Discussions on influencing swarm through self-organization principles

  • How to influence the swarm through distributed control, meaning that can I dispatch a leader agent to a location and then ask this leader agent to gather agents along the way to work on a task at a location. I wonder how this could be done effectively.
  • If we assume that the operator can talk to all the agents then there are three potential ways.
    • Case a: In the simulator, when the operator right clicks, the circle acts like a virtual agent and the radius of attraction is modified depending on the target/task. To implement this action, the gui creates a comm packet with the location of the virtual agent (mouse click) and the radius of attraction. This packet is broadcast to the network. Since, each agent knows where the virtual agent is and its field of attraction, they compute their force/movement vector. Those that are affected by the virtual agent, move to towards the virtual agent.

:: However, this is good for static virtual agent. Since, the agents have to perform a task, we can add kinematic to the virtual agent. As time t changes, the location of the virtual agent changes and moves towards the target location. Since the virtual agent has attraction force, it will drag a bunch of guys towards the target.

**Case b: As you said, one can make an agent as a leader with a certain attraction force. Based on its attraction force, the leader to take a bunch of guys to the target/task.

The above cases require operator involvement. In case a - selecting the swarm and in case b - selecting the agent that becomes a leader.

**Case c: Apart from these two cases we can use the leader quality as refined by sumpter. Some of the agents in the swarm have certain type of capabilities. So, depending on the capabilities, each agent will generate their own attraction force. The greater the capability is - higher the attraction. Thus, the agents that have low/no attraction get absorbed by the higher attraction agents. And the agent with the highest attraction will move towards the target.

::But this can cause an issue as all the agents may start moving towards the task/target - with different team sizes. In order to avoid such situations once the agents makes an attraction, it creates a repulsion force from the boundary. Thus, no more agents will get attracted to it and it has a domain over which it controls - just like in lions - one king and its kingdom. Smaller teams get repulsed by the bigger team and finally only the team with the highest attraction force will tasked.

We need further analysis and thinking for case c. But the good thing about case c is that, the operator needs to specify only the task.

  • Suppose that communication is restricted so that an operator cannot broadcast any message to all agents. The operator would only be able to broadcast to a small number of agents, and then they would be responsible for recruiting other agents – and recruiting them without being able to broadcast to all of them.
hrt/events.txt · Last modified: 2015/03/26 14:04 by ryancha
Back to top
CC Attribution-Share Alike 4.0 International
chimeric.de = chi`s home Valid CSS Driven by DokuWiki do yourself a favour and use a real browser - get firefox!! Recent changes RSS feed Valid XHTML 1.0