top of page
Bastion_Main.png

Cerberus
Sovereign Strategic Defense AI

Predict and Counter Enemy Strategy
with Original Nash Equilibrium AI Algorithm

image.webp

Take Control of the
AI Driven Battlefield

Bastion Robotics is a leading provider of AI commander: Cerberus 

​

With the help from RoK Navy, our AI Commander already proved that

it can win 65:35 against generic pattern learning AI

​

Utilizing the power of cloud computer and digital twin, AI tactical commander trains millions simulated battles per hour to generate the optimal tactical suggestion to the human commanders, in real time.

Partners 

image.png
image.png
kaist_logo_trans.png
2038883238_UxD9sREX_CNTEC95A1EC8580EB9FACEBA088EC9DB4ED84B0_logo.png
KODIT.jpg
Monochrome on Transparent.webp
Screenshot 2024-06-15 114140-photoaidcom-invert.png

Cerberus AI Service Flow

Cerberus AI trains its intelligence by conducting millions of simulations per hour in a digital twin environment and immediately recommends the optimal tactics to personnel during an incident.

Cerberus_Service_flow_0804.png

Scalable AI Architecture

Designed to be work with general cloud computing service,

Cerberus can be deployed into from a DoD-scale large architecture

to optimized Special Forces / SWAT size training facility

Original Nash Strategy Algorithm,
Minimized Dependency,
Easy to Integrate

Since Cerberus AI is developed in-house since 2018, it does not depend on any specific hardware, cloud environment nor recent open-source AI framework

Cerberus AI can be integrated with any cloud computing environment including custom cloud IDC

Cerberus AI Engine Specifications

Pioneering Nash Equilibrium Strategy Reinforcement Learning Framework

Nash Equilibrium Strategy AI
Training Framework
 

  • Original Nash Equilibrium Strategic Defense AI Model

  • Adversarial Scenario Generation

  • Separated Blue and Red AI Training Pipeline

  • Explorative Reinforcement Learning

  • Separated Blue and Red Inference Pipeline

  • Training Monitoring and Alarm Functions

Easy to Integrate with End-User Application

 

  • Standardized Input - Output Data Format

  • Inference Optimization / Sorting Options

  • Inference and Performance Evaluation Facility

  • Inference Visualization / Replay

Supports Major Professional Simulation Engines with Cerberus API​

  • Native Cerberus SF Simulator

  • VBS4

  • Unreal

  • Unity

Supports Major Cloud Services & On-premise Installation

  • Azure, GCP, AWS, Oracle Cloud

  • Custom Cloud On-premise

  • Non-cloud On-premise

  • Extreme Scalability : From a single PC to 100,000 Cloud VMs

WIP: Cerberus Free Trial DEMO

* Not Available Yet. Expected Launch: 2025 1Q

1 / Download and Install Cerberus Scenario Editor

Minimum Requirement: A Mid-Range Gaming PC

  • OS: Windows 10 or later

  • CPU : Intel Core 5 / AMD Ryzen 5 or better

  • RAM : 16GB or better

  • GPU : GTX1050 or better​

 

* Note: this requirements is rough equivalent with one of many VM machines which will run on cloud environment for the full version of Cerberus, minus GPU, as Cerberus SF Simulator supports headless mode.

2 / Setup Native Cerberus SF Simulation Scenario

As a demo version, Cerberus DEMO has only limited set of functions 

  • Can place only 2 type of unit : Soldier and Hostage

  • Can only have 2 teams, 2 units for each team

  • Only the Red team can be trained. Blue team only uses FSM & predetermined patrol path

  • Only one type of weapon (assault rifle) is available

  • Only Deathmatch mission type is provided​

  • Only 2x and 5x time acceleration is available

  • Only basic sample ML algorithm is provided.

