Orbit Logic (United States)

Space exploration missions to the moon and Mars cannot count on having a surgeon physically present to perform surgical procedures. Furthermore, telesurgery is not feasible at these distances due to network latency ranging from 3 seconds to several minutes. To address this need, a robotic surgical system can be trained by expert surgeons prospectively so that a health-preserving or life-saving procedure can be performed semi-autonomously when needed. The approach of path generation using artificial neural networks allows for an effective and scalable solution for the supervised learning and real-time performance of a surgical procedure. This study makes use of long short-term memory (LSTM) recurrent neural networks (RNNs) in conjunction with Evolino or back propagation learning algorithms for end-effector path optimization. The RNN-generated path is trained from human-performed procedures in a simulated environment. Changes in movement of path markers are accounted for by adjusting the end-effector acceleration with respect to target markers along the path. Results include smooth generated paths successfully meeting the defined procedure requirements of accuracy and speed in both static and dynamic environments.

A key requirement for the current generation of artificial decision-makers is that they should adapt well to changes in unexpected situations. This paper addresses the situation in which an AI for aerial dog fighting, with tunable parameters that govern its behavior, must optimize behavior with respect to an objective function that is evaluated and learned through simulations. Bayesian optimization with a Gaussian Process surrogate is used as the method for investigating the objective function. One key benefit is that during optimization, the Gaussian Process learns a global estimate of the true objective function, with predicted outcomes and a statistical measure of confidence in areas that haven't been investigated yet. Having a model of the objective function is important for being able to understand possible outcomes in the decision space; for example this is crucial for training and providing feedback to human pilots. However, standard Bayesian optimization does not perform consistently or provide an accurate Gaussian Process surrogate function for highly volatile objective functions. We treat these problems by introducing a novel sampling technique called Hybrid Repeat/Multi-point Sampling. This technique gives the AI ability to learn optimum behaviors in a highly uncertain environment. More importantly, it not only improves the reliability of the optimization, but also creates a better model of the entire objective surface. With this improved model the agent is equipped to more accurately/efficiently predict performance in unexplored scenarios.

Commercial imaging satellites must respond to thousands of requests to image different areas of the earth within tight time constraints. To effectively satisfy and manage these requests, geodynamic, camera, and spacecraft constraints must be reconciled within a matter of minutes to create a workable plan for capturing images during a particular spacecraft pass. These constraints include customer-defined target viewing constraints, spacecraft power limitations, data storage attributes, pointing constraints, spacecraft agility limitations, high gain antenna blockages, cloud cover and more. Orbit Logic developed and delivered a high fidelity spacecraft simulator that is tightly integrated with scheduling optimization algorithms to meet this challenge. Dubbed the OrbView Tasking System (OTS), it is used today by GeoEye for OrbView-3 image collection planning. OTS incorporates spacecraft slew, scan, and other models, reconciles all constraints and efficiently generates valid collection plans using multiple optimization algorithms and/or manual planning inputs. It also incorporates innovative visualization capabilities for intuitive assessment of collection plans. This system is being upgraded for use with GeoEye-1, in addition to OrbView-3, and will be incorporated into a new system called the Collection Planning System (CPS). CPS includes a planning server that will manage multiple instances of OTS for increased computational performance and system scalability. This paper will describe the OTS and the overall approach to optimizing collection planning for OrbView-3 as well as the system's overall applicability for use with any imagery, radar, or remote sensing mission.