This chapter presents new topics related to the research field of haptics and general biomechanics and deriving from the experiments the experiments described in this work, which are designed to answer more questions than approached in this thesis. Since this work is dealing with interactions between a single human and a machine, it might be interesting to include a second haptic device controller to the system environment. Melendez-Calderon (2011) already introduced a novel, versatile dual wrist interface at the International Conference on Intelligent Robots and Systems (IROS), in order to investigate the control of human-human or bimanual interaction. His goal was to investigate the mechanisms of human-human interaction in collaborative and competitive tasks. Knowledge gained from the experiments with the dual-wrist haptic device controller, can then be used for consecutive experiments. As this Thesis is only covering the simulation for one-legged running, the next step would be to extend this model into a two-legged running model.

It would be interesting as well to see the model walking, instead of running. In this particular case of human locomotion, it is necessary to consider the second leg for the calculation of the movement trajectories. While walking both feet touch the ground at some point in time. Therefore, it is mandatory to work with a two-legged model. Another interesting perspective would be to exchange the spring for a Hill-type muscle model, in order to see, how the different components affect the movement dynamics of running, respectively walking.

During each experimental trial, the start configuration remains the same, except for two initial parameters: The horizontal velocity 1R and the initial height $y_0$ of the CoM. Within the scope of this thesis, the variety of different parameter sets as initial values is limited. The main goal of this work was to implement a SLIP model and to couple it with a one- wrist haptic device controller. Subsequent works may use this SLIP model implementation with different parameter sets. Therefore, it may be interesting to discover new initial state values, which are also suitable for stable running, e.g. the initial vertical velocity, which has been set to 0. Setting a different value could give some indication of what would happen, if the SLIP model is already falling. Will this model fall down immediately, or is it possible by the means of haptic interaction to increase the spring stiffness within the decompression phase during stance? Are there more parameters, than adjusting the stiffness of the spring or adding an external force to the equation of motion, which may be modified in real-time? All these questions can be answered by setting up an experimental design with a wide spread range of different parameter values. Exploring different initial state variables, might bring new insights into the theory of stable running. While manipulating variables in real-time, several trials should be executed with different start parameter sets. Initial state parameters should change within a rational range before each trial block, e.g. horizontal velocity $v_x = 4 m/s$ may be combined with an untypical initial CoM height of $y_0 = 1.2 m$. It would be interesting as well to manipulate various parameters simultaneously. Therefore, at least a two-dimensional haptic hardware controller is needed.

Furthermore, the SLIP model used within this work takes a fixed angle of attack ($\alpha$) as a touchdown parameter during each ground contact phase. A possible model extension might be to develop a dynamically adaptable angle of attack, which is taking the leg orientation during the flight phase into consideration as well.

An interesting future work is to develop a multi-segmented SLIP model with at least a knee joint. Another important joint, that can be introduced to the system is the ankle joint. Simulations based on correct algorithms and calculations, would be able to represent the locomotion dynamics of the human lower limb more precisely. Therefore, a new experimental plan should be designed.

Eventually, the spacing between the velocity values can be reduced as well. Within the scope of this work, relatively big gaps between the velocity values during each setup have been used. A proper solution might be to reduce the incrementation interval to 0.1 m/s. Especially, when it comes to the examination of sensimotor learning tasks, it might be interesting to see, which initial state values lead to significantly better results. In both setup trials with an initial velocity of 2.5 m/s, where external forces have been applied, as well as the spring stiffness has been manipulated, the results deviate remarkably from the previous setups (i.e. $v_x = 2.0 m/s$). This leads to the question, if there is a certain velocity value between these two initial state conditions, which facilitates the user to achieve a greater running distance. As mentioned before, it might be due to sensimotor learning skills as well. When running with an initial speed of 1R = 2.0 m/s, the SLIP model falls down predominantly, because the user has a lack of experience manipulating the parameters in real-time. After the first setup is done, a certain controlling pattern is recorded by the user, which enables him to achieve more steps during a trial.

This work provides a general overview of the possibilities using a haptic device robot, and demonstrates the coupling with a biomechanical SLIP model. Additionally, the point mass trajectories could be evaluated in more detail with respect to topics such as sensimotor learning and training.

In order to treat haptic real-time manipulation as a sensimotor learning task, a single subject is not sufficient for the development of a hypothesis. Based on the human-in-the- loop (HITL) mechanism, which applies to the whole system setup, subjects generally improve the ability to command the haptic device robot with respect to achieve stable locomotion. Hence, a future work might setup an experimental design with multiple subjects (e.g. 8-10 subjects), who are trying to manipulate certain SLIP parameters during stance phase. After performing all experiments, it would be interesting to find out, if the subjects could learn to control the SLIP runner under unstable conditions.

indexmenu_n_80