An Adaptive Proportional BCI-Controller for Linear Dynamic Systems (original) (raw)

2018 World Automation Congress (WAC), 2018

Abstract

Brain computer interfaces (BCIs) have attracted great attention for human computer interaction. In BCIs that are based on electroencephalography (EEG), low signal-to-noise ratio causes user intent inference to be error-prone and uncertain. Thus, the control of such systems becomes a challenging task. This paper presents an adaptive proportional BCI controller for linear dynamical systems. An approach to optimize the closed-loop control performance of dynamic systems driven by probabilistic desired/reference inputs is illustrated by designing an adaptive proportional controller that exerts forces on a mass moving linearly in one dimensional space, where the user intent is inferred by an EEG-based BCI. The BCI provides a continuous-valued output by fusing all available EEG evidence; the adaptive proportional controller optimally selects a time-varying gain that considers prior/context knowledge of user intent/environment and the probabilistic error characteristics of the BCI in inferring user intent. Such an adaptive controller that considers human-intent-inference error statistics in human-in-the-loop control systems, especially relevant for BCI-based user intent inference where accuracies may be low, improves overall closed-loop system performance measured by mean-squared tracking error. The performance of the proposed BCI-controller has been evaluated by computer simulations using code based visual evoked potentials (c-VEPs). The employed datasets were collected from healthy individuals between 20 and 30 years old with normal or corrected normal vision during calibration sessions for BCI. Results indicate that this adaptive control scheme is particularly beneficial for users with poor BCI performance.

Bahram Shafai hasn't uploaded this paper.

Let Bahram know you want this paper to be uploaded.

Ask for this paper to be uploaded.