20206-04-27 北京大学（PKU）
Figure 1. Recording fish muscle signals and converting them into body motion.
＜関連情報＞
- https://newsen.pku.edu.cn/news_events/news/research/15486.html
- https://advanced.onlinelibrary.wiley.com/doi/10.1002/aisy.202501085
- https://royalsocietypublishing.org/rspb/article/292/2057/20250474/234827/Beyond-propulsion-muscle-proprioception-enables
- https://advanced.onlinelibrary.wiley.com/doi/10.1002/aisy.202501117
魚類向けEMG駆動型テレメトリおよび推論システム：姿勢再構築と流れ検知 EMG-Driven Telemetry and Inference System for Fish: Pose Reconstruction and Flow Sensing
Rahdar Hussain Afridi, Waqar Hussain Afridi, Muhammad Hamza, Ahsan Tanveer, Mingxin Wu, Xingwen Zheng, Liang Li, Guangming Xie
Advanced Intelligent Systems Published: 16 April 2026
DOI:https://doi.org/10.1002/aisy.202501085
Abstract
Intelligent sensing systems that integrate biological signals with machine learning open new opportunities to understand and replicate animal locomotion in natural environments. Conventional telemetry methods capture only limited variables and cannot reconstruct detailed kinematics or hydrodynamic context. An electromyography (EMG)-driven intelligent telemetry framework is introduced that decodes both body pose and environmental conditions in freely swimming fish. A custom 16-channel telemetry unit recorded intramuscular EMG synchronized with kinematics across laminar flows at multiple speeds, two Kármán vortex streets, a reverse Kármán vortex street, and free-swimming trials. A deep neural network mapped feature-augmented EMG to joint angles in a head-fixed frame, enabling midline reconstruction with sub-centimeter accuracy (∼3.8% body length) and joint angle prediction within 4° root mean squared erroFir (R ≈ 0.81). The same pipeline classified flow regimes and discrete flow speeds with high accuracy. Channel-efficiency analysis identified mid-body axial electrodes as sufficient to capture most flow-relevant information, guiding minimizing electrode count and invasiveness. Predicted kinematics were validated through computational fluid dynamics simulations and robotic embodiment that replayed decoded swimming motions. These results establish EMG as a dual-purpose bio-signal for locomotor and environmental inference, demonstrating an AI-driven telemetry framework that links muscle activity, kinematics, and fluid interactions.
推進力を超えて：筋肉の固有受容感覚が魚の体内の流体力学的感覚を可能にする Beyond propulsion: muscle proprioception enables hydrodynamic sensing in fish body
Rahdar Hussain Afridi;Waqar Hussain Afridi;Muhammad Hamza;Mingxin Wu;Li-Ming Chao;Yufan Zhai;Liang Li;Guangming Xie
Proceedings of the Royal Society B: Biological Sciences Published:29 Oct 2025
DOI:https://doi.org/10.1098/rspb.2025.0474
Abstract
In aquatic environments, muscle activity in free-swimming fishes not only propels body undulations to generate thrust but also serves as proprioceptive sensors for detecting surrounding fluid dynamics. Testing the proprioceptive function of the muscle is challenging owing to its deep integration with swimming activity. To address this, we introduce an experimental platform that records up to 12-channel electromyography (EMG) signals synchronized with detailed kinematics in koi and carp. We first apply various neural networks to map densely collected EMG signals to synchronized video-based body kinematics, thereby validating our EMG collection system. We then compare EMG data from fishes swimming in various laminar flows and within Kármán vortices. Our results show that the phase of muscle activity consistently precedes body kinematics in various laminar flows. While within Kármán vortices, we observe a mixed phase relationship, where muscle activity sometimes leads and at other times lags behind body kinematics. This suggests that fishes may use muscle proprioceptive sensing when interacting with complex flows, such as nearby vortices. Our research not only introduces novel methods for biological EMG studies but also offers insights that could influence the design of bio-inspired underwater sensory systems.
解釈可能なモデルを介した魚類の感覚運動ダイナミクスのバイオからロボットへの伝達 Bio-to-Robot Transfer of Fish Sensorimotor Dynamics via Interpretable Model
Waqar Hussain Afridi, Ahsan Tanveer, Rahdar Hussain Afridi, Muhammad Hamza, Mingxin Wu, Liang Li, Guangming Xie
Advanced Intelligent Systems Published: 02 December 2025
DOI:https://doi.org/10.1002/aisy.202501117
Abstract
Swimming in fish arises from tightly integrated neural, muscular, skeletal, and hydrodynamic processes that are difficult to capture in compact, transferable models for robotics. An interpretable system identification (SySID) is presented that bidirectionally maps between electromyography (EMG) and kinematics in freely swimming koi and further tests its generalization to a robotic fish. Synchronized EMG and kinematic are collected across laminar, Kármán vortex, and reverse Kármán vortex flows spanning 0.146–0.274 m s−1. A linear autoregressive with exogenous input (ARX) model architecture is chosen to capture both feedforward (EMG to kinematics) and feedback (kinematics to EMG) pathways, enabling the extraction of key system parameters, such as natural frequency, damping ratio, and input–output delays. Cross-individual validation demonstrates robust performance and identifies the best-performing fish-trained model, which is then evaluated for cross-domain transfer by replacing EMG input with processed pulse width modulation actuation signals from a robotic fish. Despite differences in mechanics and actuation physics, predictions closely match measured trajectories (mean R2 = 0.86 ± 0.13), substantially outperforming a deep neural network (97.8% higher percentage fit index) trained on the same biological datasets. These findings show that compact, interpretable SySID models enable accurate bio-to-robot transfer without robot-specific retraining, grounding robotic motion models directly in biological function rather than imitation.
