The complex functions of human sensory and motor systems enable people to coordinate different segments of the body to achieve different tasks such as stable locomotion. However, they are also vulnerable to diverse disease such as SCI (spinal cord injury), stroke, or aging. These diseases which create deficits in motor control and muscle weakness affect independent living of patients. In recent years, robotic technologies have been applied to build gait driven rehabilitation robots which drive the subject's joint through normal human walking trajectories to increase the chances and effectiveness of training compared to the high labor-cost traditional manual training. However, learning balance control is limited by implementing this kind of rehabilitation strategy. Stroke-related balance deficits comprise reduced postural stability during quiet standing and delayed and less coordinated responses to both self-induced and external balance perturbations. It highlighted the importance for real life postural perturbations during walking in rehabilitation, as majority of falls occurs during walking are induced by perturbations. Thus, to recreate the real-life fall-triggering events, a platform which is able to generate different real life perturbations is needed. In the paper, we present the biomechanical background a perturbation-based balance rehabilitation strategy which helps patient to learn how to adjust posture and shift body weight by introducing different perturbations during walking. Biomechanical models are established to study the principle of human balance during walking. In particular, in order to introduce real life perturbation to increase the effectiveness of training, the biomechanics of different perturbations like slipping, stumbling, etc. are investigated. At last, the realization of perturbation-based balance rehabilitation into a robotics system is presented.

View full abstract