The aging society puts an increasingly heavy demand on the health care system. Elderly often have difficulties in mobility, reducing opportunities for exercise and social interaction that activities like running errands or shopping normally offer. This thesis proposes a novel concept for an assistive motorised shopping trolley, called Pull-E. Based on interview results with elderly (n=40) the Pull-E should be able to carry heavy loads (~20 kg) and still move easily and lightly while pulled over flat ground, curbs, and even stairs; and should ideally also be able to move autonomously, without being pulled. It should therefore have capabilities to balance itself, climb the stairs, carry items, and resist
perturbations.
First, a small proof-of-concept of the Pull-E was built, using tri-wheels which are separately powered by motors, that allow assistive stair-climbing when the Pull-E is pulled horizontally.
In order to balance itself the tilt angle of the trolley needs to be measured. Therefore, the Pull-E was equipped with a gyroscope and an accelerometer. From this prototype one of the most challenging aspects emerged: stabilising the trolley.
The next step was to choose a suitable control system. Since humans can adapt to the
movement of other humans, climb the stairs, and resist perturbations of other humans, the
human inspired control theories have been investigated.
Based on the literature search the risk-aware control was the most promising one. In case
of risk-aware control a value function is provided to the system, by which it can avoid the
undesirable states. By using a value function it does not require the calculation of exact
trajectories and therefore, it becomes computationally more efficient, than the other human
inspired control methods. In addition it learns and adjusts its control parameters by which
it can adapt to different situations.
In order to test its efficacy, it was implemented and tested in a simulation of the Pull-E,
where a planar projection of the tri-wheeled actuated trolley was dynamically simulated.
Force inputs could act on the trolley by using the mouse, as well as pre-determined sinusoids
acting on its centre of mass. The simulation showed the risk-aware control method already
resulted in instabilities when no perturbations affected the system. So far, the risk-aware control method was tested only for zero order systems. Therefore an adaptation of the riskaware
control method was made to allow the 2nd order system, by including angular velocities
in the state and modifying the value function. This control method is able to balance the
assistive trolley for forces that cause lower angular perturbations than 15 degrees.
In the future this control method needs to be improved to become more robust against
perturbation. The prototype needs to be rebuilt with stronger motors, and the control system
needs to be tested on the prototype. Finally, this prototype will be evaluated when human is
in the loop.
To conclude, Pull-E can facilitate elderly’s longer independence, but still a lot of work needs
to be done before it becomes a viable product.