A Study of Marginal Performance Properties in Robotic Groups
Avi Rosenfeld, Gal A. Kaminka, Sarit Kraus · 2004
To date, only limited work has been performed on studying how performance scales with the addition of robots to groups. Our work focuses on studying this issue within one domain, robotic foraging. Foraging has been extensively studied, and is formally defined as consisting of locating target items from a search region S, and delivering them to a goal region G [3]. Previous work by Rybski et al. [5] demonstrated that groups of foraging robots do at times demonstrate marginal returns. As such, their productivity curves resembled logarithmic functions; the first several robots within their group added the most productivity per robot and each additional robot added successively less. In contrast, Fontan and Mataric’s [6] foraging robots contained a certain group size, a point they call ”critical mass”, after which the net productivity of the group dropped. Similarly, Vaughan et al. [7] wrote that the rule of ”too many cooks” applies to their groups and adding robots decreases performance after a certain group size. Economists have studied the gains in productivity within groups. According to their Law of Marginal Returns, if one factor of production is increased while the others remain constant, the overall returns will relatively decrease after a certain point. This model contains no reference to a concept similar to a ”critical mass” group size after which the added worker decreases the total productivity of the group. Our aim is to understand when the marginal returns predicted by the economic model would be consistently realized as work by [5] found they were, and when adding robots would decrease performance as [6] and [7] described.