Detection, Tracking, and Identification of Asteroids through On-board Image Analysis

Our research objective is to develop effective, efficient, autonomous asteroid detection, tracking, and identification algorithms that can be hosted on-board a spacecraft. This includes explicitly designing, optimizing, and testing the candidate algorithms on selected flight heritage avionics that the team has identified. The elements of NASA’s space technology roadmaps that are being addressed include the following:
•TA04.5: Autonomy This research objective directly applies to Autonomy, as the goal is to develop algorithms that can identify and track asteroids without human intervention. 
•TA05.4: Position, Navigation and Timing. While this research objective is not intended directly for on-board autonomous navigation and maneuvering systems, the algorithms used to detect, identify and track asteroids can also be utilized to autonomously identify small bodies for hazard avoidance or for identifying small bodies (while they are still point sources) during the early stages of optical navigation. 
•TA11.1: Computing Understanding and optimizing algorithms for multi-core processors and other radiation- hardened avionics is critical to the success of deploying these algorithms.
•TA11.4: Information Processing Our research objective directly maps to the Intelligent Data Understanding portion of Technology Area 12. Our goal is to intelligently understand the image data to identify an asteroid, track its trajectory, and prioritize the data to be downlinked.
We have accomplished the tasks that we set out to achieve for this grant. Our first task was to determine the viability of a CubeSat like architecture for detection and identification of asteroids. We determined during the first quarter that a CubeSat architecture could not host a large enough aperture payload to acquire signals of smaller asteroids. 
•We then identified a baseline concept for a Data Processing Unit to host our on-board algorithms. 
•We obtained applicable time series imagery for use in our algorithmic development. 
•We simulated imagery for use in algorithmic development in addition to providing a mechanism for trade analysis on detector, telescope and mission concepts.
•We prototyped a few different algorithmic chains using MATLAB and tested them against ground-based data and NEOWISE in addition to simulated imagery.
•We prototyped a set of algorithms and benchmarked them on a flight-like processor, the MCP-750, demonstrating that on-board processing can be achieved.
•We identified a data acquisition and data flow concept for use of our on-board algorithms as a triage/data compression technique to maximize use of the limited downlink bandwidth without sacrificing the science data return. The chief critique of our algorithmic development is the lack of viable imagery obtained by existing telescopes. We obtained data as a result of the generosity of other organizations (shown in Table 1); however due to funding constraints, we did not pursue spending the grant funds on searching for ground-based imagery captured in a time sequence manner. We opted to demonstrate the feasibility of our approach on a proven spacecraft processor (RAD750 surrogate: the MCP750), which we successfully accomplished.