Recurrent Neural Network Control of a Hybrid Dynamical Transfemoral Prosthesis with EdgeDRNN Accelerator

  title={Recurrent Neural Network Control of a Hybrid Dynamical Transfemoral Prosthesis with EdgeDRNN Accelerator},
  author={Chang Gao and Rachel Gehlhar and A. Ames and Shih-Chii Liu and Tobi Delbr{\"u}ck},
  journal={2020 IEEE International Conference on Robotics and Automation (ICRA)},
Lower leg prostheses could improve the life quality of amputees by increasing comfort and reducing energy to locomote, but currently control methods are limited in modulating behaviors based upon the human’s experience. This paper describes the first steps toward learning complex controllers for dynamical robotic assistive devices. We provide the first example of behavioral cloning to control a powered transfemoral prostheses using a Gated Recurrent Unit (GRU) based recurrent neural network… 
EdgeDRNN: Recurrent Neural Network Accelerator for Edge Inference
A lightweight Gated Recurrent Unit (GRU)-based RNN accelerator called EdgeDRNN that is optimized for low-latency edge RNN inference with batch size of 1 and a wall plug power efficiency that is over 4X higher than the commercial edge AI platforms.
A Survey of Human Gait-Based Artificial Intelligence Applications
An electronic database search of published works from 2012 to mid-2021 that focus on human gait studies and apply machine learning techniques is performed to provide a single broad-based survey of the applications of machine learning technology in gait analysis and identify future areas of potential study and growth.
Efficient Memory Organization for DNN Hardware Accelerator Implementation on PSoC
This work presents a software API that efficiently organizes the memory, preventing reallocating data from one memory area to another, which improves the native Linux driver with a 85% speed-up and reduces the frame computing time by 28% in a real application.
Spartus: A 9.4 TOp/s FPGA-based LSTM Accelerator Exploiting Spatio-temporal Sparsity
This paper presents a new accelerator called “Spartus” that exploits spatio-temporal sparsity to achieve ultra-low latency inference in LSTM and creates the DeltaLSTM, which saves on weight memory access and associated arithmetic operations simultaneously.
An Acceleration Based Fusion of Multiple Spatiotemporal Networks for Gait Phase Detection
This paper proposes an effective hybrid deep-learning framework based on the fusion of multiple spatiotemporal networks (FMS-Net), which is used to detect asynchronous phases from IMU signals, and has better identification accuracy than other comparative methods.
Learning Predictive Models for Ergonomic Control of Prosthetic Devices
We present Model-Predictive Interaction Primitives -- a robot learning framework for assistive motion in human-machine collaboration tasks which explicitly accounts for biomechanical impact on the
sEMG Gesture Recognition with a Simple Model of Attention
The proposed methodology's model simplicity represents a compelling alternative to the convolutional neural network (CNN) approaches utilized in recent research.


Real-Time Speech Recognition for IoT Purpose using a Delta Recurrent Neural Network Accelerator
A continuous speech recognition hardware system that uses a DeltaRNN implemented on a Xilinx Zynq-7100 FPGA to enable low latency recurrent neural network (RNN) computation and is suitable as an IoT computing platform.
Delta Networks for Optimized Recurrent Network Computation
It is shown that a naive run-time delta network implementation offers modest improvements on the number of memory accesses and computes, but optimized training techniques confer higher accuracy at higher speedup.
Human-Inspired Control of Bipedal Walking Robots
  • A. Ames
  • Computer Science
    IEEE Transactions on Automatic Control
  • 2014
The main results are that, in the case of both under and full actuation, the parameters of this controller can be determined through a human-inspired optimization problem that provides the best fit of the human data while simultaneously provably guaranteeing stable robotic walking for which the initial condition can be computed in closed form.
EdgeDRNN: Enabling Low-latency Recurrent Neural Network Edge Inference
A Gated Recurrent Unit (GRU) based recurrent neural network (RNN) accelerator called EdgeDRNN designed for portable edge computing that adopts the spiking neural network inspired delta network algorithm to exploit temporal sparsity in RNNs.
RNN-Based On-Line Continuous Gait Phase Estimation from Shank-Mounted IMUs to Control Ankle Exoskeletons
  • K. Seo, Young Jin Park, +5 authors Y. Shim
  • Computer Science, Medicine
    2019 IEEE 16th International Conference on Rehabilitation Robotics (ICORR)
  • 2019
This work proposes to estimate the gait phase as a continuous variable progressing over a gait cycle, hence allowing immediate response to frequency changes rather than iteratively correcting it after each cycle, and uses recurrent neural networks to estimate gait phases out of an inertial measurement unit every 10 ms.
Agile Autonomous Driving using End-to-End Deep Imitation Learning
This work presents an end-to-end imitation learning system for agile, off-road autonomous driving using only low-cost sensors and shows that policies trained with online imitation learning overcome well-known challenges related to covariate shift and generalize better than policies training with batch imitation learning.
An Algorithmic Perspective on Imitation Learning
This work provides an introduction to imitation learning, dividing imitation learning into directly replicating desired behavior and learning the hidden objectives of the desired behavior from demonstrations (called inverse optimal control or inverse reinforcement learning [Russell, 1998]).
DeltaRNN: A Power-efficient Recurrent Neural Network Accelerator
The implementation of an RNN delta network update approach in a hardware accelerator called "DeltaRNN" (DRNN), which leads to a 5.7x speedup compared to a conventional RNN update because of the sparsity created by the DN algorithm and the zero-skipping ability of DRNN.
Recurrent Neural Network Models for Myoelectricbased Control of a Prosthetic Hand
A set of recurrent neural networks capable of replicating the non-linear mechanism of a prosthetic hand based on surface myoelectric sensors is proposed and a comparison between the developed RNNs and a similar size non-recurrent neural network is included.
A Survey of Preference-Based Reinforcement Learning Methods
A unified framework for PbRL is provided that describes the task formally and points out the different design principles that affect the evaluation task for the human as well as the computational complexity.