Modelling of bipedal locomotion for the development of a compliant pelvic interface between human and a balance assistant robot
Date of Issue2018-02-01
School of Mechanical and Aerospace Engineering
Robotics Research Centre
With the rapidly ageing of the world' s population, the WHO predicted that, by 2050, there will be about one billion people who are 65 years or older suffering from mobility-related problems. Robotic rehabilitation has been proven to provide effective therapies for people with motor deficiencies. Although multiple solutions have been developed for both balance and gait rehabilitation, both exoskeletons and robotic walkers have been shown to not have any benefit when compared with current therapies. A possible justification is a significant alteration of the activity dynamics (i.e. balance and locomotion) that prevents the skill translatability in daily living. The current knowledge about bipedal dynamics was reviewed to identify a viable solution for the improvement of both locomotion and balance therapy. The review revealed a gap in the current knowledge in the human balance and locomotion motor control, which does not provide a satisfactory explanation of how humans control and optimize their locomotion. The following hypotheses have been developed to address the limitations of bipedal models: first, the brain accounts for fixed fulcrum for the legs' inverted pendulum model, and second, the legs are not simply synchronised but are deployed as coupled oscillators. These hypotheses have enabled the formulation of an algebraic model for human locomotion that has been tested against data from multiple experiments. The results show that the proposed model can produce outputs that are fully contained within the variance of human data in both the centre of mass transverse plane trajectories and the antero-posterior swing trajectories of the feet. On the contrary, the estimation of the centre of mass of the vertical trajectories has not given definitive results, but it provided proof that most of the energy misestimation can be associated with the ankle strategies. The deeper understanding of human locomotion strategies has led to identify in the high stiffness interfaces used in most of current devices a probable cause of the altered locomotion strategies. A 5 degrees of freedom selective compliance mechanism has been designed to improve the interface between the mobile robotic base and the patient. The proposed design introduces a buffer area to minimise the interaction forces and to increase manoeuvrability. Two prototypes of 5 degrees of freedom interface have been developed. Although they share the same mechanism design, they have different joint range of motion and inertias. The first prototype has a larger motion range and weighs about 12 kg, while the second prototype has a lower inertia and a smaller range of motion. The two prototypes have been mounted on two manual carriages to study the alteration of the gait strategies generated in healthy subjects. The first prototype has been tested with 5 female subjects. The data show an alteration of their speed, step length, step width and step frequency by -23%, -22%, -7% and -2.8%, respectively. The second prototype, on the other hand, has been tested with 12 healthy subjects (7 males and 5 females) on a highly manoeuvrable carriage that allows the regulation of the interface height. The data report a reduction in the walking speed and step length of 7.6% and 9.7%, respectively; they instead increase the step width and the step frequency of 9.2% and 2.55%, respectively. Moreover, the data also suggest that the alterations generated by the proposed interface may be due to the selection of a more conscious movement strategies made from the subjects rather than their response to an external perturbation. In conclusion, the knowledge gathered about both bipedal equilibrium and human motor-control of locomotion served as a guide to build a better interface for the pelvis that is able to provide harness support and which was the main scope of this work. Furthermore, this knowledge has beneficial outcomes not only in the development of rehabilitative technologies but also in improving bipedal robots' controllers.