Sezaki & Nishiyama Laboratory

Xiuwen GU(M1) presented “A Study on Detecting Postpartum Depression Symptoms Using Passive Mobile Sensing” at The 86th National Convention of IPSJ

Zhengbo WANG(M1) presented “Toward the Detection of Mental and Physical Stress and Recovery States in Student Athletes: Comparative Analysis Across Different Periods” at The 86th National Convention of IPSJ

【AWARD】Subaru Atsumi(M1) won the Best Paper Award at the 81th IPSJ SIGUBI!

Subaru Atsumi(M1) presented his research “UV Index Estimation Using Point Cloud Neural Networks for Signal Information of Each GNSS Satellite” at the 81st IPSJ SIGUBI.

Eri Hosonuma(D2) presented her research “Opportunistic Division and Allocation of Machine Learning Task for WSN” at The 14th International Conference on Ubiquitous and Future Networks (ICUFN 2023)!

Overview

Machine learning (ML)-applied sensing systems are widely deployed in real environments in the research and development of wireless sensor networks (WSNs).  However, in these systems, the central server must deal with the large amounts of sensing data and high processing costs to execute ML tasks. To deal with these issues, in-network processing methods of ML tasks have been proposed for WSNs.  However, their main focus is to decide the division points and allocation strategy of ML tasks, and therefore they do not consider the routing algorithm in distributed environments.  This paper proposes an opportunistic division and allocation method of ML tasks in distributed WSN environments that does not rely on a specific path. In the proposed method, each node autonomously makes a forwarding decision based on the remaining computational resource and hop count to distribute the computational load while considering the number of relays. A simulation was performed, and it revealed that the proposed method can appropriately allocate the computational processes of ML tasks and distribute them to WSN nodes.