Choosing Autonomy Modes for Multirobot Search

Abstract

Objective: The number of robots an operator can supervise increases with the robots' level of autonomy. The reported study investigates multirobot foraging to identify aspects of the task most suitable for automation. Background: Many envisioned applications of robotics involve multirobot teams. One of the simplest of these applications is foraging, in which robots are operated independently to explore and discover targets. Depending on levels of autonomy and task, operators have been found able to manage 3 to 12 robots. Method: The foraging task can be functionally subdivided into visiting new regions and identifying targets. In the reported experiment, full-task foraging performance was compared with exploration and perceptual search performance for 4-, 8-, and 12-robot teams in a between-groups repeated measures design. Results: Operators in the full-task condition could not successfully manage 12 robots, finding only half as many victims as perceptual search operators. Exploration performance was roughly the same in the fulltask and exploration conditions, suggesting that performance of this subtask was limiting the number of robots that could be controlled. Conclusion: Performance and workload measures indicate that exploration (navigation) tasks are the limiting factor in multirobot foraging. This finding suggests that robot navigation is the best candidate for automation. Application: Search tasks, such as foraging or perimeter control, account for many of the near-term applications envisioned for multirobot teams. The results support the choice of task-centered architectures in which the control and coordination of robotic platforms is automated, leaving search and identification of targets to human operators.

Attention allocation for human multi-robot control: Cognitive analysis based on behavior data and hidden states

“…The situation in which one operator controls a team of robots is considered to be a more exhaustive and complex task than managing a single robot, which requires the operator to simultaneously manage attentional resources among robots to maintain necessary SA. Various factors affecting human-robot supervisory control processes (such as perceived cognitive load, allocation of attention, and cognitive capacity) have been studied in previous research (Donmez et al, 2010;Lewis et al, 2010;Nagavalli et al, 2015;Visser and Parasuraman, 2011). Attentional control has been identified as one of the most critical factors influencing human supervision of robot teams, since most of the HRI tasks inevitably involve multitasking conditions (Chen andBarnes, 2014, Chappelle et al, 2011).…”

Teamwork in controlling multiple robots

“…The number of independent homogenous robots that a single operator can control is calculated by NT/IT+1. Further studies showed that the number of robots an operator can control is affected by other factors such as the nonlinear increasing complexity [9] and switching cost between robots [2].…”

“…Thirty seconds was chosen as the threshold criteria based on previous study [1,8,9] and pilot test of the experiment. In previous studies on visual search tasks [1], the possibility of finding a target was shown to decrease as more time was spent on the visual search task.…”

Scalable target detection for large robot teams

“…In previous experiments [10] it was demonstrated that a high degree of autonomy can improve performance in foraging tasks, especially in search and rescue scenarios. Consequently, in the current version of MrCS we included further autonomous functions while also improving on existing capabilities.…”