eduzhai > Applied Sciences > Engineering >

Robustifying the Deployment of tinyML Models for Autonomous mini-vehicles

  • king
  • (0) Download
  • 20210505
  • Save

... pages left unread,continue reading

Document pages: 5 pages

Abstract: Standard-size autonomous navigation vehicles have rapidly improved thanks tothe breakthroughs of deep learning. However, scaling autonomous driving tolow-power systems deployed on dynamic environments poses several challengesthat prevent their adoption. To address them, we propose a closed-loop learningflow for autonomous driving mini-vehicles that includes the target environmentin-the-loop. We leverage a family of compact and high-throughput tinyCNNs tocontrol the mini-vehicle, which learn in the target environment by imitating acomputer vision algorithm, i.e., the expert. Thus, the tinyCNNs, having onlyaccess to an on-board fast-rate linear camera, gain robustness to lightingconditions and improve over time. Further, we leverage GAP8, a parallelultra-low-power RISC-V SoC, to meet the inference requirements. When runningthe family of CNNs, our GAP8 s solution outperforms any other implementation onthe STM32L4 and NXP k64f (Cortex-M4), reducing the latency by over 13x and theenergy consummation by 92 .

Please select stars to rate!

         

0 comments Sign in to leave a comment.

    Data loading, please wait...
×