Discrete Collective Estimation in Swarm Robotics with Ranked Voting Systems

Benchmarking Effects of Communication Bandwidths and Number of Options on Collective Decision-making Strategies in Discrete Collective Estimation

“…Our experiments are done in a simulated physics-based environment, as shown in Figure 1 (left). The experimental setting is similar to previous studies on the subject [4,8]. The environment is a 2-dimensional 2m × 2m arena covered by 400 tiles.…”

“…This parameter controls the level of randomness in the decision-making process. Results from previous studies [4,8] have shown that the settings of τ determines the trade-off between reliability and accuracy, and higher values of τ leads to an increase in the rate of failing to achieve a consensus but also a decrease in error.…”

“…Under DMVD, the robot automatically switches to received opinion (line 13). The implementations of the decision rules are different from in previous works on the subject [4,8], where the robot can collect opinions from all robots in its vicinity during a control loop. In this work, to keep the communication paradigm the same across considered strategies, we have limited the robot to only picking up a random broadcast per control loop, thus limiting the maximum neighborhood size to 2 and eliminating the need for unique identifications for the robots during communication.…”

“…Algorithm 2 shows our implementation of a ranked voting-based decisionmaking strategy. We consider the best performing voting system investigated in previous work on the subject [8], Borda count, with similar modifications made to bring exploration and dissemination processes into one control loop as in DC and DMVD.…”

“…These opinion-based strategies are inspired by the house hunting behavior of honeybees and are characterized by an individual agent having an explicit representation of a favored option at any given point of time, and trying to recruit its peers to its opinion during the decision-making process. There are also decision-making strategies proposed that take inspiration from sensor fusion as well as electoral process, such as distributed Bayesian belief sharing (DBBS) [4] and ranked voting strategies [8]. These decision-making strategies were proposed to specifically address the multi-option decision-making scenario of discrete collective estimation.…”