Results for “decision making under deep uncertainty”

NASA NTRSDate not supplied

Uncertainty Quantification using Deep Ensembles for Decision Making in Cyber-Physical-Human Systems

In this paper and its companion, Differential Equation Approximation Using Gradient-Boosted Quantile Regression, Robison et al., we examine an approach to quantifying model uncertainty with the aim of increasing the trustworthiness of computational models in human-machine interactions. In Differential Equation Approximation Using Gradient-Boosted Quantile Regression, we focus on gradient-boosted decision trees, while in this one, we give more details about deep ensembles. Uncertainty quantification is crucial for building trustworthy autonomous decision-making agents in human-machine teams. There are two types of uncertainties: aleatoric and epistemic. The former is related to the inherent stochasticity (noise) of the process, whereas the latter is associated with the lack of knowledge or representation capability of models, such as neural networks. By lack of knowledge, we mean the model’s inability to accurately predict outputs for all possible inputs. The aleatory uncertainty can be estimated fairly easily with, for example, filters, whereas epistemic uncertainty is challenging to compute. This paper uses deep ensembles to quantify both aleatory and epistemic uncertainty. It can act as an uncertainty-aware surrogate transition model for decision-making frameworks. "Uncertainty-aware" means that the surrogate transition model should make predictions along with confidence in those predictions. In the context of decision-making, the transition models are ordinary differential equations (ODEs). Since ODEs can be simulated to make one-step or multi-step predictions, a good surrogate model for them should perform reasonably well in both modes. In a multi-step approach, the trajectory sampling method TS∞ was used to propagate uncertainty over multiple steps. The cartpole dynamical system was selected to demonstrate the ability of deep ensembles as good surrogate transition models for decision-making frameworks. The deep ensembles modeled the dynamics of cartpole ODEs and made uncertainty-aware predictions in single-step and multi-step transition modes.

CPH systems↗

NASA NTRSDate not supplied

Collision Avoidance Approach Using Deep Reinforcement Learning

A method to enable autonomous robots moving in a 2D space collision free motivates the purposed approach for collision avoidance for autonomous UAM vehicles. Challenges of autonomous collision free navigation for both problems are similar. Agents in each environment do not know the intent, or goal, of the other. Finding the time efficient paths require some level of anticipation with neighboring agents which is computationally expensive. In the original work, these obstacles were overcome with a novel application of deep reinforcement learning which offloads the online computation to an offline learning algorithm. A value network that encodes the estimated time to the goal given the agent’s state and the observable portion of the other agent’s state is trained on a baseline policy and further refined with reinforcement learning to promote time efficient collision free navigation. Online, the value network efficiently informs the agent’s decision making in the face of uncertainty of the other agent’s next move. In this paper, challenges extending this methodology to the 3D environment of autonomous UAM vehicles with kinematic constraints are discussed and initial results shown.

Collision Avoidance↗

NASA NTRSDate not supplied

Collision Avoidance Approach Using Deep Reinforcement Learning

A method to enable autonomous robots moving in a 2D space collision free motivates the purposed approach for collision avoidance for autonomous UAM vehicles. Challenges of autonomous collision free navigation for both problems are similar. Agents in each environment do not know the intent, or goal, of the other. Finding the time efficient paths require some level of anticipation with neighboring agents which is computationally expensive. In the original work, these obstacles were overcome with a novel application of deep reinforcement learning which offloads the online computation to an offline learning algorithm. A value network that encodes the estimated time to the goal given the agent’s state and the observable portion of the other agent’s state is trained on a baseline policy and further refined with reinforcement learning to promote time efficient collision free navigation. Online, the value network efficiently informs the agent’s decision making in the face of uncertainty of the other agent’s next move. In this paper, challenges extending this methodology to the 3D environment of autonomous UAM vehicles with kinematic constraints are discussed and initial results shown.

Collision Avoidance↗

NASA NTRSDate not supplied

Model Based Approaches for Fault Detection, Prognostics, Decision Making in Complex Systems

The presentation discusses application of model based approaches to complex systems. The model is composed of physics-derived and empirical equations, integrated with connected networks that are strategically placed within the model to substitute equations that are subject to large uncertainty. Polynomial fit driven by heuristics or empirical observations can be substituted by more flexible networks that can minimize the error between model predictions and observations without being restricted to a predefined functional form. This modeling strategy allows training of networks deep inside the model and unknown parameters in a single learning stage.

