2.
Reactive Agent
Practical Reasoning Agent – BDI
Hybrid Agent
Performance Analysis
Communication and Memory
Demonstration of hybrid agent
Demonstration of multiple agents
Contents
3.
Respond to environmental changes via agent’s
percepts.
It will always react to the nearest object.
If distance to both tile and hole are the same, picking
up the tile will be of higher priority.
Reactive Agent
4.
It will make decisions based solely on the present
and does not consider past events.
Memory is not utilized.
Limitations
5.
Decisions are made based on logical reasoning.
Intentions are decided based on current situation.
Plans are made based on past memory.
BDI Agent improves Reactive Agent by
incorporating past memory for planning.
BDI Agent
6.
A vertical layering of Reactive Agent and BDI Agent.
Hybrid Agent
9.
Communication by sending “mail” to each other
The mail contains 3 TWAgentPercept
Sender’s target:
TWHole or TWTile (null if none)
Sender itself as a percept
(For receiver to access sender location)
TWHole nearest to the receiver
(With reference to sender’s memory)
Communication
10.
Memory of area in sensor range will be “refreshed”
No memory decay function as deemed unnecessary
Mail processing:
TWHole percept from mail is added to memory
Sender’s target is removed from memory unless
receiver position is closer to the target
Memory
Editor's Notes
Example: After the agent picks up a tile, instead of re-visiting a hole which it passed by previously, it will find a new hole at random.
Appropriate plans are then made to fulfill these intention and these plans are made based on its past memory, such as a previously visited hole or tile.The limitations of the bdi agent are its long decision time but this is not of an issue to our project.
Our hybrid agent is a vertical layering of our reactive agent and our bdi agent.The only improvement from bdi agent to this is that it will pick up a tile or fill the hole if it spawns right underneath the agent without interfering its current plan
Single agent runAs you can see:BDI Agent performance drastic improvement from Reactive agentHybrid agent is a combination for Reactive and BDIThe performance improvement not too drastic for config 12) The performance improvement is more drastic for config 2These are the configurations for the 2 environmentsThus this is why we chose Hybrid agent
This is the result for the double-agent run in these 3 configurationsEven though 2 agents are on the field, the individual score does not differ much from a single agent run.These are the settings for configuration 3.