Militaries are facing complex operational and tactical environments such as the increasing challenge of non-state/lone actors; persistent rapid technological development along with its broad availability via the internet; and the advent and expansion of new domains of possible threat such as information systems, space, cyberspace, electronic warfare, and autonomous weapons.
A COA is a plan describing the selected strategies and operational actions designed to accomplish the mission according to the commander’s intent. A commander’s intent is defined in terms of the goal and the end state. The goal is what the military campaign is expected to achieve. The end state is what the conditions are expected to be after the military campaign is over. Due to the existence of a typical military administrative hierarchy in command of a specific military campaign, a commander’s intent may also be represented in different levels of a hierarchy, from the strategic level through the operational level to the tactical level.
The strategic level of the commander’s intent refers to a high level commander’s intent, such as
the president’s intent. DARPA gives example: We will liberate Orangeland, restore power and control to her rightful government, and then punish the aggressor nation for its unlawful attack and occupation by significantly reducing his ability to wage war such that he is no longer a regional threat.
The operational level commander’s intent refers to the actual execution commander’s intent, i.e.,
the intent of the commander in charge of the specific military campaign. At this level, the
commander’s intent may be represented in several ways, such as end state, purpose, method, and
risk. DARPA gives example:
End state: a. freedom to operate forces starting with pre-deployment activities; b. no Weapons of Mass Destruction (WMD)/Theater Ballistic Missile (TBM) or terrorist threat to region/US
Purpose: regional stability and US security
Method: Global Strike Task Force initial strikes followed by Air and Space Epeditionary Task
Force/Carrier Battle Group persistence forces
Risk: low to US forces; medium for collateral damage.
The tactical level of commander’s intent refers to the specific objectives that the staff of the
commander in charge of the military campaign has outlined in terms of the operational level
commander’s intent. Disrupt enemy TBM Command & Control (C2) systems.
Militaries then form teams to brainstorm possible COAs and to develop a range of tentative COAs. The process of developing COAs is designed to encourage creative thinking and the application of operational art to open up the range of possibilities that could be considered. The suggested COAs are subsequently analyzed, compared, and evaluated by the staff using the method of war-gaming. After wargaming, the staff uses a decision matrix to determine which COA to recommend to the commander who makes the final decision.
However, militaries are still bound by traditional hierarchical structures, conservative organizational cultures linear-thinking doctrines and traditional strategies, methods, and procedures which privilege previous experience. Relying too exclusively on lessons from past experience will not adequately meet the operational and tactical challenges of today.
There are five fundamental issues that must be considered when developing COAs. A valid
COA should be suitable, feasible, acceptable, distinguishable and complete . A COA is
suitable if it is in alignment with the commander’s intent and will accomplish the mission when
carried out successfully. A COA is feasible if it can be achieved with the given resources. A
COA is acceptable if it balances cost and risk with advantages gained through execution. A COA
is distinguishable if it is significantly different from others and a COA is complete if it
incorporates major operations and tasks to be accomplished to accomplish the desired end state.
The military planning process depends upon analysis systems to be able to anticipate and
respond in real-time to a dynamically changing battlespace with counteractions. Complex
technical challenges exist in developing automated processes to derive hypotheses about future
alternatives for mission scenarios.
The military conducts combat operations in the presence of uncertainty and the alternatives that might emerge. It is virtually impossible to identify or predict the specific details of what might transpire. Plans and strategies, which result in COAs, are evaluated to determine the necessary steps to meet the overall strategic objectives. COA analysis is the process of performing “what if” analysis of actions and reactions and is designed to visualize the flow of the battle and evaluate each friendly COA. Due to the dynamic nature of military campaigns, COAs are continuously generated, developed and analyzed prior to execution. For each mission, thousands of COAs could be automatically generated.
