving device creation, evaluation and model-based design optimization. Subjects are expected to strongly adapt their motions under the influence of assistive technologies in an a priori unknown way which makes prediction a very challenging problem.