Search and rescue, autonomous construction, and many other semi-autonomous multirobot applications can benefit from proximal interactions between an operator and a swarm of robots. Most research on proximal interaction is based on explicit communication techniques such as gesture and speech. This study proposes a new implicit proximal communication technique to approach the problem of robot selection. We use electroencephalography (EEG) signals to select the robot at which the operator is looking. This is achieved using steady-state visually evoked potential (SSVEP), a repeatable neural response to a regularly blinking visual stimulus that varies predictively based on the blinking frequency. In our experiments, each robot was equipped with LEDs blinking at a different frequency, and the operator’s SSVEP neural response was extracted from the EEG signal to detect and select the robot without requiring any conscious action by the user. This study systematically investigates several parameters affecting the SSVEP neural response: blinking frequency of the LED, distance between the robot and the operator, and color of the LED. Based on these parameters, we study two signal processing approaches and critically analyze their performance on 10 subjects controlling a set of physical robots. Our results show that despite numerous artifacts, it is possible to achieve a recognition rate higher than 85 % on some subjects, while the average over the ten subjects was 75 %.

中文翻译：

---

脑电图作为人类与机器人群体之间近距离交互的隐式交流渠道

搜索和救援，自主建造以及许多其他半自主多机器人应用程序都可以从操作员与机器人群之间的近距离交互中受益。关于近端交互的大多数研究都基于诸如手势和语音之类的显式通信技术。这项研究提出了一种新的隐性近距离通信技术来解决机器人选择问题。我们使用脑电图（EEG）信号来选择操作员正在看的机器人。这是通过使用稳态视觉诱发电位（SSVEP）来实现的，SSVEP是对规则眨眼视觉刺激的可重复神经反应，它会根据眨眼频率而预测性地变化。在我们的实验中，每个机器人都配备了以不同频率闪烁的LED，并且从EEG信号中提取了操作员的SSVEP神经反应，以检测和选择机器人，而无需用户有意识地采取任何行动。这项研究系统地研究了影响SSVEP神经反应的几个参数：LED的闪烁频率，机器人与操作员之间的距离以及LED的颜色。根据这些参数，我们研究了两种信号处理方法，并严格分析了它们在控制一组物理机器人的10个对象上的性能。我们的结果表明，尽管有许多伪像，但在某些对象上仍可以实现高于85％的识别率，而在十个对象上的平均识别率为75％。