robust-motor

Trustworthy Learned Models for Actuators

Every robot ultimately moves the world through electric motors. Learning-based low-level control increasingly relies on neural networks that model motor dynamics from time-series signals — but a model that is accurate on clean benchmarks is not one you can trust inside an actuator loop, where sensor noise, distribution shift, and hardware faults are the norm.

“Robustness of Neural Networks used in Electrical Motor Time-Series” (NeurIPS 2022 RobustSeq Workshop) studies how these models behave once the data stops being pristine, across three complementary datasets:

• Motor dynamics / denoising / speed-torque — core behavior characterization.
• Temperature — thermal dynamics governing safe operating limits.
• Broken bars — a classic fault-detection setting for degraded machines.

Why It Matters for Robotics

A motor controller is a safety-critical component: if a learned dynamics or fault-detection model degrades ungracefully under noise or partial faults, the failure propagates straight into the robot’s motion. Understanding where these models are brittle — and building in robustness — is a prerequisite for putting learning-based control near real actuators. The work traces back to my Ph.D. modeling heavy electrical motors with neural networks (funded by Schneider Electric).

Code, datasets, and pretrained weights: github.com/sagarverma/robust-motor · Read the blog post. </content>