Human-Robotic Prosthesis as Collaborating Agents for Symmetrical Walking

Authors: Ruofan Wu, Junmin Zhong, Brent Wallace, Xiang Gao, He Huang, Jennie Si

NeurIPS 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental In this section, we provide results of a large set of simulation studies aiming at answering the following questions:... To provide answers, we show three sets of evaluations: benchmark, ablation and reliability. The main evaluation results are presented in Table 1.
Researcher Affiliation Academia Ruofan Wu Arizona State University Junmin Zhong Arizona State University Brent Abraham Wallace Arizona State University Xiang Gao Arizona State University He Huang North Carolina State University Jennie Si Arizona State University
Pseudocode No The paper describes the algorithm using mathematical equations and textual explanations of the neural network structures and update rules, but it does not include any clearly labeled pseudocode or algorithm blocks.
Open Source Code No The paper does not include an explicit statement about the release of its own source code, nor does it provide a link to a code repository.
Open Datasets No The paper mentions that 'sensor and actuator noise data extracted from real human experimental testing sessions is applied to all the simulations in this study' and 'Appendix A.2 provides the complete procedure followed of extracting noise data from experiments involving human subjects and injecting it into all the simulations.' However, it does not provide concrete access information (link, DOI, specific citation with authors/year) for this or any other dataset used in the experiments.
Dataset Splits No The paper describes different walking tasks used for evaluations (level ground, slope walking, increased pace) and notes that learning curves are averaged over multiple random seeds, but it does not provide specific details on training, validation, or test dataset splits (e.g., percentages or sample counts).
Hardware Specification No The paper states that simulations were conducted using Open Sim but does not provide any specific details about the hardware used to run these simulations, such as CPU or GPU models, or memory specifications.
Software Dependencies No The paper mentions 'Open Sim' as a tool but does not provide a specific version number. It also refers to algorithms like DDPG, PPO, SAC, TD3, MADDPG, and COMA, and an 'SGD optimizer,' but no other specific software dependencies with version numbers are listed.
Experiment Setup Yes The critic neural network is a three-layer MLP with 6 hidden units and uses linear activation function in the output layer. Therefore, we have the approximated cost to go value represented by:... Similar to the critic network, the actor networks for the human and the robot, respectively are threelayer MLP with 6 hidden units with hyperbolic activation function in the output layer to bound the action output. The same SGD optimizer [36] can be applied to the actor networks as well. lc > 0 is the learning rate of the critic network. and la > 0 is the learning rate of the actor. To make the simulations reflective of real world conditions, sensor and actuator noise data extracted from real human experimental testing sessions is applied to all the simulations in this study. Appendix A.2 provides the complete procedure followed of extracting noise data from experiments involving human subjects and injecting it into all the simulations. and Impedance parameters are reset to initial impedance values if any of the state variables exceed the safety bounds. The safety protocols followed in this work are fully described in Appendix A.4.