Background on Team Members
Our team comprises 5 Active Duty O-1s from the Navy, Air Force, and Army. Each of us are currently in graduate school for our first assignment. 3 of us are studying Business Analytics and the other 2 are studying Finance. With little experience with hardware or computer science, we framed the hackathon through the lens of our graduate school disciplines. As a result, we focused on a mathematical optimization formulation and its associated finances.
Inspiration
Our project was inspired by the dual challenges faced in modern urban environments: the need for effective civil infrastructure monitoring in smart cities like Boston, and the specialized requirements of military operations in dense urban terrain. The methodology was particularly influenced by the submarine detection research work on placement of undersea surveillance assets. We recognized that similar mathematical optimization techniques could be adapted to address the complex problem of urban sensor grid deployment, where building density, RF propagation characteristics, and overlapping coverage areas create significant placement challenges.
In discussions with operators from USASOC and AFSOC, we learned about how the deployment of a sensing grid around an objective can be complex and dangerous. Operators have limited resources and can face extreme risk when operating in hostile environments.
We developed SPOT to have a dual use application that can be used by the city of Boston and USSOCOM. The following two use cases are examples of how SPOT can be utilized to support decisions of resource allocation in sensor placement.
Use Case #1: Boston Marathon Finish Surveillance
The Boston Marathon bombing in 2013 occurred when surveillance and security of the race’s finish line were not prioritized. Given budget constraints, difficulties with field of view, and limited limited personnel to monitor the camera feeds, where should security personnel place cameras for situational awareness?
Use Case #2: SOF Sensing Grid Deployment
Special Operators work in a much more constrained and hostile environment than the city of Boston. For example, a high value target (HVT) mission entails establishing a perimeter around an objective, deploying a sensing grid, and monitoring the information coming from the environment. Operators face significant risk when deploying a sensing grid around the objective, especially when an adversary wants to control the area. Given an objective area, limited personnel, and risk associated with deployment, where should operators place sensors to support their mission?
In both of these use cases, SPOT is a framework that provides decision support for the allocation of sensor resources.
What it does
This proof of concept model does the following: divides an area of responsibility (AOR) into a grid of discrete cells and determines binary coverage for each cell based on camera placement. The model uses simple geospatial relationships to establish whether each grid cell is visible to a camera placed at a specific location. It provides a basic framework for optimizing camera placement by identifying which combination of locations can cover the maximum number of cells with a limited number of assets. This approach works equally well for city planners evaluating security camera networks and for Special Operations Forces determining optimal surveillance positions in urban environments. While basic in its implementation, the model demonstrates the fundamental principle of coverage optimization and provides the foundation for more sophisticated sensor placement algorithms. By visualizing which areas remain uncovered with different camera configurations, users can quickly identify gaps in surveillance coverage and adjust accordingly.
How we built it
We built this proof of concept model by:
Creating a geospatial grid network representing the urban environment Establishing binary decision variables for camera placement locations Developing a coverage matrix identifying which camera positions can observe which grid cells Attaining granular crime data for the neighborhood of Back Bay Simulated danger/priority weights to different areas based on importance Implementing an integer programming optimization model using Gurobi Setting constraints on the maximum number of available cameras Formulating an objective function to maximize coverage of high-priority areas Solving the optimization problem to determine optimal camera placement Visualizing results through network graphs showing camera positions and coverage areas The model's simplicity allows for quick implementations and iterations while demonstrating the core concept of optimized sensor placement.
Financial Feasibility:
In business, financial justification is essential for project execution. While SPOT is a promising concept, its profitability must be demonstrated. In 2023, the Air Force and AFWERX awarded a similar contract for a virtual sensor optimization project. To minimize SPOT’s project risk, conservative assumptions were made, including a high discount rate of 10% and a net working capital amount of $750,000. Costs are detailed in the attached pdf, with straight-line depreciation applied over a 10-year lifespan. The project is expected to be funded through an SBIR/STTR contract, which pays out in incremental phases. Since equity financing is used, neither the WACC nor the cost of debt is considered in the analysis. Key financial metrics include a net present value (NPV) of $997,426.75 and an internal rate of return (IRR) of 53.96%. NPV represents the difference between the present value of cash inflows and outflows, while IRR indicates the breakeven rate of return. Higher values for both metrics suggest a more attractive investment. Although contract specifications can be uncertain, similar ventures have demonstrated profitability. When combined with our high positive NPV and IRR, this provides strong financial justification for SPOT.
