Bridging the Gap: Embedded Deployment of PyTorch Deep Learning Models in Complex Engineered Systems
summary
Workflow trains a PyTorch model, imports it into Simulink/MathWorks model‑based design to replace the state‑of‑charge estimator with an AI virtual sensor, adds an out‑of‑distribution binary detector, runs closed‑loop simulation against a battery plant, automatically generates C++ code and performs processor‑in‑the‑loop on a TI C2000 microcontroller, and applies calibration + quantization to shrink the model. Result: a compressed PyTorch-derived estimator that matches simulation numerics and runs on the target hardware after iterative validation.
AI-written summary · about
chapters (15)
- 0:00Embedded PyTorch deployment intro
- 1:06Embedded AI integration challenges
- 3:34Model-based design for embedded systems
- 6:27Replacing sensors with PyTorch models
- 7:56Post-training model deployment challenges
- 8:39Integrating PyTorch models into systems
- 11:27Sensor failure detection in simulations
- 13:31Testing models on microcontrollers
- 15:37Compressing PyTorch models for hardware
- 17:24Validating compressed models in simulation
- 18:48Processor-in-the-loop testing
- 19:54Model-based design workflow benefits
- 20:15Q&A: Error detection and recovery
- 22:43Q&A: Virtual vs physical sensors
- 24:07Q&A: Deep learning vs traditional estimators
related talks
14mFlexible Deployment of PyTorch Models on MCU-Class...
10mExecuTorch on Microcontrollers: Deploying PyTorch To...
32mDeploy AI with Simulink
35mClient Side Deep Learning Optimization with PyTorch
20mPyTorch Edge: Developer Journey for Deploying AI Models Onto Edge Devices
37mHow to Optimize, Validate and Deploy ML Models On Device
19mBoosting confidence in tiny AI deployment in embedded control systems
13mAI inference for the smallest processors