The AEA TAC - trading agent competition - skills demonstrate an interaction between multiple AEAs in a game.

There are two types of AEAs:

• The tac_controller which coordinates the game.
• The tac_participant AEAs which compete in the game. The tac_participant AEAs trade tokens with each other to maximize their utility.

Discussion

The scope of this specific demo is to demonstrate how the agents negotiate autonomously with each other while they pursue their goals by playing a game of TAC. Another AEA has the role of the controller and it's responsible for calculating the revenue for each participant and if the transaction messages are valid. Transactions are settled with the controller agent rather than against a public ledger.

Communication

There are two types of interactions: - between the participants and the controller, the game communication - between the participants, the negotiation

Registration communication

This diagram shows the communication between the various entities during the registration phase.

sequenceDiagram participant Agent_2 participant Agent_1 participant Search participant Controller activate Search activate Controller Controller->>Search: register_service activate Agent_1 Agent_1->>Search: search Search-->>Agent_1: controller Agent_1->>Controller: register activate Agent_2 Agent_2->>Search: search Search-->>Agent_2: controller Agent_2->>Controller: register Controller->>Controller: start_game Controller->>Agent_1: game_data Controller->>Agent_2: game_data deactivate Agent_1 deactivate Agent_2 deactivate Search deactivate Controller

Transaction communication

This diagram shows the communication between two AEAs and the controller. In this case, we have an AEA in the role of the seller, referred to as Seller_Agent. We also have an AEA in the role of the biyer, referred to as Buyer_Agent. During a given TAC, an AEA can be in both roles simultaneously in different bilateral interactions.

sequenceDiagram participant Buyer_Agent participant Seller_Agent participant Search participant Controller activate Buyer_Agent activate Seller_Agent activate Search activate Controller Seller_Agent->>Search: register_service Buyer_Agent->>Search: search Search-->>Buyer_Agent: list_of_agents Buyer_Agent->>Seller_Agent: call_for_proposal Seller_Agent->>Buyer_Agent: proposal Buyer_Agent->>Seller_Agent: accept Seller_Agent->>Buyer_Agent: match_accept Seller_Agent->>Controller: transaction Controller->>Controller: transaction_execution Controller->>Seller_Agent: confirm_transaction Controller->>Buyer_Agent: confirm_transaction deactivate Buyer_Agent deactivate Seller_Agent deactivate Search deactivate Controller

In the above case, the proposal received contains a set of good which the seller wishes to sell and a cost of them. The buyer AEA needs to determine if this is a good deal for them and if so, it accepts.

There is an equivalent diagram for seller AEAs set up to search for buyers and their interaction with AEAs which are registered as buyers. In that scenario, the proposal will instead, be a list of goods that the buyer wishes to buy and the price it is willing to pay for them.

Preparation instructions

Dependencies

Follow the Preliminaries and Installation sections from the AEA quick start.

Launch an OEF search and communication node

In a separate terminal, launch a local OEF search and communication node.

python scripts/oef/launch.py -c ./scripts/oef/launch_config.json

Keep it running for the following demo.

Demo instructions:

Create TAC controller AEA

In the root directory, fetch the controller AEA:

aea fetch fetchai/tac_controller:0.3.0
cd tac_controller
aea install

Alternatively, create from scratch.

The following steps create the controller from scratch:

aea create tac_controller
cd tac_controller
aea add connection fetchai/oef:0.5.0
aea add skill fetchai/tac_control:0.3.0
aea install
aea config set agent.default_connection fetchai/oef:0.5.0
aea config set agent.default_ledger ethereum

Create the TAC participant AEAs

In a separate terminal, in the root directory, fetch at least two participants:

aea fetch fetchai/tac_participant:0.4.0 --alias tac_participant_one
aea fetch fetchai/tac_participant:0.4.0 --alias tac_participant_two
cd tac_participant_two
aea install

Alternatively, create from scratch.

In a separate terminal, in the root directory, create at least two tac participant AEAs:

aea create tac_participant_one
aea create tac_participant_two

Build participant one:

cd tac_participant_one
aea add connection fetchai/oef:0.5.0
aea add skill fetchai/tac_participation:0.4.0
aea add skill fetchai/tac_negotiation:0.4.0
aea install
aea config set agent.default_connection fetchai/oef:0.5.0
aea config set agent.default_ledger ethereum

Then, build participant two:

cd tac_participant_two
aea add connection fetchai/oef:0.5.0
aea add skill fetchai/tac_participation:0.4.0
aea add skill fetchai/tac_negotiation:0.4.0
aea install
aea config set agent.default_connection fetchai/oef:0.5.0
aea config set agent.default_ledger ethereum

Update the game parameters in the controller

Navigate to the tac controller project, then use the command line to get and set the start time (set it to at least two minutes in the future):

aea config get vendor.fetchai.skills.tac_control.models.parameters.args.start_time
aea config set vendor.fetchai.skills.tac_control.models.parameters.args.start_time '01 01 2020  00:01'

Run the AEAs

The CLI tool supports the launch of several agents at once.

For example, assuming you followed the tutorial, you can launch all the TAC agents as follows from the root directory:

aea launch tac_controller tac_participant_one tac_participant_two

You may want to try --multithreaded option in order to run the agents in the same process.

Cleaning up

When you're finished, delete your AEAs:

aea delete tac_controller
aea delete tac_participant_one
aea delete tac_participant_two

Previous:
Ml Skills Next:
Tac Skills Ledger-Based