Robot Swarm Design Patterns

Individualist

Category:

Information Transmitter pattern

Problem:

Robots need to find and exploit worksites as quickly as possible.

Applicability:

Information about worksites is easily obtainable, for example when worksite density is high (Winfield, 2009a; Pitonakova et al., 2014, 2016a, Pitonakova et al., 2018). Also recommended when continuous exploration of the environment is important, e.g., when new worksites appear over time (Pitonakova et al., 2018).

Solution:

A robot scouts for worksites in the environment. A successful scout stores information about its worksite, such as its location, in an internal data structure ("Worksite data int."), and begins work. The data structure may be updated and utilised periodically while the robot works. For example, if the robot uses odometry to localise itself relative to the worksite, the relative vector to the worksite should be updated periodically. The robot ignores any information and actions of other members of the swarm.

BDRML representation of the Individualist design pattern.

Feedback Loops:

-

Parameters:

-

Forces:

-

Consequences:

• Leads to a low information gain rate, which is why information about worksites needs to be relatively easy to find (Pitonakova et al., 2018)
• Minimises displacement and misinformation costs (Pitonakova et al., 2018)
• The spread of robots across worksites only depends on their scouting movement pattern. For example, an even spread across the environment may be achieved when robots utilise random walk (Sugawara and Watanabe, 2002; Kernbach et al., 2012; Pitonakova et al., 2018)
• Prevents the spread of erroneous information among robots (Pitonakova et al., 2014)

Known Uses:

Often used when simple foraging algorithms are needed as a basis for robot behaviour, while other swarm behaviours, such as self-regulation or task-allocation, are explored (Krieger and Billeter, 2000; Labella et al., 2006; Lerman et al., 2006; Campo and Dorigo, 2007; Kernbach et al., 2012). Also used in studies that compare swarms that do and do not utilise robot-robot recruitment (Balch and Arkin, 1994; Rybski et al., 2007; Gutiérrez et al., 2010; Lee et al., 2013; Fujisawa et al., 2014; Amato et al., 2015) and in scenarios where robots can infer information about others and about the environment through sensing (e.g., by using a camera), rather than through communication (Jones and Mataric, 2003).

Related Patterns:

Serves an alternative to the Broadcaster and Information Storage patterns, that, in general, is easier to implement and provides collective performance that is less difficult to understand due to the lack of parameters.