Physics Informed↗

NASA NTRS2020

Improving Trust in Deep Neural Networks with Nearest Neighbors

Deep neural networks are used increasingly for perception and decision-making in UAVs. For example, they can be used to recognize objects from images and decide what actions the vehicle should take. While deep neural networks can perform very well at complex tasks, their decisions may be unintuitive to a human operator. When a human disagrees with a neural network prediction, due to the black box nature of deep neural networks, it can be unclear whether the system knows something the human does not or whether the system is malfunctioning. This uncertainty is problematic when it comes to ensuring safety. As a result, it is important to develop technologies for explaining neural network decisions for trust and safety. This paper explores a modification to the deep neural network classification layer to produce both a predicted label and an explanation to support its prediction. Specifically, at test time, we replace the final output layer of the neural network classifier by a k-nearest neighbor classifier. The nearest neighbor classifier produces 1) a predicted label through voting and 2) the nearest neighbors involved in the prediction, which represent the most similar examples from the training dataset. Because prediction and explanation are derived from the same underlying process, this approach guarantees that the explanations are always relevant to the predictions. We demonstrate the approach on a convolutional neural network for a UAV image classification task. We perform experiments using a forest trail image dataset and show empirically that the hybrid classifier can produce intuitive explanations without loss of predictive performance compared to the original neural network. We also show how the approach can be used to help identify potential issues in the network and training process.

Lee, Ritchie↗

NASA NTRS2017

Multi-Objective Reinforcement Learning-Based Deep Neural Networks for Cognitive Space Communications

Future communication subsystems of space exploration missions can potentially benefit from software-defined radios (SDRs) controlled by machine learning algorithms. In this paper, we propose a novel hybrid radio resource allocation management control algorithm that integrates multi-objective reinforcement learning and deep artificial neural networks. The objective is to efficiently manage communications system resources by monitoring performance functions with common dependent variables that result in conflicting goals. The uncertainty in the performance of thousands of different possible combinations of radio parameters makes the trade-off between exploration and exploitation in reinforcement learning (RL) much more challenging for future critical space-based missions. Thus, the system should spend as little time as possible on exploring actions, and whenever it explores an action, it should perform at acceptable levels most of the time. The proposed approach enables on-line learning by interactions with the environment and restricts poor resource allocation performance through virtual environment exploration. Improvements in the multiobjective performance can be achieved via transmitter parameter adaptation on a packet-basis, with poorly predicted performance promptly resulting in rejected decisions. Simulations presented in this work considered the DVB-S2 standard adaptive transmitter parameters and additional ones expected to be present in future adaptive radio systems. Performance results are provided by analysis of the proposed hybrid algorithm when operating across a satellite communication channel from Earth to GEO orbit during clear sky conditions. The proposed approach constitutes part of the core cognitive engine proof-of-concept to be delivered to the NASA Glenn Research Center SCaN Testbed located onboard the International Space Station.

space archtiecture↗

NASA NTRS2017

Multi-Objective Reinforcement Learning-based Deep Neural Networks for Cognitive Space Communications

Future communication subsystems of space exploration missions can potentially benefit from software-defined radios (SDRs) controlled by machine learning algorithms. In this paper, we propose a novel hybrid radio resource allocation management control algorithm that integrates multi-objective reinforcement learning and deep artificial neural networks. The objective is to efficiently manage communications system resources by monitoring performance functions with common dependent variables that result in conflicting goals. The uncertainty in the performance of thousands of different possible combinations of radio parameters makes the trade-off between exploration and exploitation in reinforcement learning (RL) much more challenging for future critical space-based missions. Thus, the system should spend as little time as possible on exploring actions, and whenever it explores an action, it should perform at acceptable levels most of the time. The proposed approach enables on-line learning by interactions with the environment and restricts poor resource allocation performance through virtual environment exploration. Improvements in the multiobjective performance can be achieved via transmitter parameter adaptation on a packet-basis, with poorly predicted performance promptly resulting in rejected decisions. Simulations presented in this work considered the DVB-S2 standard adaptive transmitter parameters and additional ones expected to be present in future adaptive radio systems. Performance results are provided by analysis of the proposed hybrid algorithm when operating across a satellite communication channel from Earth to GEO orbit during clear sky conditions. The proposed approach constitutes part of the core cognitive engine proof-of-concept to be delivered to the NASA Glenn Research Center SCaN Testbed located onboard the International Space Station.

