Expressive Swarms: towards building better human-swarm communication

Lead Research Organisation: University of Bristol
Department Name: Aerospace Engineering


Project description:
Swarm robotic systems aim to produce collective behaviour through self-organised and decentralised mechanisms of control. The agents in this system depend on local interactions among one another, and interactions with the environment, to make decisions. If one agent fails, the rest of the system continues to function. Therefore, swarm robotic systems aim to be robust, scalable and flexible. The complexity of self-organised, decentralised control means that ordinarily these systems are created with a very high degree of autonomy.

Such systems contain large number of agents, which produces a challenge when human interaction is introduced to the system. It is not feasible for the human to control each and every agent in the system due to their large number. Additionally, the task of integrating human interaction into the robotic system becomes more challenging because it interferes with the mechanisms of self-organisation. However, when a swarm robotic system is applied to a problem or a task, it would be useful to be able to steer, influence or understand the system. Hence, the field of human-swarm interaction (HSI) arose. In HSI, the aim is to give the human the ability to influence the swarm system as a whole, as opposed to controlling each individual robot, or imposing a centralised control scheme.

Swarm systems may not behave perfectly, and they experience failure modes. Similarly, a human might perform errors during interaction, especially with increased numbers of agents. Hence, to reach a balance, human-swarm interaction must provide a human with a method of influencing the swarm behaviour as a whole, and at the same time, preserve the autonomy and self-organisation of the system. Striking this balance effectively is the core of the research domain.

A key parameter of human-swarm interaction is the mutual understanding of the state of the interaction between the human and the swarm. Just as it is crucial to ensure that the swarm system understands the human's interaction, it is also crucial that the human understands the state of the swarm in order to determine whether their interaction took effect, and what action to take to assert their influence. Therefore, the idea of "Expressive Swarms" in this project aims at creating a swarm system that communicates its state in an expressive manner that is easily readable by a human operator. Since creating expressive swarms has the potential of easing the human-swarm interaction task, different real-world scenarios will be studied and deployed during the work of this project, where expressing the state of the system is crucial. The research will necessitate measures of interactivity and expressivity, and whether these were congruous with the task/performance of the swarm.

This project contributes to the area of human-swarm interaction by developing a swarm system that is decentralised yet allows a human to effectively interact with it without breaking its decentralisation. Furthermore, the contribution of this project is to break new ground on measures and quantifiable human-swarm interaction.

The project will develop a number of human-swarm interaction modes that enable the human to interact successfully with the robots. Those modes will each create a feedback (such as visualisation) which expresses the internal state of the swarm system as a whole. These modes will be experimentally investigated for their effectiveness in communication, as well as the effects of the human interaction on the swarming performance. All the interaction modes will be developed within the core constraints of a swarm robotic system - the autonomy and decentralisation of the swarm will be preserved. In other words, the robots will use local interactions among one another, and interaction with their environment around them, in order to make a decision on what type of human interaction took place, if any, and consequently provide an expressive feedback.


10 25 50

Studentship Projects

Project Reference Relationship Related To Start End Student Name
EP/S021795/1 01/10/2019 31/03/2028
2282777 Studentship EP/S021795/1 28/10/2019 27/10/2022 Merihan Ahmed Alhafnawi