Skip to main content
arc logo
Automotive Research Center
hero image
Back to all projects

Systems of Systems & Integration

Annual Plan

Multi-Phase Vector Symbolic Architectures for Distributed and Collective Intelligence in Multi-Agent Autonomous Systems

Project Team

Principal Investigator

Maryam Parsa, George Mason University


David Gorsich, Stephen Rapp, Matt Castanier, US Army GVSC


Andrew Capodieci, Neya Robotics


Shay Snyder, George Mason University

Project Summary

Project begin 2023.

The key research questions we are addressing in this project are: (1) How to obtain collective intelligence by accounting for the shared experiences among all individuals (such as main battle tanks, Infantry fighting vehicles, or any other armored combat vehicles) within a swarm while distributing computation among all agents? and (2) How can asset-level decision-making algorithms utilize real-time statistics of power system capabilities to forecast power requirements in stochastic and noisy environments to adapt their behavior and capability accordingly? Battlefield environments are inherently stochastic due to a variety of factors such as an adversary destroying paths, moving obstacles, or obstructing the combat vehicle’s vision with smoke.

In this project, we will address the aforementioned research questions through a novel distributed but collective intelligence (DCI) framework. The building block of this framework is transferring our spatially and temporally complex problem to hyperdimensional space and analyze it through multi-level vector symbolic architectures (VSAs) by encoding the multi-dimensional perceptions, RGB cameras, infrared sensors, temperature sensors, energy meters, etc., into a series of symbolic vectors. If successful, this increases the situational awareness, reduces soldier load, and enables greater levels of sustainability and maneuverability on the battlefield. To the best of our knowledge, this would be the first application of a multi-level VSA for heterogeneous fleets of intelligent agents.

We expect VSAs to provide a distributed means of memory, computation, and intelligence for both the group and asset-level lifelong learning policies. As previous works have shown, this architecture should provide a drastic reduction in computational complexity as fleet and asset level experience increases throughout time.