Robot Swarm Design Patterns


Design Pattern Catalogue

Information Exchange Centre


Information Aggregation pattern


The information exchange of robots needs to be regulated, or, robots have a low probability of meeting each other during foraging due to low density of worksites and/or robots.


Robots are able to navigate sufficiently long distances without significantly distorting their private information about worksites, e.g., as a result of cumulative effect of sensory-motor noise, which could result in incorrect information being passed to others (Pitonakova et al., 2014). Especially applicable during central-place foraging, provided that the Information Exchange Centre is identical to the place where robots need to travel to periodically in order to drop off resource (Dornhaus et al., 2006; Lemmens et al., 2008; Bailis et al., 2010; Pitonakova et al., 2014, 2016a, Pitonakova et al., 2018).


Robots meet at the Information Exchange Centre (IEC) in order to exchange information. There are two types of robots found at the IEC: informed robots, that provide information and uninformed robots that search for information. An informed robot pauses its work and returns to the IEC when its boolean recruitment initiation function, i, returns true, in order to begin providing information at the IEC. The robot leaves the IEC based on a recruitment expiry function, e, and resumes work.

BDRML representation of the Information Exchange Centre design pattern.BDRML representation of the Information Exchange Centre design pattern.

An uninformed robot located outside of the IEC, i.e., a scout, returns to the IEC based on a scouting expiry function, u, in order to check whether new information is available. If the robot finds information about where work is located, either as a result of robot-robot recruitment, or after adopting data available in a non-robot entity, it transitions to the "Work" behaviour and leaves the IEC. If no information is available in the IEC, the uninformed robot resumes scouting when a scouting initiation function, s, returns true.

Note that the relations between the data structures and other primitives have an always condition. This signifies the fact that IEC is an exchange pattern and its role is therefore to identify where information exchange takes place, but not the conditions under which information is utilised by behaviours.

Feedback Loops:

Depending on the context within which it is used and on the selected parameter values, this pattern can either enforce positive feedback loops that already exist in the swarm behaviour by designating an area where robots are likely to find information, or provide regulation of information transfer by forcing robots to travel to a designated location in order to exchange information.


  • Transmission initiation function, i: a boolean function that determines whether an informed robot returns to the IEC. For example, a robot might need to drop off resource during central-place foraging (Krieger and Billeter, 2000; Hecker and Moses, 2015).
  • Transmission expiry function, e: a boolean function that determines whether an informed robot leaves the IEC. For example, if the IEC pattern is combined with the Broadcaster pattern, expiry of a recruitment time can trigger a robot to resume work (Pitonakova et al., 2016a; Valentini et al., 2016).
  • Scouting expiry function, u: a boolean function that determines whether a scout returns to the IEC. For example, when the robot spends a certain amount of time scouting unsuccessfully (Pitonakova et al., 2016a, Pitonakova et al., 2018).
  • Scouting initiation function, s: a boolean function that determines whether an uninformed robot in the IEC becomes a scout. For example, the robot might do so with a certain probability each second (Pitonakova et al., 2016a, Pitonakova et al., 2018), or when demand for resources reaches a threshold (Krieger and Billeter, 2000).


  • The scouting efficiency of the swarm decreases due to the fact that scouts return to the IEC. The scouting expiry function, u, thus must fit the nature of the environment. For example, enough time must be given to scouts to explore a large or a dynamic working area, while at the same time ensuring that robots do not spend too much time outside of the base, where information may be readily available (Pitonakova et al., 2018).
  • The swarm size and its relation to the area of the IEC play an important role, since a large number of robots situated in the IEC at the same time can cause congestion and decrease the swarm performance (Lee et al., 2013; Pitonakova et al., 2016b).


  • Information gain rate is less dependent on the structure of the environment, on the communication range of robots and on the robot movement algorithm. The variance in information gain rate is small across different environments (Pitonakova et al., 2018).
  • Promotes spatio-temporal coordination between robots. This is advantageous when a single worksite exists in the environment. On the other hand, the swarm performance is poor when the swarm needs to concentrate on multiple worksites simultaneously (Krieger and Billeter, 2000; Pitonakova et al., 2018).
  • The amount of the incurred misplacement and misinformation costs depends on the structure of the environment, especially on the worksite distance from the IEC. A larger worksite distance generally leads to higher costs being incurred (Pitonakova et al., 2018).

Known Uses:

Most prominently used to study bee-inspired (Seeley et al., 1991) multi-robot foraging algorithms (Krieger and Billeter, 2000; Pitonakova et al., 2014; Hecker and Moses, 2015; Reina et al., 2015; Pitonakova et al., 2016a, Pitonakova et al., 2018), where robots collect items from the environment and return them to the base, where they also recruit in a peer-to-peer fashion. It has also been used to help robots recruit each other in the base during a cooperative transportation task (Amato et al., 2015).

Related Patterns:

Provides an alternative to the Information Exchange Anywhere and Information Exchange near Worksites patterns, by making the ability of robots to share information less dependent on the effectiveness of robot communication hardware and on the task parameters.

The pattern can be either combined with the Broadcaster pattern, in order to facilitate local interactions of agents in the base (Krieger and Billeter, 2000; Pitonakova et al., 2014, 2016a, Pitonakova et al., 2018), or with the Information Storage pattern, in order to turn the base into a repository of information that robots can read from without the need to meet each other (Alers et al., 2011; Hecker et al., 2012; Amato et al., 2015).

A related pattern that involved bee-inspired collective decision-making has been described in (Reina et al., 2015).