Human-Centric Evaluation of NOIR in Real-world Tasks

Table of Links

Abstract & Introduction

Brain-Robot Interface (BRI): Background

The NOIR System

Experiments

Results

Conclusion, Limitations, and Ethical Concerns

Acknowledgments & References

Appendix 1: Questions and Answers about NOIR

Appendix 2: Comparison between Different Brain Recording Devices

Appendix 3: System Setup

Appendix 4: Task Definitions

Appendix 5: Experimental Procedure

Appendix 6: Decoding Algorithms Details

Appendix 7: Robot Learning Algorithm Details

4 Experiments

Tasks. NOIR can greatly benefit those who require assistance with everyday activities. We select tasks from the BEHAVIOR benchmark [69] and Activities of Daily Living [70] to capture actual human needs. The tasks are shown in Fig. 1, and consist of 16 tabletop tasks and four mobile manipulation tasks. The tasks encompass various categories, including eight meal preparation tasks, six cleaning tasks, three personal care tasks, and three entertainment tasks. For systematic evaluation of task success, we provide formal definitions of these activities in the BDDL language format [69, 71], which specifies the initial and goal conditions of a task using first-order logic. Task definitions and figures can be found in Appendix 4.

Procedure. The human study conducted has received approval from Institutional Review Board. Three healthy human participants (2 male, 1 female) performed all 15 Franka tasks. Sukiyaki, four Tiago tasks, and learning tasks are performed by one user. We use the EGI NetStation EEG system, which is completely non-invasive, making almost everyone an ideal subject. Before experiments, users are familiarized with task definitions and system interfaces. During the experiment, users stay in an isolated room, remain stationary, watch the robot on a screen, and solely rely on brain signals to communicate with the robots (more details about the procedure can be found in Appendix 5).