Challenges we ran into
During our model development, we encountered challenges related to risk assessment granularity. Urban environments contain numerous risk factors—crime patterns, crowd densities, infrastructure vulnerabilities, temporal variations—all of which influence optimal sensor placement. Our model required quantifiable risk values for each grid cell, but the available data lacked sufficient detail to capture these nuances effectively. This wasn't a fundamental limitation of our approach but rather a data access issue; with higher-resolution inputs from city databases, incident reports, or intelligence sources, our grid-based model could incorporate sophisticated risk weightings. The framework itself is sound, but its effectiveness directly correlates with input data quality. Given more comprehensive information, our optimization algorithm could readily incorporate these factors to produce significantly more targeted placement recommendations. To get around this, we simply assumed distributions to show the framework would work given the correct data.
A major technical limitation was our inability to incorporate line-of-sight (LOS) constraints. Though we recognized that buildings and other structures significantly impact surveillance coverage in urban environments, integration issues prevented us from incorporating elevation data and 3D obstruction analysis. Properly modeling LOS would require detailed building footprints, heights, and terrain models to determine whether a camera at one location could actually observe a target area without obstruction. This capability would substantially improve the realism of our camera placement model.
Determining which sensor types would be most applicable presented another challenge. Urban environments generate diverse data streams requiring different sensor modalities (optical, infrared, acoustic, etc.). Our model simplistically assumed camera-based surveillance, but a comprehensive system would need to consider various sensor types, their coverage characteristics, and integration challenges across heterogeneous networks.
Finally, balancing the dual-use requirements for both SOCOM operations and City of Boston planning created interesting constraints. While military applications prioritize covert deployment and high-risk area monitoring, urban planning focuses more on public safety and infrastructure management. Finding common ground between these stakeholders required careful consideration of shared priorities while recognizing their distinct operational requirements and privacy concerns. Despite these challenges, this alignment of interests created valuable opportunities for knowledge sharing and resource optimization.
Accomplishments that we're proud of
We successfully developed a working mathematical optimization model that solves the complex problem of sensor placement with limited resources. Our approach efficiently balances coverage maximization with resource constraints through integer programming, providing solutions that identify the most strategically valuable camera positions. The visualization component clearly illustrates coverage patterns, making results interpretable for both technical and non-technical stakeholders. We're particularly proud of creating a dual-purpose framework applicable to both special operations planning and civilian urban management, demonstrating how similar mathematical principles can address seemingly different domains. The model's flexibility allows for easy adjustment of parameters like camera quantity and priority weightings without requiring code restructuring. Additionally, we achieved computational efficiency—our optimization runs complete within reasonable time frames even for large urban areas split into hundreds of grid cells. Finally, we built the system using industry-standard tools and libraries, ensuring maintainability and potential for future expansion as more sophisticated data becomes available.
What we learned
We first learned the importance of not being discouraged by inexperience. At first, we were nervous that we would not be able to deliver a product that was relevant to the CodeMetal Hackathon since we had little experience with hardware. We then reflected on what we did know, and pivoted towards an optimization and finance approach.
We also learned about the importance of always keeping the “end user” in mind during product development. We can come up with a concept that sounds cool on paper, but unless it increases the intended effects of a SOF operator or Boston emergency responder for their specific use case, the concept will not add any value to the mission.
What's next for SPOT: Sensor Placement Optimization Tool
The next step for spot is to actually implement the recommended sensor placement of an objective area and seeing how it improved upon a baseline. If we had more time, we would continue to refine the optimization formulation so that it could be tailored to a wider variety of use cases. We want operators to be able to specify parameters of the model so that its output is more applicable to their mission.
Log in or sign up for Devpost to join the conversation.