No Sensor is an Island
By Dejan Milojicic
The future of using consumer devices with a variety of sensors to assist in health care is extremely promising. It may solve a plethora of complicated procedures, simplify tracking human health, consolidating all the data and deploying consistent policies across all the devices to impose the regulatory privacy compliance. These sensors may be carried by a human, deployed at spaces, such as home or work, within SmartCities and Municipalities. All the data/information may be uploaded into the Cloud (see Figure 1). However, these approaches have been discussed for decades now [1, 2, 3]. Most of the problems have been exacerbated by the lack of privacy and regulations which limited the use and sharing of data to only most conservative cases. The following two paragraphs summarize past/present and future of systems support for Smart HealthCare specifically exploring sensors. There are numerous references in support of this discussion, author chose almost a dozen of his own publications over the course of past 20 years.
Figure 1. Sensors are deployed everywhere. On and around the user and his spaces, in municipalities, SmartCities and Smart Hospitals, etc. Models are deployed and information and models are moved from Edge to Cloud (North-South, as well as across edges (East-West) and across other entities, such as municipalities and Smart[Towns, Hospitals].
UI (User Interface), personalization, and functional ensemble are the core functions to enable better use of sensors, i.e. customizing and personalizing general sensors to individuals, how sensors interact with users and the environment and how they form a coherent ensemble of devices . There was a lot of progress lately with wearables and sensors in the phones, however this still remains one of the most important areas to enable adoption. Because wearables using sensors are very limited in resources, various operations needs to be offloaded , this is the simplest problem nowadays as phones and ambient servers are becoming more and more powerful devices. However, a standardized way  of doing this is still an open problem. Security and privacy  continue to be the biggest challenge, as well as the Platforms  that support them. Because these devices are very brittle, they can frequently break and new models of servicing them and prior to that making them more robust at scale are required . Finally, some means of communication with the Cloud (either direct or through proxy) is necessary .
How will situation change in the future? In terms of UIs, sensors will have to be integrated with new display technologies, e.g., AR/VR (Augmented Reality/Virtual Reality), with technologies such as machine learning, robots, and drones, and with a multitude of sensors outside, on the skin and inside the body . Offloading will take place in terms of model offloading to execute a degree of training at the edge using deep learning accelerators. New ways of standardization are primarily based on working code, as an evolution of open source with the introduction of Open Neural Network Exchange (ONNX) format and a benchmark for measuring machine-learning performance (MLPerf). Security and privacy will continue to be a major issue, but they will be assisted with active security prevention and sensors built into human bodies . New platforms will migrate towards AI/ML/DL (Artificial Intelligence/Machine Learning/Deep Learning) platforms, running in public and private Clouds. Most servicing/support will be based on over provisioning and decommissioning systems once a large % of the deployed infrastructure (sensors, compute, storage) are dysfunctional. Finally, 5G will dramatically help in connectivity, but local communication networks will still dominate due to the cost and accessibility.
In summary, a very interesting and promising times are in front of us that will finally enable democratization of HealthCare. This will be accomplished by delivery at home with remote access to professionals who can assist, track and facilitate non-routine activities. Systems software will transform into AI-driven, solving some old obstacles but also introducing new ones. Issues such as bias, personalization, customization will continue to be among the top of challenges. While promising, at least a decade plus is required for this to become mainstream.
- 1. Milojicic, D., Kalogeraki, V., Lukose, R., Nagaraja, K., Pruyne, J., Richard, B., Rollins, S., Xu, Z., “Peer-to-Peer Computing”, HPL Technical Report HPL-2002-57
2. Faraboschi, P., Frachtenberg, E., Laplante, P., Mansfield, K., Milojicic, D., “Technology Predictions Art, Science, and Fashion,” to appear at IEEE Computer, December 2019.
3. Alkhatib, H., Faraboschi, P. Frachtenberg, E., Kasahara, H., Lange, D., Laplante, P., Merchant, A., Milojicic, D., Schwan, K., “What Will 2022 Look Like? The IEEE CS 2022 Report,” IEEE Computer, vol 48, no 3, pp 68-76. DOI: 10.1109/MC.2015.92.
4. Kumar, R., Po4ladian, V., Greenberg, I., Messer, A., Milojicic, D., “Selecting Devices for Aggregation,” Proc. Fifth IEEE Workshop on Mobile Computing Systems & Applications (WMCSA), pp 150-169, October 2003.
5. Gu, X., Nahrstedt, K., Messer, A., Greenberg, I., Milojicic, D., “Adaptive Offloading Inference for Delivering Applications in Pervasive Computing Environments,” IEEE Pervasive, vol 3, no 3, July-September 2004, pp 66-73.
6. Milojicic, D., Breugst, M., Busse, I., Campbell, J., Covaci, S., Friedman, B., Kosaka, K., Lange, D., Ono, K., Oshima, M., Tham, C., Virdhagriswaran, S., and White, J., “MASIF – The OMG Mobile Agent System Interoperability Facility”, Personal Technologies, Springer Verlag, (1998), 2:117-129.
7. Bresniker, K., Gavrilovska, Holt, J., Milojicic, D., Tran, T., “Grand Challenge: Applying AI/ML to Cybersecurity Empowering Cyber Analysts using Standardized AI,” to appear at IEEE Computer, December 2019.
8. Messer, A., Greenberg, I., Bernadat, P., Milojicic, D., Chen, D., Giuli, T.J., Gu, X., “Towards a Distributed Platform for Resource-Constrained Devices,” Proceedings of the ICDCS 2002, July 2002, Vienna, Austria, pp 43-51.
9. Connelly, C., Cox, B., Forell, T., Liu, R., Milojicic, D., Nemeth, A., Piet, P., Shivanna, S., and Wan, W.-H., “Reiki: Serviceability Architecture and Approach for Reduction and Management of Product Service Incidents,” Proceedings of the IEEE ICWS, pp 775-782, July 2009.
10. Santana, EFZ, Chaves, AP, Gerosa, MA, Kon, F., Milojicic, D.S., “Software Platforms for Smart Cities: Concepts, Requirements, Challenges, and a Unified Reference Architecture,” accepted for publication at ACM Computing Surveys.
11. IEEE Computer Society Technology Predictions: https://www.computer.org/press-room/2018-news/ieee-cs-top-technology-trends-2019.
Hewlett Packard Labs