Towards Designing Index-free : Formation Control

Abstract

An agent must be aware of its location and surroundings within a formation. The awareness can be managed in a centralized or decentralized manner. We focus on a distributed approach due to its scalability, lower hardware requirements and inspired by swarms of animals in nature that are able to perform complex actions without a leader or external input. In the decentralized approach, an agent usually communicates with other agents to make decisions of its objective position within the multi-agent system, in order to achieve consensus. This reserves computational and communication resources. To reduce the use of resources, we introduce a formation control algorithm with low sensing and a priori information requirements that requires no communication between agents. Our algorithm needs only local coordinate systems with a common orientation and global a priori information of the formation configuration. It is able to produce almost-arbitrary formation configurations for multi-agent systems. We show examples of non-trivial formation shapes that can be achieved with an index-free approach and based only on local position measurements referenced in individual oriented coordinate systems. Both the position assignment and movement path are self-calculated by the individual agents without communication with their neighbors and based only on position measurements. In the last chapter of this document, we introduce a vision-based algorithm that could be implemented in drones or over robots to have a reliable tool for measuring distance and bearing to other near robots. These measurements can then be used in the formation control algorithm in order to achieve different formation configurations. We have developed a vision-based algorithm that uses on-board cameras and 2 LED markers on each of the agents. We have tested this algorithm for detection of LED markers in environments with different light conditions and distances from 20 cm to 6 m, with satisfactory results.