A high-power plantar energy harvester with LSTM-KAN adaptive damping control.

Plantar energy harvesters demonstrate significant potential for continuously powering portable electronic devices. However, low output power and poor wearing comfort remain critical challenges hindering the practical deployment of plantar energy harvesters. To effectively address the aforementioned issues, this paper proposes a novel adaptive damping high-power plantar energy harvesting (DPEH) system. A bidirectionally extensible cross-beam array structure is designed, combined with pulley block-belt composite transmission system. It achieves a 17.6-fold stroke amplification, improving the power output of the DPEH significantly. Furthermore, a hybrid long short-term memory-Kolmogorov-Arnold network (LSTM-KAN) framework was developed to dynamically identify movement speeds during walking and running, achieving an accuracy rate of 99.84%. Real-Time control of DPEH electromagnetic damping via STM32 microcontroller, optimizing comfort-power tradeoffs. Experimental results demonstrate that ergonomic comfort improved by 12%. A 75 kg tester wearing the DPEH generated a peak power of 20.81 W and an average power of 2.75 W while running at 7 km/h. This strategy constitutes a new paradigm for self-regulating human-machine collaborative energy systems.