    • No Nash equilibrium advanced AI algorithm or other advanced algorithm is available for free trial DEMO

3 / Run the Training on the Cloud using Cloud Credit

  • Free trial credit will be given for the law enforcement organizations, military organizations and PMCs. (specific amount will vary)

  • For personal use or general purpose use other than specific application areas stated above, please contact us 

  • 100k iteration minimum, 1M iteration recommended for  a simple scenario

4 / End training and Try Inference

  • Try to win against the AI which you have trained

    • Note: playing against the trained AI also consumes free trial credit for making inferences​

article-2094871-118CE43C000005DC-673_634x400.jpg
cerberus-SF-simulator-Screenshot 2024-07-26 092731.png
79627564-be0f1580-8139-11ea-814a-d431d544f742.png

Testimonial

Yang_Uniform.jpg

As a naval ship commander who has served for a long time, I often thought that in the future, there would be a need for a military AI advisor capable of quickly and accurately assisting in the complex and diverse battlefield environments.

During that time, I became aware of a startup company that specializes in research in the field of tactical AI, and I had the opportunity to deeply engage with their research processes and outcomes.

They successfully conducted a Proof of Concept (PoC) by utilizing global top tier talents, developing and simulating an AI agent tailored to the missions and characteristics of naval ships. They then developed a tactical-level AI with a high win rate through countless simulations, which left a deep impression on me.

I believe that the tactical AI being developed by Bastion Robotics has the potential to be widely used not only in the military field but also in various areas such as counterterrorism, security, and defense. I am confident that it will dominate the future battlefield in the near future.

Col. Yang Seung Ryong, RoK Navy
Commander Cheonghae Task Force

The Team

Lead by AI and Defense Industry experts,

Bastion Robotics is building the most powerful tactical AI the world has ever seen

CEO & CTO

Marcus Kim

  • Ph. D Information Economics, KAIST

  • BS Mechanical Engineering, KAIST

  • Founder, Former CEO of Sentience, $20M Advanced AI Startup

  • 4 U.S. AI Patents, 4 Korea Patents

Big Tech AI Expert

Amy Yum

  • Kakao Dataproduct PM

  • Hyundai Motors Data Analytics Lead

  • MS Management Science KAIST

  • BS Computer Science KAIST

AI Team Manager

Junho Lee

  • Gensler Next Gen AI Team

  • MS Computer Science UC

  • BS Architect Engineering SNU

Electronics & Robotics Specialist

Youngwoo Kwon

  • Robotics and Semiconductor Expert

  • Ph.D. Busan University

  • Published papers in Advanced Science, Advanced Materials, ACS Nano, Science Advances and more

  • Serial Entrepreneur

Defense Industry Strategy Officer

Ilwon Yoon

  • RoK DoD Director of Counter Cyber Warfare

  • RoK DoD Director of Destructive Technology

  • RoK DoD Senior Manager of Compatibility Military Operation

Defense Industry Expert & Head Consultant

Jungwoo Kim

  • RoK Army Manager of C4I Systems

  • RoK DoD Senior Manager of Software Policy

  • RoK DoD Anti Cyber Threat Task Force

  • KCA Senior Consultant of Defense Industry IT Business

National Security Advisor

Sangho Park

  • Artificial Intelligent in PPS Systems - National Security Research Center 

  • Next Gen PPS System Design - National Security Research Center 

  • Ph. D.  SNU Public Policy

Ex-707 Veteran, Global PMC Advisor

Chan-il Park

  • ROK 707th anti-terrorism special force

  • Iraq War Veteran (ROK Zaytun Division Peacekeeper)

  • Hanwha Systems EOD Consultant

  • EOD Director at Global Tactics, Inc.

  • CEO, Aidios Korea

DevSecOps Expert

Kurt Dusek

  • Appdome Senior Product Manager, DevSecOps

  • GitLab Senior Solutions Architect, DevOps

  • Spectrum Principal Application Architect 

Business Development Manager

Kelly Kim

  • B.S. Global Business Administration &English Language and Literature, Sungshin Univ.

NEWS

"Si Vis Pacem, Para Bellum"

Publius Vegetius Renatus

bottom of page