Empathic agents: running example in JavaScript (Node.js)

Author: Timotheus Kampik

This repository contains an example implementation of empathic agents--agents that use to make trade-offs between their own utility and the utility of other agents in their environment when deciding which actions they should execute--implemented in JavaScript (for Node.js).

Empathic agents

This repository is meant as supplementary material to a book chapter, in which we provide a detailed explanation of the empathic agent concept. A colloquial description of some of the underlying concepts can be found in this README.

Architecture

The example is implemented in JavaScript for Node.js, as a web-socket based client-server application. The agents are clients, while the server models the environment and manages agent communications:

• Socket server:

  The socket server consists of a generic server module and a environment specification that describes a particular scenario.

  • Generic server module:

    The generic server module provides the communication backbone for agent interaction, and also manages the share value system.

  • Environment specification:

    The environment specification describes the world the agents "live" in, i.e. the shared value system the agents rely on.

• Agents

  Each agent depends on the generic agent module and has its own agent specification.

  • Generic agent module:

    The generic agent module provides the empathic agent algorithms to the agents.

  • Agent specification:

    An agent specification defines an agent's utility function and registers the agents with the server.

Currently, the application is limited to two agents. For this basic demonstrations, both socket server and clients run on the same machine. The default scenario this project implements is a one-off vehicle coordination scenario (see below). It is possible to adjust the agent and environment specification to run more advanced custom scenarios; however, this scientific prototype takes some short cuts that make it infeasible for production use.

Requirements

The examples are implemented in JavaScript and require Node.js, version 10.11.0 or later, as well as npm or yarn. The instructions below assume Node.js and npm (or yarn, respectively) are installed.

Installation

Clone this repository with git@github.com:TimKam/empathic-agents-js.git. Then, install program and its dependencies with npm install or yarn. You find an overview of the required npm packages that will be automatically installed in the package.json file.

Running the example

In this project, we provide the vehicle scenario of the corresponding research paper as a running example. To run the example proceed as follows:

• Navigate to the project's root directory and open three terminal windows.

• In your first window, start the environment server by executing node ./src/server.js.

• In your second and and third window, start two agent by running node src/agent.js <scenarioType> <agentId> <agentType>, with:

  • <scenarioType>: the scenario to be executed (either vehicles, or music);

  • <agentId>: the ID of the agent (either 0 or 1);

  • <agentType>: the type of the agent (either naive, lazy, or full).

  For example, you could run node src/agent.js music 1 full in one window and node src/agent.js music 0 full in the other.

Note that you can adjust the example specification to run other 2-agent scenarios.

Testing

The project contains a set of Jasime tests. To execute the tests, run npm test.

Generating JSDoc

The code of this project is documented with JSDoc. To generate the documentation, run npm run doc. You will find the generated documentation in the out directory.

Acknowledgements

This work was partially supported by the Wallenberg AI, Autonomous Systems and Software Program (WASP) funded by the Knut and Alice Wallenberg Foundation.