space archtiecture↗

NASA NTRS1997

Assessment and Management of the Risks of Debris Hits During Space Station EVAs

The risk of EVAs is critical to the decision of whether or not to automate a large part of the construction of the International Space Station (ISS). Furthermore, the choice of the technologies of the space suit and the life support system will determine (1) the immediate safety of these operations, and (2) the long-run costs and risks of human presence in space, not only in lower orbit (as is the case of the ISS) but also perhaps, outside these orbits, or on the surface of other planets. The problem is therefore both an immediate one and a long-term one. The fundamental question is how and when to shift from the existing EMU system (suit, helmet, gloves and life support system) to another type (e.g. a hard suit), given the potential trade-offs among life-cycle costs, risks to the astronauts, performance of tasks, and uncertainties about new systems' safety inherent to such a shift in technology. A more immediate issue is how to manage the risks of EVAs during the construction and operation of the ISS in order to make the astronauts (in the words of the NASA Administrator) "as safe outside as inside". For the moment (June 1997), the plan is to construct the Space Station using the low-pressure space suits that have been developed for the space shuttle. In the following, we will refer to this suit assembly as EMU (External Maneuvering Unit). It is the product of a long evolution, starting from the U.S. Air Force pilot suits through the various versions and changes that occurred for the purpose of NASA space exploration, in particular during the Gemini and the Apollo programs. The Shuttle EMU is composed of both soft fabrics and hard plates. As an alternative to the shuttle suit, at least two hard suits were developed by NASA: the AX5 and the MRKIII. The problem of producing hard suits for space exploration is very similar to that of producing deep-sea diving suits. There was thus an opportunity to develop a suit that could be manufactured for both purposes with the economies of scale that could be gained from a two-branch manufacturing line (space and deep sea). Of course, the space suit would need to be space qualified. Some of the problems in adopting one of the hard suits were first that the testing had to be completed, and second that it required additional storage space. The decision was made not to develop a hard suit in time for the construction and operation of the ISS. Instead, to improve the safety of the current suit, it was decided to reinforce the soft parts of the shuttle EMU with KEVLAR linings to strengthen it against debris impacts. Test results, however, show that this advanced suit design has little effect on the penetration characteristics.

Pate-Cornell, Elisabeth↗

NASA NTRSDate not supplied

Early Information Parameter-Set Analysis for Satellite Close Approaches using Machine Learning

Understanding orbital mechanics is essential in space flight and navigation applications, and leveraging modern force models for flight path projection remains an important aspect in space mission design and operation. However, force models do not capture all the dynamics or perturbations in the space environment and thus are subject to errors in predicting the state vectors. The further out the predicted miss distance between spacecraft is from the time of closest approach (TCA), the larger the propagated errors in the predicted miss distance at TCA is. The dependency on these force models for spacecraft flight state prediction calls for a more reliable method that can quantify, or even reduce, these propagated errors. With recent advances in the field artificial intelligence, specifically in machine and deep learning algorithms, a model that implements these approaches can improve on the modern force model approach. The goal for this work is to provide an early-information decision-making threshold, in order to prioritize risk assessment implementation, given the ongoing increase of space objects. In analyzing the relationship of several parameters from conjunction data messages(CDMs) and solar information, early information becomes viable in miss distance prediction with unsupervised learning techniques, which learn the parameters that are linked together with miss distance and probability of collision (Pc) variables. Another approach implemented for identifying relationships within CDMs is supervised learning, in which a shallow neural network binary classifier learns to distinguish events with Pc values¡108. These parameters detected in the unsupervised process are then applied to a regression neural network, which predicts the miss distance at TCA for a specific event within a given uncertainty bound. For the regression neural network, a Long Short Term Memory (LSTM) neural network is implemented, which yields memory about each time step in an event. Using an LSTM network, the model learns to predict miss distance within 0.2km of the value measured at TCA. Although there is a limited amount of "close miss" data to train a network, the network learns to associate parameters, like large energy dissipation rates with the secondary object, with an elevated Pc

Brianna I. Robertson↗

NASA NTRS2008

The NASA Energy and Water Cycle Extreme (NEWSE) Integration Project

Skillful predictions of water and energy cycle extremes (flood and drought) are elusive. To better understand the mechanisms responsible for water and energy extremes, and to make decisive progress in predicting these extremes, the collaborative NASA Energy and Water cycle Extremes (NEWSE) Integration Project, is studying these extremes in the U.S. Southern Great Plains (SGP) during 2006-2007, including their relationships with continental and global scale processes, and assessment of their predictability on multiple space and time scales. It is our hypothesis that an integrative analysis of observed extremes which reflects the current understanding of the role of SST and soil moisture variability influences on atmospheric heating and forcing of planetary waves, incorporating recently available global and regional hydro- meteorological datasets (i.e., precipitation, water vapor, clouds, etc.) in conjunction with advances in data assimilation, can lead to new insights into the factors that lead to persistent drought and flooding. We will show initial results of this project, whose goals are to provide an improved definition, attribution and prediction on sub-seasonal to interannual time scales, improved understanding of the mechanisms of decadal drought and its predictability, including the impacts of SST variability and deep soil moisture variability, and improved monitoring/attributions, with transition to applications; a bridging of the gap between hydrological forecasts and stakeholders (utilization of probabilistic forecasts, education, forecast interpretation for different sectors, assessment of uncertainties for different sectors, etc.).

House, P. R.↗