SEARCH · Search NASA

Results for “partially observable Markov decision process”

Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.

SourceResearch connectionFrom yearThrough year

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

14 records

All sources NASA NTRS DOE OSTI

DOE OSTI2021

Distributionally Robust Partially Observable Markov Decision Process with Moment-Based Ambiguity

In this paper, we consider a distributionally robust partially observable Markov decision process (DR-POMDP), where the distribution of the transition-observation probabilities is unknown at the beginning of each decision period, but their realizations can be inferred using side information at the end of each period after an action being taken. We build an ambiguity set of the joint distribution using bounded moments via conic constraints and seek an optimal policy to maximize the worst-case (minimum) reward for any distribution in the set. We show that the value function of DR-POMDP is piecewise linear convex with respect to the belief state and propose a heuristic search value iteration method for obtaining lower and upper bounds of the value function. We conduct numerical studies and demonstrate the computational performance of our approach via testing instances of a dynamic epidemic control problem. Our results show that DR-POMDP can produce more robust policies under misspecified distributions of transition-observation probabilities as compared to POMDP but has less costly solutions than robust POMDP. The DR-POMDP policies are also insensitive to varying parameter in the ambiguity set and to noise added to the true transition-observation probability values obtained at the end of each decision period.

97 MATHEMATICS AND COMPUTING↗

NASA NTRS1999

Cooperation and Coordination Between Fuzzy Reinforcement Learning Agents in Continuous State Partially Observable Markov Decision Processes

Successful operations of future multi-agent intelligent systems require efficient cooperation schemes between agents sharing learning experiences. We consider a pseudo-realistic world in which one or more opportunities appear and disappear in random locations. Agents use fuzzy reinforcement learning to learn which opportunities are most worthy of pursuing based on their promise rewards, expected lifetimes, path lengths and expected path costs. We show that this world is partially observable because the history of an agent influences the distribution of its future states. We consider a cooperation mechanism in which agents share experience by using and-updating one joint behavior policy. We also implement a coordination mechanism for allocating opportunities to different agents in the same world. Our results demonstrate that K cooperative agents each learning in a separate world over N time steps outperform K independent agents each learning in a separate world over K*N time steps, with this result becoming more pronounced as the degree of partial observability in the environment increases. We also show that cooperation between agents learning in the same world decreases performance with respect to independent agents. Since cooperation reduces diversity between agents, we conclude that diversity is a key parameter in the trade off between maximizing utility from cooperation when diversity is low and maximizing utility from competitive coordination when diversity is high.

Berenji, Hamid R.↗

NASA NTRS2000

Back to the Future: Consistency-Based Trajectory Tracking

Given a model of a physical process and a sequence of commands and observations received over time, the task of an autonomous controller is to determine the likely states of the process and the actions required to move the process to a desired configuration. We introduce a representation and algorithms for incrementally generating approximate belief states for a restricted but relevant class of partially observable Markov decision processes with very large state spaces. The algorithm presented incrementally generates, rather than revises, an approximate belief state at any point by abstracting and summarizing segments of the likely trajectories of the process. This enables applications to efficiently maintain a partial belief state when it remains consistent with observations and revisit past assumptions about the process' evolution when the belief state is ruled out. The system presented has been implemented and results on examples from the domain of spacecraft control are presented.

Kurien, James↗

DOE OSTI2023

Bayesian sequential optimal experimental design for nonlinear models using policy gradient reinforcement learning

We present a mathematical framework and computational methods for optimally designing a finite sequence of experiments. This sequential optimal experimental design (sOED) problem is formulated as a finite-horizon partially observable Markov decision process (POMDP) under a Bayesian setting and with information-theoretic utilities. The formulation is general and may accommodate continuous random variables, non-Gaussian posteriors, and nonlinear forward models. The sOED design policy incorporates elements of feedback and lookahead simultaneously, and we show it to generalize the commonly-used batch and greedy design strategies. We solve for the sOED policy using the policy gradient (PG) method from reinforcement learning, and provide a derivation for the PG expression in the sOED context. Adopting an actor-critic approach, the policy and value functions are parameterized using deep neural networks and improved via PG estimates produced from simulated episodes of designs and observations. The new PG-sOED algorithm is first validated on a linear-Gaussian benchmark, and then compared against other design baselines on a sensor movement problem for contaminant source inversion in a convection-diffusion field. As a result, we provide explanation for the policy behaviors using knowledge of the underlying physical process.

