Meeting with Will, 24-Apr @ Reading …

Admin.
- Jon to set up weekly phone meetings with Will, every Thursday

• 1st one to be set up 4-May @10.30am
• {Jon noted Will's phone numbers}

- Jon to set up monthly meetings with Will

• every 3rd Thursday of the month
• Thu 11-May though will be a good one to attend, as there is a relevant talk
• Jon to send Will a list of suitable days, and Will to respond with clashses, etc.

Study
- {Jon outlined the intent of the project, to look at dynamic changes on the fly within flocks}
- using hard code rules to perform changes in flocking behaviour are relatively straightforwards; naturalistic rules are more challenging goals

- however, would be a good idea to use hard code rules as a baseline metric, then look to improve with a naturalistic/novel way of approaching problem

- heterogeneous robots were discussed, with herd particles having different skillsets in the herd

• different skillsets within the group
• different goals and different approaches

- two quick anecdotes were mentioned, regarding differing and combining problems

• the donkey, who is equidistant between water and food: it makes a decision to move towards the food, then realises it's thirsty, so changes direction to go to the water; on it's way to the water, it realises it's hungry, and changes direction — it never reaches either
• the rhino in the middle of the road: automated car knows it can go round either the left of the right of the rhino, but ends up hitting the rhino, as there's no reason to choose one of the separate directions
• {cf. the choice between two seemingly identical choices}

How to test the ultimate deliverable

- looking at existing problem space:

• robocup
• search and rescue

- flocking algorithms

- webbots

• khepra
• k-team robots
• EPFL

- simulation packages

• either webbots above, or …
• in-house bespoke system (need to know C#)
• roll your own {NB issue with use of bespoke packages is the fact that difficult for community to compare what you've done if non-standard simulation environment}

- Conferences to look at:

• ALife10 / ALifeX
• EvoCop
• From animals to animats (artificial animals)

Things to dig into:

• Learning and evolving
• deterministic approach (cf. chaotic) to flocking
• {Note that a lot of work has already been done in the deterministic flocking space, so you may struggle to find a new avenue}
• How distributed animats works on a real life problem (e.g. search and rescue, etc)
• ??What is the difference between flocking and swarming??
• * Evolution and Computation journal