You are here

Machine Learning Techniques for Tactical Mission Command



OBJECTIVE: Perform research into Machine Learning and its applicability to Mission Command in tactical environments. Improve Mission Command which includes tools, processes, and personnel across echelons involved in all phases of operations. Develop a study that considers operational environments, soldier needs and tasks, existing systems, availability of data, and the feasibility to apply Machine Learning in the Mission Command domain. 

DESCRIPTION: Human reaction time is just too slow during critical Military operations and decision making. An autonomous learned system (i.e. Machine Learning) can understand large amounts of data, manage the results, and react faster to cyber defense, electronic warfare, and large raid attacks. The Army wants to assess the potential costs, benefits, and risks of applying Machine Learning to Mission Command, the operations process, and decision making. Machine Learning relies on models that consume real-time operational data to provide predictions, alerts, and recommendations. The ultimate goal of this SBIR is to develop strategic insights into the incorporation of Machine Learning techniques to enhance human performance in the processing of information management and knowledge management in the exercise of Mission Command. Ultimately, an analysis of the cost, benefits, and risks in applying specific Machine Learning techniques to specific tasks across the planning and operations phases is required. The domain of the research is the tactical environment, specifically at the Brigade and lower levels. Machine Learning techniques applicable to Mission Command and soldiers in specific echelons should be studied, with a consideration of the tasks and data that can drive Machine Learning models. The study should also assess aspects of Machine Learning and its associated models that may be appropriate for Mission Command cross-echelon collaboration and problem solving. Fundamental to this study are (1) a deep understanding of Machine Learning techniques, (2) an understanding of peculiar Mission Command tasks in the tactical environment, and (3) a consideration of data availability. Specific data types and sources to drive the Machine Learning techniques should be delineated. 

PHASE I: This Phase will develop a methodology to assess the applicability of specific Machine Learning techniques to various Mission Command planning and operational tasks in specific echelons and environments. This should include insights into the costs and benefits of specific Mission Command task / Machine Learning combinations, and begin to highlight opportunities for possible development, as well as gaps for future research. Machine Learning Techniques - Review of Techniques, Priorities, and Justification: The contractor shall analyze Machine Learning techniques and tools by assessing their applicability to data environments and solider needs such as those found at the Brigade and lower echelons. The contractor is expected to bring a deep and broad body of Machine Learning knowledge to the research tasks. While this is not intended to be comprehensive, the contractor should build confidence in his ability to consider enterprise-level Machine Learning techniques for the less data-rich environment. Mission Command Tasks - Review of Tasks, Priorities, and Justification: The contractor shall explore the differences in Mission Command tasks by warfighting function, both within and across echelons. An assessment of the data and information that could drive Machine Learning in notional Mission Command tasks is desired. Identifying the right types of Machine Learning tasks for the various data environments across the lower echelons is key. Methodology to Predict Cost, Benefits, and Risk for Each Machine Learning Technique vs. Mission Command Task: The contractor shall develop a methodology for assessing the applicability of individual Machine Learning Techniques vs. individual Mission Command tasks / goals. Reasonable ways to assess or measure potential costs and benefits for a combination should be presented, and a way to assess risks for a specific combination should also be explored. Methodology for Validation of Cost, Benefits, and Risk Predictions for Each Machine Learning Technique vs. Mission Command Task: The contractor shall develop a detailed approach to validate the methodology for measuring costs/benefits/risks in the preceding paragraph. There may be different techniques for doing this ranging from Subject Matter Expert Review to development and use of a data-driven Machine Learning model by a prototypical user. The contractor should build confidence that the analytical approach for task and technique is sound. 

PHASE II: This Phase will develop a cross-walk of Mission Learning techniques and key Mission Command tasks in specific echelons based upon the approach and conclusions from Phase I. Using insights gained from Phase I, as well as government oversight, the contractor is expected to highlight promising Mission Command task / Machine Learning technique combinations. The Government may exercise a subset of the task/technique combinations identified in Phase I with representative data sets to develop models appropriate to specific task support. The contractor will work with the Government to validate the approach and conclusions using the methodologies from Phase I. These methodologies will be refined and matured as part of the validation. The end goal of Phase II to integrate Machine Learning into the Army tactical environment. The contractor will develop a concept demonstrator based on a set of recommendations and technical guidance to validate the integration. 

PHASE III: During Phase III of the SBIR, the contractor will mature and develop concept demonstrator(s) for integration into the Command Post and Mounted Computing Environment systems of record. Additionally, the contractor must identify potential commercial applications for the Machine Learning techniques. 


1: ADP 6-0, Mission Command, May 2012

2:  ADP 5-0 The Operations Process, May 2012

3:  ADRP 3-0 Unified Land Operations, May 2012

4:  ADRP 6-0 Mission Command, May 2012

5:  Twenty-First Century Information Warfare and the third offset Strategy (page 16 of the Joint Force Quarterly issue 82 3rd Quarter 2016

KEYWORDS: Machine Learning, Mission Command, Autonomous, Learned, Decision Making, Models, Information Management, Human Performance, Knowledge Management, Planning, Operational, Tasks 


Todd Urness 

(443) 395-0376 

Donovan Sweet 

(443) 395-0398 

US Flag An Official Website of the United States Government