Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Goal Estimation-based Adaptive Shared Control for Brain-Machine Interfaces Remote Robot Navigation

Tomoka Muraoka, Tatsuya Aoki, Masayuki Hirata, Tadahiro Taniguchi, Takato Horii, Takayuki Nagai

TL;DR

The proposed shared control method significantly reduced obstacle collisions in all experiments and markedly shortened path lengths under almost all conditions in simulations and, in participant experiments, especially when user inputs become more discrete and noisy.

Abstract

In this study, we propose a shared control method for teleoperated mobile robots using brain-machine interfaces (BMI). The control commands generated through BMI for robot operation face issues of low input frequency, discreteness, and uncertainty due to noise. To address these challenges, our method estimates the user's intended goal from their commands and uses this goal to generate auxiliary commands through the autonomous system that are both at a higher input frequency and more continuous. Furthermore, by defining the confidence level of the estimation, we adaptively calculated the weights for combining user and autonomous commands, thus achieving shared control.

Goal Estimation-based Adaptive Shared Control for Brain-Machine Interfaces Remote Robot Navigation

TL;DR

The proposed shared control method significantly reduced obstacle collisions in all experiments and markedly shortened path lengths under almost all conditions in simulations and, in participant experiments, especially when user inputs become more discrete and noisy.

Abstract

In this study, we propose a shared control method for teleoperated mobile robots using brain-machine interfaces (BMI). The control commands generated through BMI for robot operation face issues of low input frequency, discreteness, and uncertainty due to noise. To address these challenges, our method estimates the user's intended goal from their commands and uses this goal to generate auxiliary commands through the autonomous system that are both at a higher input frequency and more continuous. Furthermore, by defining the confidence level of the estimation, we adaptively calculated the weights for combining user and autonomous commands, thus achieving shared control.
Paper Structure (37 sections, 7 equations, 7 figures, 1 table)

This paper contains 37 sections, 7 equations, 7 figures, 1 table.

Table of Contents

  1. INTRODUCTION
  2. PROPOSED METHOD
  3. Overall View of the Proposed System
  4. Goal Estimation
  5. Definition of the Probability Distribution
  6. Calculating $P(\bm{g} \mid \bm{v}^{user}_{t}, \bm{x}_t, M)$ Using a NN
  7. How to Train the NN
  8. Goal Selection
  9. Confidence level of estimation
  10. Method of Generating SC Commands
  11. EXPERIMENT
  12. Experimental Setup
  13. Robot
  14. Experimental Environment
  15. Experimental Conditions
  16. ...and 22 more sections

Figures (7)

  • Figure 1: Overview of the proposed system: (1) The user module receives inputs from the user and defines them as velocity commands $\bm{v}^{user}_t$ for the robot, which are then output to the other modules. (2) The goal module estimates the user's intended goal $bm{g}_t$ using environmental information and the user command, as well as calculates the confidence level $c(\bm{g}_t)$ of this estimation. (3) The robot autonomy module generates autonomous commands $\bm{v}^{auto}_t$ to navigate towards the goal estimated by the goal module. (4) The shared controller combines the user command and the autonomous command based on the confidence level of the estimation calculated by the goal module to produce the shared control command $\bm{v}^{shared}_t$, which dictates the robot's movement.
  • Figure 2: Neural network architecture. The network takes as input the map $M$, the robot's initial position $\bm{x}_0$, the robot's self-position $\bm{x}_t$ at time $t$, and the user command $\bm{v}^{user}_t$ at time $t$, and outputs $P(\bm{g} \mid \bm{v}^{user}_{t}, \bm{x}_t, M)$, given the user command at time $t$.
  • Figure 3: Data Preparation for Neural Network Training
  • Figure 4: Experimental Setup
  • Figure 5: The interfaces used to generate user commands
  • ...and 2 more figures