cfaed Seminar Series

MSc. Kai Geißdörfer , TU Berlin

Predictive Energy Allocation for Solar Energy Harvesting Sensor Nodes in Long-Term Applications

23.11.2017 (Thursday) , 10:30 - 12:00
2nd floor, room 204 , Georg-Schumann-Str. 7A , 01069 Dresden


Wireless sensor network applications increasingly rely on harvesting energy from the environment to sense, process and transmit data. Forecasting energy availability and optimally managing energy are critical system services in this context, to ensure high energy efficiency and long-term energy-neutral operation. So far, application power requirements have been assumed to be uniform in time. The talk brings this assumption into question and shows how prior knowledge of the temporal structure of a process can help to maximize information gain from a sensing system. The dimensions of the corresponding optimization problem are outlined, before a novel algorithm is presented, which uses long-term energy predictions to optimally allocate energy according to a time varying utility profile. After demonstrating the practical benefit of the proposed algorithm for a simulated GPS tracking application, the talk is concluded with a discussion of remaining challenges and future work.


Kai Geißdörfer received his Bachelor of Science in Electrical Engineering from TU Berlin in 2015. He wrote his thesis on classification of cardiac events using neural networks on mobile sensor nodes for patient monitoring. Since then he's been working as a student research assistant under an EU-funded project with the Telecommunication Networks Group at TU Berlin, where he is the responsible developer of a novel wireless testbed for heterogeneous experiments. In 2017 he received his Master of Science, also in Electrical Engineering from TU Berlin, after spending half a year as visiting research student at the Distributed Sensing Systems Group of CSIRO in Australia. During his stay abroad he wrote his thesis on predictive energy management for solar energy harvesting sensor nodes. Kai's main interests are embedded communication systems, signal processing and machine learning.

Host: Dr. Marco Zimmerling, Independent Research Group Leader, Networked Embedded Systems Group, TU Dresden / Center for Advancing Electronics Dresden

Go back