Toward Autonomous Agent Simulation
Updated 3 months ago
Methodologies for Simulation of Complex Autonomous Agents of Mobility
*still many things TODO...

article links:

The AVCB Project
The Complex Autonomous System
On the Nature of Complexity
The Autonomous Mobility Agent: A story (part I)
Toward Procedural Generation
A Question of Mobility
On Complex Systems Engineering
On Heterogeneous Complex Systems
On Generalism
software components
AgentSim Simulation - The GE Component
AgentSim Simulation - The DES Component
Toward the Simulation of Vehicular Flocks (Crowds)
Toward Decentralized Complex Autonomous Organizations
The Complex Autonomous City
Notes on Complex Autonomous Systems
Notes from Neuroscience, Chemistry and Biology and the Complex System
Physics, Math and Complex Systems notes
Economics, the Social sciences, Language and complex systems notes
Notes on theory
Notes on Mobility and Transportation
Correspondences: on Agents, Energy and Entropy I
Correspondences: on Comparison and Consistency
Correspondences: an aside I

Preliminary Notes

*A continuation (or origination) of the article:
*AgentSim Middleware vs AgentSim Simulator [Simulation Component] vs AgentSim Ecosystem (clearer "distinguishments" / terminologies in future works...)




Functional Requirements

  • Automated Generation of complex environments
  • test-bed for development and design of multi-agent autonomous [vehicular] systems
  • Point of connection / integration for various software[s] involved in the design, development and [software] testing of complex autonomous [vehicular] systems
  • ...

Use Cases

  • Simulation and Analysis of networking agent applications (V2V, VANET, MANET, ...)
  • Simulation and analysis of machine learning applications for multi-agent autonomous system
  • ...

Simulation Threads

Within the ecosystem shall the study / simulation of complex systems be logically divided among a series of simulation [sub]components / threads, where each of these sub-simulators are to simulate some aspect of the overall system. These components are ideally to be made as to be compo-sable, and interactive with one another and they could be ran in a distributed manner and connect with one another during concurrent simulations.

The Software Module Components (Software Tool-Chain)

Brief Overview
Currently, are there 7 key components of functionalities identified from which a cascading chain of development tools is formed. A quasi-parallel, quasi-sequential [simulation] development cycle allows for study of complex autonomous [mobility] systems from several perspectives. The low level processing and connections are mainly encompassed within the Core Component. The development and integration of hardware systems (or devices) is to be captured by the DP Component, The generation and visualization of complex [3D] environments is performed by the GE Component, which also includes physical simulations and mathematical computations. Structuring and organization of a procedurally generated complex environment is one of the main purposes of the DES component. This component also provides a connection between the "lower level/raw" creation and processing of complex environments and provides a mechanisms for organizing, storing, and distributing information form the GE component, which in turn can be used the higher components in the tool-chain stack. The SA Component processes and analyzes all generated data from any given simulation, and is also intended to provide the capability of data curation and pre-processing for further applications. The AIP Component has the intention of providing more advanced computational processing of simulation information in the form of various AI / ML applications. The top level APP Component are where applications for the simulation component exists. This component interacts closely with the higher level interface, and the component integrates and [pseudo]directs the lower level components while implementing a [custom created] application. It is a goal to streamline the creation and testing of simulations for multi agent complex systems, and the APP component along with the high level user interface are to be the main vehicle to accomplish this, in which users shall be able to design and select desirable component for a given simulation, and the platform shall then generate scenes and scenarios as according to the desired parameters.
Ideally, due to the modular structure of these components, it is desired that a user shall be able to pick and choose which features, capabilities, or components are to be enabled, at any given time, with greater fidelity. Also, may these components, and the simulation platform as a whole, be more extensible and manageable given such a structure.
A software of softwares (A Middleware...)
The platform is a combination of many open-source softwares and packages, and does it not necessarily seek to be a purely "stand alone" platform on its own. Is there more of a desire to synthesize the plethora of already existing tools which currently exist in a fragmented state, and re-purpose these tools as to make them into a smooth continua of utilities for which can be used to aid in the development and study of complex autonomous systems. It is hoped that the great overhead that is incurred using current tools used in the development and study of these systems can be effectively eliminated and the ease for which an individual or a group can either enter into this field, or study existing applications, shall be maximal.

Core Component

Base functionality

DP Component

device integration

GE Component

core visualization

DES Component

structure / organization of scenes / scenarios / etc...

SA Component

Systems Analysis Component
  • TDA, Homological Algebra, Persistent Homology

AIP Component

AI / ML functionality

APP Component


High Level User Interface

principle user interface; complexity hiding; connective software substrate
  • GUi simulation formation
  • high level scripting API

Management System

platform management functionality

Internal Scripting

The many, various software components, libraries, applications, modules, etc... are to be coalesced into a single, high level scripting library for which can serve as a single "all in one" package for interfacing and interacting with the platform at large. As is suggested by the name, this central library works internally within the structure of the AgentSim Simulation component platform. Future works shall follow a similar hierarchical pattern, which is to say that there is the intention to have a common development language/framework for different simulation components, and their interactions and connections between the other components (Controllers, Devices, Applications, etc...) of the larger AgentSim Ecosystem.



Further Reading


End Notes

Ecosystem-Platform Development Schedule

Phase I: AVSIM (MobilitinAV) [Simulator]
  • initial ideas / preliminary work;
  • August 2018-December 2018
Phase II: AgentSim (AgentaSim Platform) [Middleware]
  • AVSIM re-factorization (version 1.xx)
  • prototype simulation platform,
  • preliminary work on Ecosystem
  • preliminary work on theory
  • preliminary work on device / controller design and development
  • AV Agent architecture / conceptual models...
  • January 2019 - December 2019
Phase III: Agentās [Ecosystem]
  • AgentSim re-factorization (version 2.xx)
  • deployment of initial simulation platform
  • finalization of Ecosystem design / etc...
  • January 2020 - December 2020
Phase [IV]: Palchya [Research, Production Environment...]
  • "Agentās Ecosystem" re-factorization (version 3.xx) [master deployment...]
  • Streamline creation of AA devices / controllers ...
  • Streamline simulation services...
  • The AVCB (Synebrum) Project (support, development, research,...)...
  • publications, manual[s], documentation, etc...
  • book[s]...
  • AV Agent architecture / conceptual models testing and verification...
  • January 2021 - ...
Phase [X]: Palchya [Future Developments]

*Notes on Complex Autonomous Systems