How naman1034/a-neuroadaptive-closed-loop-wearable-device-integrating-bio-inspired-reinforcement-learning-and-pred works
This repository combines Teensy sensor acquisition, Raspberry Pi real-time tremor monitoring/control, and offline model training and controller tuning. The principal intervention workflow is sensor packets into the Pi, feature/control computation, then stimulation commands; separate offline workflows train a tremor model and tune MPC parameters from recorded traces. The README is empty, and several source excerpts are partial, so links between unsampled firmware behavior and runtime components are limited to relationships explicit in the excerpts.
Main entry points: realtime.py Pi runtime · imuandemgdata.csv Recorded traces · teensy.ino Teensy firmware · imuemgdatacollect.py CSV collector
Components 15
Sensor acquisition
- Teensy firmware
Science-Fair-Project2526/teensy code/teensy.ino - CSV collector
Science-Fair-Project2526/training/imuemgdatacollect.py
Live monitoring and control
- Pi runtime
Science-Fair-Project2526/raspberrypi code/realtime.py - Serial frame reader
Science-Fair-Project2526/raspberrypi code/realtime.py - Signal features
Science-Fair-Project2526/raspberrypi code/realtime.py - Tremor inference
Science-Fair-Project2526/raspberrypi code/realtime.py - MPC controller
Science-Fair-Project2526/raspberrypi code/realtime.py - Stimulation output
Science-Fair-Project2526/raspberrypi code/realtime.py
Offline training and tuning
- CNN-LSTM training
Science-Fair-Project2526/training/cnnlstm.py - DQN training
Science-Fair-Project2526/training/rl.py - MPC tuning
Science-Fair-Project2526/training/mpc.py
Data and models
- Recorded traces
Science-Fair-Project2526/training/imuandemgdata.csv - Tremor model
Science-Fair-Project2526/training/models/tremor_model.keras - Tuned MPC parameters
Science-Fair-Project2526/training/models/mpc_cem_best_params.json
Other components
- Operator
Connections 15
- Operator → Pi runtime: starts runtime
- Teensy firmware → Serial frame reader: sends packets
Science-Fair-Project2526/raspberrypi code/realtime.py - Serial frame reader → Pi runtime: queues frames
Science-Fair-Project2526/raspberrypi code/realtime.py - Pi runtime → Signal features: computes features
Science-Fair-Project2526/raspberrypi code/realtime.py - Pi runtime → MPC controller: solves control
Science-Fair-Project2526/raspberrypi code/realtime.py - Pi runtime → Stimulation output: sends control
Science-Fair-Project2526/raspberrypi code/realtime.py - Pi runtime → Tremor inference: uses inference
Science-Fair-Project2526/raspberrypi code/realtime.py - Teensy firmware → CSV collector: streams samples
Science-Fair-Project2526/training/imuemgdatacollect.py - CSV collector → Recorded traces: writes CSV
Science-Fair-Project2526/training/imuemgdatacollect.py - Recorded traces → CNN-LSTM training: trains from
Science-Fair-Project2526/training/cnnlstm.py - CNN-LSTM training → Tremor model: saves model
Science-Fair-Project2526/training/cnnlstm.py - Recorded traces → DQN training: loads trace
Science-Fair-Project2526/training/rl.py - Recorded traces → MPC tuning: tunes from logs
Science-Fair-Project2526/training/mpc.py - MPC tuning → Tuned MPC parameters: writes parameters
Science-Fair-Project2526/training/mpc.py - Tuned MPC parameters → Pi runtime: loads parameters
Science-Fair-Project2526/raspberrypi code/realtime.py