Products

Strategy Robot’s technology and products apply to all decision-making levels: tactical, operative, and strategic.

They apply to war gaming, planning, doctrine optimization, force design, basing, beddown, allocation and sequencing, C2/JADC2, reveal/conceal, EW/cyber, intelligence, security, deterrence/world stability, etc.

They do not need data about how Red has played previously (unlike supervised learning which requires prohibitive amounts of data) so they can be used in novel settings such as with hypersonic, autonomous, swarm, directed energy, long-range precision fires, novel combat vehicles, modern underground, space, offensive mining, and nuclear (also tactical nuclear where no experience exists).

The company has developed a platform technology, the Game Solving System (GSS) on which it develops product lines, see figure below. GSS has

  • game-theoretic reasoning AI technologies,

  • opponent exploitation AI technologies, and

  • hybrids of the two.

The company uses productization to achieve software quality, cost savings, and fast time to warfighter impact. We have four product lines shown in blue above. We also have additional products planned.

The “game” (i.e., scenario) can be described to Strategy Robot’s AI using several alternative approaches:

  • easy-to-use user interfaces,

  • by hooking Strategy Robot’s AI on top of an external simulation such as OneSAF, AFSIM, Command, STORM, NGTS, VBS, VR-Forces, Pioneer, etc.

  • the Strategy Robot Game Description Language, or

  • the Strategy Robot Game Programming Language.

CGES

Course-of-Action Generation and Execution System

Strategy Robot has developed a unique next-generation COA Generation and Execution System (CGES). It is the only game-theoretic COA-generation product. It computes COAs for Blue and Red simultaneously—taking into account that Blue's optimal COA depends on Red's COA and that Red's optimal COA depends on Blue's COA. So, unlike simulation, single-sided optimization, and reinforcement learning, it does not need to assume that the user has correctly guessed the adversary's COA (but it can further capitalize on aspects of it being known).

Strategy Robot has hooked CGES on top of, for example, OneSAF, the main constructive simulation tool used by the US Army. (This is a task that many others had failed at for at least 21 years and with over a hundred million dollars spent.)

Figure. Blue’s best COA depends on Red’s COA and Red’s best COA depends on Blue’s COA.

CGES is able to generate large-scale end-to-end COAs while requiring no involvement from human subject matter experts (SMEs). However, CGES also has state-of-the-art centaur capability. Users can constrain COAs to be doctrinal in several ways and make COA suggestions that our algorithm is able to include in its considerations and improve.

CGES has generated superhuman COAs, for example, for DoD scenarios with 1,400 entities. By now, our AI’s superhuman performance has been quantitatively evaluated in six sets of experiments with multiple different DoD organizations. Note that the AI has an advantage in decision speed over humans. This advantage was neutralized in these experiments by allowing the human players to pause the simulation at will. Therefore, only decision quality was evaluated and it was superhuman, that is, better than that of best humans.

AI-Generated COAs: Surprising and Successful Strategies

The AI-generated COAs were highly successful. They used surprising and varied strategies, including deception and a careful hybrid of force concentration and surprise. At the same time, many aspects of doctrine arose organically within the AI-generated COAs.

Unlike deep learning, the COAs of this AI are human understandable and explainable. They can be shown to humans via our UIs or in data or textual formats. They can be fed to downstream systems. Our system can show the human commander what the AI is going to suggest to do next so there are no sudden surprises. For example, deep reinforcement learning cannot do that.

Unlike reinforcement learning, CGES can work even with the speed of detailed DoD simulators (OneSAF only runs 2-30x real time). With faster simulations, CGES will generate even better COAs, that is, even closer approximations of game-theoretically optimal strategies.

CGES can also be integrated on top of other simulators such as AFSIM, Command, STORM, NGTS, VBS, VR-Forces, or Pioneer.

CGES can be used for:

  • wargaming

  • training

  • evaluating officers

  • evaluating COAs

  • doctrine optimization / insight generation

  • planning and dynamic replanning in command-and-control settings

The tool can be used at the tactical, operative, and strategic decision-making levels. Furthermore, the AI can be used in hexagon-based war games.

Portfolio Planning and Deployment

Strategy Robot has developed the only game-theoretic portfolio planning and deployment product family.

It computes optimal portfolios for Blue and Red simultaneously—taking into account that Blue’s optimal portfolio depends on Red’s portfolio and that Red’s optimal portfolio depends on Blue’s portfolio. So, unlike simulation and single-sided optimization, it does not need to assume that the user has correctly guessed the adversary's portfolio (but it can further capitalize on aspects of it being known).

The tool also simultaneously game-theoretically optimizes aspects of deployment with the portfolios, such as asset and munition allocation to waves of battle, and their sequencing and target assignment within each wave—taking into account that Blue’s optimal portfolio and deployment strategy depends on Red’s portfolio and deployment strategy, and vice versa.

The tool can be used operationally to decide what assets to include in an operation. It can also be used for acquisition, and it is uniquely capable of strategic planning of acquisition across multiple sequential acquisition periods.

It can be used for all domains and joint. It can also be used for multi-theater asset allocation.

Revolutionizing DoD Strategies with Innovative Tool

The tool has its own play-through model based on costs, pd’s, pk’s, group effects, ranges, etc., and does not need an external simulation tool.

The tool has already found strategies that exhibited novel aspects that had not been previously considered in that setting, such as constructing Blue portfolios that cause Red’s financial resources to drain—in some models so much that Red’s best decision is not to build any assets, which means that Blue achieved deterrence at the acquisition level.

The tool has also helped the DoD significantly sanity check and modernize the utility functions used for planning. Strategy Robot’s AI uncovered that the traditional utility functions (based on kill ratios) used by militaries have nonsensical aspects in them. This had not been discovered before because planning had been heuristic and manual. This discovery then led to an AI-driven utility function refinement process in the Pentagon where commander’s intent is rigorously captured into the utility function—beyond what was possible before.

So far Strategy Robot has developed three products in this product family:

  • Multi-Base Defense System (MBDS). It optimizes the portfolio, basing/beddown, allocation of assets to waves of battle, and targeting within each wave for both Blue and Red simultaneously. It finds the optimal defense against ballistic, cruise, and hypersonic missiles, drones, planes, helicopters, etc.

  • Portfolio Planning System. Optimizes asset portfolios for maximum effectiveness and deterrence. For air, ground, sea, space, EW, cyber, and joint. It has generated novel portfolios and deployment strategies.

  • Missile Defense and Offense System.

Electronic warfare (EW) is included in all of these products.