Learning for Microrobot Exploration: Model-based Locomotion, Robust Navigation, and Low-Power Deep Classification

Download the paper!Read on ArXiv!Run the code!Video available!
hide & show
Building intelligent autonomous systems at any scale is challenging. The sensing and computation constraints of a microrobot platform make the problems harder. We present improvements to learning-based methods for on-board learning of locomotion, classification, and navigation of microrobots. We show how simulated locomotion can be achieved with model-based reinforcement learning via on-board sensor data distilled into control. Next, we introduce a sparse, linear detector and a Dynamic Thresholding method to FAST Visual Odometry for improved navigation in the noisy regime of mm scale imagery. We end with a new image classifier capable of classification with fewer than one million multiply-and-accumulate (MAC) operations by combining fast downsampling, efficient layer structures and hard activation functions. These are promising steps toward using state-of-the-art algorithms in the power-limited world of edge-intelligence and microrobots.

What you need to know:

  1. Many current machine learning methods do scale down to low compute regimes, even though that is not where the money is at.
  2. Tightly integrating such software with the niche hardware can make more effective tools, but is a challenging interdisciplinary problem.


 author={N. O. {Lambert} and F. {Toddywala} and B. {Liao} and E. {Zhu} and L. {Lee} and K. S. J. {Pister}},
 booktitle={2020 International Conference on Manipulation, Automation and Robotics at Small Scales (MARSS)},
 title={Learning for Microrobot Exploration: Model-based Locomotion, Sparse-robust Navigation, and Low-power Deep Classification},
 doi={10.1109/MARSS49294.2020.9307921} }