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 the specific demo is to demonstrate how the agents negotiate autonomously with each other while they pursue their goals by playing a game of TAC. An other 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.

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 all the following demos.

Demo instructions 1: no ledger transactions

This demo uses another AEA - a controller AEA - to take the role of running the competition and validating the transactions negotiated by the AEAs.

Create the TAC controller AEA

In the root directory, create the tac controller AEA and enter the project. 

aea create tac_controller
cd tac_controller

Add the tac control skill

aea add connection fetchai/oef:0.1.0
aea add skill fetchai/tac_control:0.1.0
aea install

Add the following configs to the aea config: 

ledger_apis:
  ethereum:
    address: https://ropsten.infura.io/v3/f00f7b3ba0e848ddbdc8941c527447fe
    chain_id: 3
    gas_price: 20

Set the default ledger to ethereum: 

aea config set agent.default_ledger ethereum

Update the game parameters

You can change the game parameters in tac_controller/skills/tac_control/skill.yaml under Parameters.

You must set the start time to a point in the future start_time: 12 11 2019 15:01.

Alternatively, use the command line to get and set the start time:

aea config get skills.tac_control.models.parameters.args.start_time
aea config set skills.tac_control.models.parameters.args.start_time '21 12 2019  07:14'

Run the TAC controller AEA

aea run --connections fetchai/oef:0.1.0

Create the TAC participants AEA

In a separate terminal, in the root directory, create the tac participant AEA. 

aea create tac_participant_one
aea create tac_participant_two

Add the tac participation skill to participant one

cd tac_participant_one
aea add connection fetchai/oef:0.1.0
aea add skill fetchai/tac_participation:0.1.0
aea add skill fetchai/tac_negotiation:0.1.0
aea install

Set the default ledger to ethereum: 

aea config set agent.default_ledger ethereum

Add the tac participation skill to participant two

cd tac_participant_two
aea add connection fetchai/oef:0.1.0
aea add skill fetchai/tac_participation:0.1.0
aea add skill fetchai/tac_negotiation:0.1.0
aea install

Set the default ledger to ethereum: 

aea config set agent.default_ledger ethereum

Run both the TAC participant AEAs

aea run --connections fetchai/oef:0.1.0

Using aea launch

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

For example, assuming you followed the tutorial, you can launch the TAC agents as follows:

  • set the default connection fetchai/oef:0.1.0 for every agent;
  • run: 
    aea launch tac_controller tac_participant_one tac_participant_two

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->>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 the two AEAs and the controller. In this case, we have a Seller_Agent which is set up as a seller (and registers itself as such with the controller during the registration phase). We also have the Searching_Agent which is set up to search for sellers.

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.

Negotiation skill - deep dive

The AEA tac_negotiation skill demonstrates how negotiation strategies may be embedded into an Autonomous Economic Agent.

The tac_negotiation skill skill.yaml configuration file looks like this.

name: tac_negotiation
authors: fetchai
version: 0.1.0
license: Apache-2.0
description: "The tac negotiation skill implements the logic for an AEA to do fipa negotiation in the TAC."
behaviours:
  behaviour:
      class_name: GoodsRegisterAndSearchBehaviour
      args:
        services_interval: 5
  clean_up:
    class_name: TransactionCleanUpTask
    args:
      tick_interval: 5.0
handlers:
  fipa:
    class_name: FIPANegotiationHandler
    args: {}
  transaction:
    class_name: TransactionHandler
    args: {}
  oef:
    class_name: OEFSearchHandler
    args: {}
models:
  search:
    class_name: Search
    args:
      search_interval: 5
  registration:
    class_name: Registration
    args:
      update_interval: 5
  strategy:
    class_name: Strategy
    args:
      register_as: both
      search_for: both
  dialogues:
    class_name: Dialogues
    args: {}
  transactions:
    class_name: Transactions
    args:
      pending_transaction_timeout: 30
protocols: ['fetchai/oef_search:0.1.0', 'fetchai/fipa:0.1.0']

Above, you can see the registered Behaviour class name GoodsRegisterAndSearchBehaviour which implements register and search behaviour of an AEA for the tac_negotiation skill.

The FIPANegotiationHandler deals with receiving FipaMessage types containing FIPA negotiation terms, such as cfp, propose, decline, accept and match_accept.

The TransactionHandler deals with TransactionMessages received from the decision maker component. The decision maker component is responsible for cryptoeconomic security.

The OEFSearchHandler deals with OefSearchMessage types returned from the OEF search node

The TransactionCleanUpTask is responsible for cleaning up transactions which are no longer likely to being settled with the controller AEA.

Models

The models element in the configuration yaml lists a number of important classes which are shared between the handlers, behaviours and tasks.

This class abstracts the logic required by AEAs performing searches for other buying/selling AEAs according to strategy (see below).

Registration

This class abstracts the logic required by AEAs performing service registrations on the OEF search node.

Strategy

This class defines the strategy behind an AEA's activities.

The class is instantiated with the AEA's goals, for example whether the AEA intends to buy/sell something, and is therefore looking for other sellers, buyers, or both.

It also provides methods for defining what goods AEAs are looking for and what goods they may have to sell, for generating proposal queries, and checking whether a proposal is profitable or not.

Dialogue

Dialogues abstract the negotiations that take place between AEAs including all negotiation end states, such as accepted, declined, etc. and all the negotiation states in between.

Transactions

This class deals with representing potential transactions between AEAs.

Previous:
Ml Skills Next:
Tac External App