Efficient reconfigurable CPS for monitoring the elderly at home via Deep Learning
MetadataShow full item record
AuthorDéniz Cerpa, José Daniel; Isern, Juan; Solanti, Jan; Jääskeläinen, Pekka; Hnětynka, Petr; Bulej, Lubomír; Ros Vidal, Eduardo; Barranco Expósito, Francisco
Cyberphysical systemsIndoor monitoringMachine LearningRun-time adaptation
The rapid growth of the aging population poses serious challenges for our society e.g. the increase of the health care costs. Fueled by the information and communication technologies (ICT), Smart-Health Cyber-Physical Systems (CPS) offer innovative solutions to ease this burden. This work proposes a general framework for CPS-based solutions that adapt in run-time to optimize the overall performance of the system, continuously monitoring system working qualities such as bandwidth utilization, response time, accuracy, or energy consumption. Adaptation is done through automatic reconfiguration of processing Deep Learning models deployed on local edges. Local processing is performed in embedded devices that provide short latency and real-time processing despite their limited computation capacity compared to high-end cloud servers. Furthermore, a virtualization platform allows the system to offload any occasional intensive computation to such shared high-end servers. This paper demonstrates this reconfigurable CPS in a challenging scenario: indoor ambient assisted living for the elderly. Our system collects lifestyle user data in a non-invasive manner to promote healthy habits and triggers alarms in case of emergency to foster autonomy. Local edge video processing nodes identify indoor activities powered by state-of-the-art deep-learning action recognition models. The optimized embedded nodes allow the system to locally reduce cost and power consumption but the systems needs to maximize the overall performance in a changing environment. To that end, our solution enables run-time reconfiguration to adapt in terms of functionality or resource availability. The experimental section shows a real-world setup performing run-time adaptation with different reconfiguration policies. For that example, run-time adaptation extends the working time in more than 60% or increases critical action recognition accuracy up to 3x.