COA analysis was studied through computer-generated forces in simulation using cognitive modeling. Based on individual cases, they used cognitive modeling to attempt to develop a generalized strategy for COA analysis using simulations. In a joint research project on COA analysis between Army Research Laboratory (ARL) and Ohio State University (OSU), the multi-criterial decision tool developed at OSU was used to mine ARL combat simulation data in order to gain battle-planning insights into understanding the COA space.
Recently, COA analysis has been investigated in the context of real-time decision support at the
Air Force Research Laboratory (AFRL). The current status of COA analysis .Preliminary
simulation results are reported using high performance computing facilities to achieve real-time
Due to the great challenge of the semantic gap between the commander’s intent and a COA, as
for the first phase of this investigation, we have made the following assumptions to simplify the
1. The commander’s intent is given at the tactical level. This allows a restrictive syntax to be used.
2. The COA is also given in a lower, more specific level. This also allows a restrictive syntax to be used.
3. A domain ontology must be given.
Based on these assumptions, we developed the CAFSIN solution. This solution models the
determination of the alignment problem between a commander’s intent and a COA as a fuzzified
language matching problem. This is a general approach to COA analysis and reasoning because
it addresses the uncertain and fuzzy nature of the problem using fuzzy logic analysis, and
consequently, the solution leaves a user to define what is considered as a compliant or a diverting COA. Even though CAFSIN is developed under the assumptions made above, it may also work
when the assumptions are relaxed if reliable information extraction (IE) tools are available.
DARPA launched Strategic Chaos Engine for Planning, Tactics, Experimentation and Resiliency (SCEPTER) program in Jan 2022, that seeks to develop analytic engines that will produce machine-generated strategies, capable of competing with humans in the planning of real warfare as evaluated within trusted simulation environments.
SCEPTER will discover novel and surprising courses of action (COAs) by exploring the high complexity state-action space of military engagements at machine speed. It is envisioned that high COA exploration speed will be enabled by tailorable abstraction of trusted, expert-informed models. A few of the best performing COAs will be validated in higher fidelity trusted simulators and with thorough human review.
On the other hand, a COA actually represents a specific possible option in order to achieve a
military mission, and therefore, it may also be represented in a hierarchy at different levels of
execution. For example, a higher level COA may be “attack WMD and TBM power” while a
lower level COA may be “move FA-18 at speed 500 through route 21”. Consequently, a COA
may consist of several lower level granularity COAs in sequence.
This research only addresses the suitability issue of COA analysis. In other words, given a
commander’s intent and a COA, the problem is to determine whether the COA is in alignment
with the commander’s intent, and if not, how far the COA diverts from the commander’s intent.
The challenge is that typically there are always semantic uncertainty and fuzziness for both
commander’s intent and COAs. This semantic uncertainty and fuzziness demand that not only
natural language be correctly understood, but also the semantic meaning of each word be
correctly understood, given the different context in different application.
For example, what do “control” and “ability significantly reduced” exactly mean? Due to this semantic uncertainty and fuzziness, there is a semantic gap between the commander’s intent and
a COA; the challenge is to develop a solution that overcomes this semantic gap.
In order to address the semantic uncertainty and fuzziness, we developed a fuzzified approach to
semantic inference for COA analysis, called CAFSIN, which stands for COA Analysis based on
Fuzzified Semantic INference. We demonstrated the effectiveness of the CAFSIN method
through preliminary testing and evaluations, and present the results here.
The SCEPTER program’s first phase will address two key technical focus areas: developing unscripted goal-oriented agents able to discover relevant and interpretable solutions; and managing growth of threats to achieve fast exploration of large-scale military scenarios.
SCEPTER is planned as a two-phase three-year battle planning program. This solicitation is for only an 18-month first phase. Phase 2 proposal instructions will be released to the Phase 1 performers prior to the end of Phase 1. The Phase 2 proposal instructions and program execution will be classified.
DARPA researchers say they plan to spend $39 million on the SCEPTER program over the next three years. Companies interested should upload proposals no later than 11 March 2022 to the DARPA BAA Website at https://baa.darpa.mil.