The third and final project, for the 02285 Artificial Intelligence and Multi-Agent Systems.

The project, in all its simplicity, is about planning. More specifically the actions of so-called agents, to solve goals by proximity.

offline/online planning?

independent agents (multi threading)

hericical approach

assume sub-goal independence

forward search best first search heuristic

negotiation highest workload = highest priority bidding for who solves the job

• Adjustable Resolution
• Goal state
• H: variable goal state
• K cluster

• Identify a problem
• Describe the problem
• 2 search
  • Realistic
  • Relaxed
• DS: route
• Predict a problem

• [-] Agent
• Msg: "Please move"
• Who moves least
• Box
• Who can move it?

• Parring box to goal?
• Clustering of goals.
• Clustering on color

• Refactor Agent - Kasper
  • [-] Relaxed by number of colours - Kasper
• Refactor Map - Daniel
• Route - Martin
• Heuristic
• Speed, profiler- martin
• Conflict - Kasper og Daniel
• Optimize node hashcode - daniel
• Optimize node equals - daniel
• Create expand of node, only using box-moves.
• Plan longer move for helper agent.
• NoOPs into stuck agent, for Helper Agent's route to the box in question.
• Identify "safe spot" to stash Helper Agent and box, while stuck agent moves.
• Reduce memory footprint
  • Reduce the search space?