SS04a-21 Real-time Monitoring System for HKUST Amenities

Subscribe
Notify of
10 Comments
Inline Feedbacks
View all comments
NUGRAHA, Ferris Prima(@fpnugraha)
January 19, 2022 4:51 pm

One question from me is related to the hardware implementation and if there are specific reasons that the sensors (phototransistor, ToF) are chosen in this project. Overall, amazing project! I am looking forward to experiencing the RT monitoring system at HKUST anytime soon.

CHONG, Kai Ching(@kcchongac)
January 19, 2022 5:07 pm

Hello Ferris!
The phototransistor is a simple solution for tracking inflow of people from the entrances as only one person can pass through the gates at once. However, the situation is slightly different in the exits where people can pass through the exit one by one without any gates, making the phototransistor solution inaccurate. For that, we have switched to ToF sensors that are commonly found in commercial use cases to track reliably using the distance from the ceiling to the ground. If any person passes by, the distance will change and we can detect them leaving by using a pair of ToF sensors. (Unless they open the gates the other way to get back into the library)

Last edited 10 months ago by CHONG, Kai Ching
PUROHIT, Anushka(@apurohit)
January 19, 2022 4:14 pm

Very cool! How are you tracking how many people use the exit and leave, though? For example in HKUST’s library, when you leave, it is possible that one person opens the exit but many people go out. This could be clearer, but otherwise, this is such a useful project! I know I would definitely benefit from this data.

CHONG, Kai Ching(@kcchongac)
January 19, 2022 4:45 pm

Hello Anushka, thanks for your appreciation!
The exit is definitely a bit tricky to track accurately as multiple people can pass through the exit gates one by one. Therefore, we will be employing ToF sensors from the ceiling of the exits that detects the distance to the first object it sees, which should be the floor when no one is around. If the sensor detects a significant decrease in distance, and that is within the range of a human height, it will be counted. To accurately determine that a person has left the library, we would need 2 sensors, only when both acknowledges a person passing will it be recorded.

SHEN, Shanpu(@eeshanpu)
January 19, 2022 3:39 pm

Have you consider using energy harvesting technology to solve the power consumption issue in your design?

CHONG, Kai Ching(@kcchongac)
January 19, 2022 4:02 pm
Reply to  SHEN, Shanpu

We have considered the possibilities of using such techniques in the early stages of the project.

Piezoelectric devices that detect whether a seat is used has been explored but inserting a sensor under a seat without costing too much and causing discomfort has deemed piezoelectric not particularly useful.

We have also looked into passive energy harvesting techniques. Since we would be sending data to a server, simply waking up the sensor, connecting to the WiFi, would already consume a large amount of energy, not to mention the most energy-intensive part, sending the data to the server. Even if we can store up the energy over a period of time, the updating frequency would be too slow to produce any useful data, unless coupled with a battery and the energy harvested during downtime could suffice the use for the next day. The cost of the passive energy harvesting modules themselves is another problem, as the budget of the project is extremely limited and each of our sensors will require one as the power source. Nevertheless, we still believe that energy harvesting technology like this is a game-changer in the world of IoT and we would definitely employ them as the technology matures.

ZHANG, Jun(@eejzhang)
January 19, 2022 3:35 pm

I also have a question on the privacy issue. Any idea to address it?

CHONG, Kai Ching(@kcchongac)
January 19, 2022 3:50 pm
Reply to  ZHANG, Jun

The system would be designed to provide minimal information to the users so as to fully protect the privacy of the people tracked by the system while providing sufficient information to the users to make better decisions.

To illustrate in-depth, the source code of the micro-controllers can be made open-source for anyone that is concerned with how the data is processed. Moreover, a hash could also be generated by the source code to make sure that the micro-controller is running the corresponding firmware as shown in the open-source directory.

Another way to provide minimal, yet useful data, is to use an opaque approach on the occupancy. For instance, if we were to expand this system to the use of the washroom, where a lot of people are concerned with privacy issues, only the availability of the overall washroom will be shown or even only “whether there is any vacancy in that washroom”, instead of showing data of any particular cubicles. Such data can enable users to decide which washroom they shall go to while avoiding major privacy issues.

CAO, Xuanyu(@eexcao)
January 19, 2022 3:21 pm

How are you going to address the challenges mentioned, e.g., power consumption and privacy?

CHONG, Kai Ching(@kcchongac)
January 19, 2022 3:40 pm
Reply to  CAO, Xuanyu

For energy consumption, we will be using a relatively centralized approach (focusing on entrances and exits) to reduce maintenance and installation cost. The low power mode of the ESP32 and other means to reduce power consumption will also be applied to maximize the battery life span for minimum replacement frequency. If we are able to connect to the mains power of the library, the power consumption shall not be an issue anymore.

Regarding privacy issues, it would mostly be surrounding the use of cameras in this project. We will be processing all the camera feed on the micro-controller so any data that is uploaded or potentially compromised would only contain the occupancy of the library, which is totally anonymous and available to the public. We will also adopt a more vague approach on the occupancy, for example, instead of showing which table is occupied, we would be indicating how many vacancies there are, on a certain group of tables.

Depending on the use case, a more centralized approach shall significantly reduce the number of sensors needed, with slightly lower accuracy. Bounded by the project budget, we would be leaning towards a centralized approach to minimize cost.