Using Haptic Feedback in Human-Swarm Interaction

Introduction

A robotic swarm consists of a large collection of simple robots with limited sensing, communication, actuation, and computational capabilities. Individual robots act according to simple local rules and exhibit a wide range of behaviors, such as flocking (Reynolds, 1987; Couzin, Krause, James, Ruxton, & Franks, 2002; Spears & Spears, 2012; Bruemmer, 2002) without any centralized controller. However, for performing complex tasks like search and exploration in obstacle-filled environments, it is usually difficult to design local control laws for individual swarms that guarantee good performance of the overall system. To use swarm robotic systems in a complex mission, the presence of human operators are required to guide the behaviors of the swarm towards accomplishing mission goals. A key aspect of using a human to control a swarm is the transfer of information between the human and the swarm. The human has to obtain information about the state of the swarm in order to control it. In the extant literature, experimental studies in human swarm interaction have primarily explored the use of the visual channel of the human to transfer information about the swarm state. However, the use of the haptic channel has not been studied adequately in HSI, except for formation control tasks and with small multi-robot systems (fewer than 10 robots). Therefore, along with background literature, this chapter will present experiments to explore the benefits of using a haptic device to control swarm robots (in addition to the visual channel), using large numbers of robots to demonstrate scalability.

A key aspect of using a haptic device to feed information back about the robot state is to decide on the information (or cue) that should be fed back from the robots to the human. In many swarm robotic algorithms, potential field based methods (LaValle, 2006) are used for avoiding obstacles. Roughly speaking, when a robot is near an obstacle, the robot controller computes a virtual force from the obstacle that is inversely proportional to the distance (or some superlinear function of the distance) between the robot and the obstacle. Thus, the nearer a robot is to an obstacle, the greater the force it “experiences” that makes it move away from the obstacle. Therefore, one cue that can be fed back to the human is the net obstacle forces experienced by the robots.

In formation control tasks, since the robots have to usually maintain a rigid formation as they move, each robot has to track a path in order to maintain a formation. Thus, a natural cue in formation control is the tracking error of the robots in following their paths. The use of net forces from the obstacles along with tracking errors has been explored in the context of formation control (Son et al., 2011). In many applications of swarm systems, especially in obstacle-filled environments, it is not always desirable to move the robots in a formation. In these cases, there is no natural notion of tracking error, so the force cue from the obstacles can be fed back to the humans through the haptic device instead.