Conclusion
Conclusion
People with dementia may lose the ability to take a walk on their own due to the decline in their cognitive abilities, causing them to rely on their caretakers to assist them in doing so. In this project, we aimed to design a robotic partner that could replace the caretakers during these walks. This robotic partner could potentially increase the perceived autonomy of people with dementia during walks compared to walking under the guidance of a caretaker, which could increase the overall wellbeing of people with dementia as a consequence.
The robot we used for creating this robotic partner is MiRo. MiRo is a small dog-like robot that is able to drive around on small wheels and exert dog-like behaviour such as tail-wagging, head tilting and movement of its ears. We imagined our robotic partner, called WAF, to be programmable with walking routes specified by activity coordinators or caretakers. WAF would then lead the way for PwD on walks while making sure the PwD keeps following them and alerting caretakers when something out of its control happens. Additionally, we imagined WAF to also function as a robot companion, allowing for Snoezelen.
During the design of our prototype, however, we encountered many limitations with the MiRo robot. Among other things, MiRo proved to be very fragile, disallowing for any form of cuddling, and thus, Snoezelen. We therefore excluded this from our prototype. Our ideas on this functionality can be found in the Snoezelen section below. Additionally, any physical connection such as a leash that one would have with walking an actual dog could also not be realized. Next to this, the online environment in which MiRo could be programmed proved to be very limiting, which ultimately caused us to have to 'Wizard of Oz' all of the functionalities that we imagined. Nevertheless, we set up an experiment in which we aimed to evaluate our prototype and test our claims.
In this experiment, we tasked students to act like a person with dementia while being guided by WAF and by a caretaker separately. After each walk, they were asked to fill in a questionnaire containing questions aimed to validate our claims. The results show that, with the current state of our prototype, WAF did not manage to increase the autonomy of people with dementia during walks when compared to walking under the guidance of a caretaker. The caretaker is preferred when asked if the user feels in charge, feels safe, and feels trustworthy. WAF does, however, slightly outperform the caretaker with regard to happiness, and WAF proved to be successful in guiding PwD in walks, being able to regain their attention when a PwD was distracted.
To conclude, our prototype did not manage to achieve our claims. However, the experiment itself had many limitations, described in Test, which have most definitely influenced our results. Also, the many limitations posed by MiRo caused us to not be able to implement every function that we would have liked to implement, potentially reducing the effectiveness of WAF in achieving our goals. We think that, with a more sophisticated robot, one could create a robotic partner that is able to increase the perceived autonomy of PwD in taking walks, while still ensuring safety during the walk, achieving our goals. However, a more sophisticated robot would also cost significantly more, which would deteriorate the feasibility of this solution. Perhaps, if such robots become more affordable in the future, this solution would be feasible to employ at care homes, but for now, we do not think it is.
Future Work
As described above, the MiRo robot came with many limitations that ultimately did not allow us to implement many of the functions that we had imagined. This section contains an overview of the future work that should be done to create a sophisticated robotic partner that is truly able to guide PwD on walks.
One of the limitations of the MiRo is that it is unable to walk on floors that are either uneven, carpeted or black. A new robot prototype should have some form of limbs that allow it to be able to walk up steps and traverse rough terrain, allowing the person with dementia to walk outside. Also, the robot is currently unable to walk a path that is not entirely hardcoded. The robot should be able to follow a predefined path and should be able to differ from that path if an obstacle is detected. Next to this, the robustness of a new version should be improved. This could allow for touch and/or cuddling (more on this later), but also for a physical connection such as a leash to the robot. Currently, our prototype was only able to regain the attention of the PwD by the proximity wristband and by movement. With a physical connection, this regaining of attention would be much easier. If this is not possible, however, there should an evaluation on which type of attention gaining is most effective.
Additionally, more evaluation should be done on a more representative subject group and in a more representative environment. We described our ideas for this in the conclusion of the Test section.
The robot dog should also be able to connect with the caretaker if something happens where additional assistance is necessary. Therefore, the dog should have some way of contacting the caretaker, either over data roaming or WiFi. The robot should also have a GPS or UWB support so the caretaker can locate the robot, and, more importantly, the person with dementia if they are lost or if they refuse to follow the robot.
When the person with dementia is on a walk, the robot dog should be able to listen to a name that has been created by either the care home or the individual person with dementia. When the dog can listen to a name, it can respond to its name like a real dog. The person with dementia could then get a better bond with the robot dog if they were able to give it a name to which it would respond. Finally, if the dog listens to its name, it can stop walking and start paying attention to the user to see if it would need something from the dog.
To further strengthen this bond between the person with dementia and their robot dog companion, the new version should allow for Snoezelen. The next section contains our ideas for this additional functionality, and how we would evaluate this function if our prototype allowed for this.
Snoezelen
We imagined WAF to not only be a guiding robot but also pose itself as a companion. For this, WAF should allow for 'Snoezelen'. Clients should be able to pet WAF on its head and body, and it should respond to that affection with happy noises and movements. This helps the client to create a bond with WAF and it makes walks with WAF more enjoyable. For this, WAF must not be too fragile so people can pet its head and body. Preferably, WAF's skin should be soft or nice to touch, so petting it would be preferable to petting a hard plastic dog like the MiRo.
We did take Snoezelen as a secondary function into account for our Ontology, Use Cases, Functions and Claims.
However, as we were not able to test Snoezelen, our evaluation of this function is limited to a conceptual empirical setup which is described as follows. The research question (R4) would be phrased as: how do users react to Snoezelen with a moving robot? To answer this exploratory question empirically, we would embed a between-subject study design where participants would be divided into three groups: stationary, limited movement and free movement. Multiple sessions (e.g. more than three) would be preferable to mitigate the novelty effect of engaging with WAF (i.e. the MiRo) for the first time.
In the first group, the MiRo would only make sounds and blink LEDs. In the second group, the MiRo would wag its tail and turn its head at times. In the last group, the MiRo would move in all directions, so roll forwards, backwards and side-to-side. Using the first group as a base condition, we could use metrics such as heart rate and the PwD's own experience to compare between each group and see if Snoezelen with a robot that expresses more degrees of moving freedom either overwhelms (e.g. too much stimuli caused by movement of MiRo) PwD or rather stimulates them (e.g. makes them more energetic). To allow the PwD to describe their own emotions perhaps more accurately, the AffectButton (Broekens & Brinkman, 2009) could be used to provide feedback.
Ethically, we would have to accommodate PwDs who are hypersensitive to certain stimuli by making sure they only participate in the base control group (i.e. the first one) or to exclude them if they are known to suffer from epileptic seizures. Naturally, this study would take on the form of a field study rather than a lab one as the discomfort of PwD can be a major factor in the outcome of the results of such an experiment.