Tradeoff Analysis of Space-Based Imaging Algorithms for Distributed Area Collection

Multi-satellite planning for space-based imaging has received increasing interest in recent years, not only in the optimality and computational complexity of various planning algorithms, but also in the area of distributed or swarm planning for satellite constellations. One challenge of space-based imaging is that collecting images over a large area can require multiple passes of the constellation. Collections also gather data on overlapping areas resulting in many combinations of collects to fulfill an area of interest. Ground-based centralized planning algorithms are able to make optimal plans and have fewer limitations on computational resources but lack robustness to loss of communication and face more latency with re-planning. Onboard distributed planning can adapt to changing scenarios at the cost of computational resources and the optimality of planning solutions. The choice between centralized, decentralized, and distributed formulations as well as the selection of planning algorithm presents a complex set of tradeoffs when designing for specific mission goals. Our research builds on previous work by characterizing the tradeoffs between optimality, computation time, data transfer, robustness to loss of communication, and scalability of different planning algorithms for space-based imaging. Using a combination of simulation and real-world experiments, our research focuses on the following areas: (i) Validation and comparison of optimality and computational resources of previous planning algorithms. The earth observation satellite scheduling problem is NP-hard, meaning that it is computationally intractable to find the optimal solution in all but small cases. This has led to the development of a number of heuristics and approximation algorithms that can find near-optimal solutions in a reasonable amount of time. We investigate the optimality of previous planning algorithms by comparing them to a Mixed Integer Linear Program (MILP) approach which achieves the global optimum but does not scale well with problem dimension. We also compare the computation time of different planning algorithms on both ground and onboard systems to determine which algorithms are most suitable for real-time applications. (ii) Data transfer for centralized and decentralized planning configurations. Both centralized and decentralized planning algorithms need to know the state information of the constellation and their collection capabilities including slew rates and image sensor quality. This research characterizes the amount of data transfer required for various configurations and explores methods to reduce overall data transfer. (iii) Quantifying the robustness of different planning algorithms to losses in communication and the limitations of space-to-space and ground-to-space communication. (iv) Investigating the scalability of different planning algorithms for increasing constellation sizes. This research provides a better understanding of the tradeoffs between optimality, computation time, data transfer, robustness, and scalability of different planning algorithms for space-based imaging. Utilizing this information, we are able to integrate a lightweight, robust, distributed area image collection algorithm for satellite constellations into Auria Innovations’s onboard planning and scheduling architecture.