Robust multi-robot formations under human supervision and control

Abstract

Abstract-There is considerable interest in real-world formation-maintenance tasks, where robots move together while maintaining a geometric shape. This interest is motivated by promise of robustly and efficiently moving multiple robots along a path, guided by a human operator. This paper presents a comprehensive set of techniques that fulfill this promise: (i) a novel method for fusing open-and closed-loop controllers, for robust formation-maintenance; (ii) an ecological display, allowing a human operator to monitor and guide robots, while improving their performance and reducing the failure rate; and (iii) a set of methods for interacting with the formation in the case of a disconnect in the formation. We evaluate each of these contributions in extensive experiments, including 25 human operators. We show significant improvements in performance (in terms of movement time), robustness (both in number of failures, as well as failure rate), and consistency between operators.

I Have a Robot, and I'm Not Afraid to Use It!

“…Most multirobot research to date, within the robotics community, focuses on a single task at a time. Some examples of such canonical tasks include moving while maintaining formations (Balch and Arkin 1998; Fredslund and Mataric 2002; Desai, Ostrowski, and Kumar 2001; Carpin and Parker 2002; Inalhan, Busse, and How 2000; Kaminka, Schechter‐Glick, and Sadov 2008; Elmaliach and Kaminka 2008), multirobot coverage (Williams and Burdick 2006; Ferranti, Trigoni, and Levene 2007; Rekleitis et al 2004; Zheng et al 2005; Rekleitis, Dudek, and Milios 2001; Batalin and Sukhatme 2002; Wagner and Bruckstein 1997; Butler, Rizzi, and Hollis 2000; Hazon and Kaminka 2008; Agmon, Hazon, and Kaminka 2008), foraging (Goldberg and Mataríc 2001; Rybski et al 1998; Rosenfeld et al 2008; Zuluaga and Vaughan 2005; Schneider‐Fontan and Mataríc 1996; Jager and Nebel 2002; Ostergaard, Sukhatme, and Mataríc 2001; Kaminka, Erusalimchik, and Kraus 2010), and patrolling or surveillance (Elmaliach, Shiloni, and Kaminka 2008; Agmon, Kraus, and Kaminka 2008; Agmon et al 2008; Jensen et al 2011; Basilico, Gatti, and Amigoni 2009; Smith, Schwager, and Rus 2011; Agmon, Urieli, and Stone 2011; Marino et al 2009; Delle Fave et al 2009). Many of these are approached from the perspective of distributed control.…”

“…In formation maintenance, robots must move in unison along a given path, while maintaining a given geometric shape. Various formation maintenance methods have been investigated (for example, Balch and Arkin [1998]; Desai [2002]; Fredslund and Mataric [2002]; Balch and Hybinette [2000]; Desai, Ostrowski, and Kumar [2001]; Carpin and Parker [2002]; Inalhan, Busse, and How [2000]; Tabuada, Pappas, and Lima [2005]; Kaminka, Schechter‐Glick, and Sadov [2008]; Elmaliach and Kaminka [2008]). All of these schemes are distributed; all require each robot to run a local control process, which executes the controller that fits the role of the robot.…”

“…A right‐following robot would do the same, but its controller would maintain the leader at a bearing of 330 degrees (that is, 30 degrees to the left). Figures 1–4 show a number of formations, the basis for the work in Kaminka and Frenkel (2005; 2007); Elmaliach and Kaminka (2008); Kaminka, Schechter‐Glick, and Sadov (2008); and Traub (2011).…”

“…Thus my research group used a teamwork architecture to manage the joint switching of controllers and the allocation of roles and tasks. We utilized the CogniTAO (CogniTeam, Ltd. 2009) commercial teamwork architecture to integrate together robust SBC (Kaminka, Schechter‐Glick, and Sadov 2008) and communication‐based formation‐maintenance controllers (Elmaliach and Kaminka 2008). The details of the integration are well beyond the scope of this article, but the lessons are not.…”

Multi-agent versus multi-robot and other Byzantine discussions

“…Hence, the issue includes various theoretical papers, such as the contribution by Bruno Lacerda et al [9], alongside others that are much more applied, for example the piece by Patrick Beeson et al [3]. An overview of the field is also included, in the shape of the invited paper by Lynne Parker [10], and a more applied contribution from Y. Elmaliach et al [6] , emphasizing the problem of human-robot interaction in this domain.…”