Common Challenges

The challenges faced by the floor monitoring system have the potential to be very complex as the research not only looks at the structure vibrations itself but aims at inferring the activities that people perform as they interact with their environment and those objects in it. These challenges are augmented when the sensors are in- stalled in a hospital setting where patients and visitors can directly interact with the system. The challenges of having a successful monitoring system are discussed in three categories: (1) Environment, focusing on installed location; and (2) Hardware focusing on the sensing system itself.

2.4.1 Environment Challenges

One of the first challenges phased during sensor installation was the limitation of which outlets were available for use. This limited the location of the sensors as the cables for power were kept to a minimal distance. Spare outlets out of the way, say behind a bed, were few. Sensors were not permitted in particular outlets because they were for emergency power, medical equipment, etc. This left mostly outlet that were in the open or in a hallway where a higher probability of tampering exists because of the visibility of the phone chargers used to power the sensors. Sensors in these areas could also be accidentally hit when a hospital bed is wheeled out of a room or someone thinks the power source is a complementary charger provided by the hospital and attempt to use it. Many of the sensors experiencing the USB Port Detached failure mode discussed in Section 2.5 were found in unprotected areas (e.g. hallway). When the sensors were hidden, people had more of a ‘out of sight, out of

mind’ mentality as they did not know the sensors or their chargers existed.

The user interaction with the sensors changed when sensors were installed in patients’ homes. Each patient was advised to not touch the equipment. There were

zero USB Port Detached failure modes in patient homes even when sensors were

easily visible. In comparison, only hospital staff (e.g. nurses) were informed about the system equipment in the hospital setting. Movement of patients through the hospital is variable, and sometimes rapid, making informing each individual patient impractical for the research team. The conclusion drawn is that equipment used in a hospital setting needs to look foreign enough to patients that they do not tamper with it, or at the very least, all equipment needs to be well hidden out of the view of patients and visitors.

Another challenge was the low amount of vibrations transfered through the rein- forced concrete flooring present in the hospital. The floors design of hospitals tend to be stricter on the amount of vibration allowed because of potentially sensitive med- ical instruments and patient comfort [18]. Hence, each patient room needed several sensors and a lower threshold level to adequately detect signals.

2.4.2 Hardware Challenges

The use of a star network as outlined in [15], provides challenges due to the operation of a base station in the form of a laptop. Restrictions placed by the hospital removed the possibility of sending data through the Internet to a centralized server. Thus, a laptop was required to collect and store data from the wireless sensors. The limited laptop WiFi range increased the number of laptops required to cover the entire hospital wing to six. This becomes one more item, and a very vital data collection piece, to keep functional; something that can be removed with access to hospital WiFi.

running when electricity goes out) and cleaning out dust accumulation. In addition, the computers needed to be hidden to prevent tampering, locked down to prevent theft, and thoroughly encrypted to protect the data according to the hospital’s reg- ulations. These devices were undisturbed to the best of our knowledge due to the proactive stance taken to secure the equipment.

The sensors themselves experienced significant failures over their operating life span. Failures include wireless communication errors and accelerometer reading fail- ures with the Mean Time To Failure (MTTF) of the sensors being approximately 170 days. These failures are most likely because the type of sensors used are prototypes for research purposes and not a commercial standard. The reliability of the sensors can be improved when using industrial level sensors. More on this is discussed in Section 2.5.

Various operational errors of the sensors lead to false-positives as readings crossed the threshold level. Some sensors reported extremely rapid oscillations to the maxi- mum swing level of the hardware as depicted in Figure 2.8. These were certainly not generated by any human activity as the cycling rate is extremely high.

Accelerometers were sometimes over-clocked or malfunctioned to produce a single high value point reading that would trigger the system. Figure 2.9 displays the error. The plot shows that the trigger could not have been caused by human activity as there is no oscillation around zero which would be expected from a floor excitation.