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2022

Robotic Planning under Uncertainty in Spatiotemporal Environments in Expeditionary Science

In the expeditionary sciences, spatiotemporally varying environments -- hydrothermal plumes, algal blooms, lava flows, or animal migrations -- are ubiquitous. Mobile robots are uniquely well-suited to study these dynamic, mesoscale natural environments. We formalize expeditionary science as a sequential decision-making problem, modeled using the language of partially-observable Markov decision processes (POMDPs). Solving the expeditionary science POMDP under real-world constraints requires efficient probabilistic modeling and decision-making in problems with complex dynamics and observational models. Previous work in informative path planning, adaptive sampling, and experimental design have shown compelling results, largely in static environments, using data-driven models and information-based rewards. However, these methodologies do not trivially extend to expeditionary science in spatiotemporal environments: they generally do not make use of scientific knowledge such as equations of state dynamics, they focus on information gathering as opposed to scientific task execution, and they make use of decision-making approaches that scale poorly to large, continuous problems with long planning horizons and real-time operational constraints. In this work, we discuss these and other challenges related to probabilistic modeling and decision-making in expeditionary science, and present some of our preliminary work that addresses these gaps. We ground our results in a real expeditionary science deployment of an autonomous underwater vehicle (AUV) in the deep ocean for hydrothermal vent discovery and characterization. Our concluding thoughts highlight remaining work to be done, and the challenges that merit consideration by the reinforcement learning and decision-making community.

Preston, Victoria↗

NASA NTRS1998

Markov Tracking for Agent Coordination

Partially observable Markov decision processes (POMDPs) axe an attractive representation for representing agent behavior, since they capture uncertainty in both the agent's state and its actions. However, finding an optimal policy for POMDPs in general is computationally difficult. In this paper we present Markov Tracking, a restricted problem of coordinating actions with an agent or process represented as a POMDP Because the actions coordinate with the agent rather than influence its behavior, the optimal solution to this problem can be computed locally and quickly. We also demonstrate the use of the technique on sequential POMDPs, which can be used to model a behavior that follows a linear, acyclic trajectory through a series of states. By imposing a "windowing" restriction that restricts the number of possible alternatives considered at any moment to a fixed size, a coordinating action can be calculated in constant time, making this amenable to coordination with complex agents.

Washington, Richard↗

NASA NTRS1998

Making the Impossible Possible: Strategies for Fast POMDP Monitoring

Systems modeled as partially observable Markov decision processes (POMDPs) can be tracked quickly with three restrictions: all actions are grouped together, the out-degree of each system state is bounded by a constant, and the number of non-zero elements in the belief state is bounded by a (different) constant. With these restrictions, the tracking algorithm operates in constant time and linear space. The first restriction assumes that the action itself is unobservable. The second restriction defines a subclass of POMDPs that covers however a wide range of problems. The third restriction is an approximation technique that can lead to a potentially vexing problem: an observation may be received that has probability according to the restricted belief state. This problem of impossibility will cause the belief state to collapse. In this paper we discuss the tradeoffs between the constant bound on the belief state and the quality of the solution. We concentrate on strategies for overcoming the impossibility problem and demonstrate initial experimental results that indicate promising directions.

Washington, Richard↗

NASA NTRS2003

Optimal Limited Contingency Planning

For a given problem, the optimal Markov policy over a finite horizon is a conditional plan containing a potentially large number of branches. However, there are applications where it is desirable to strictly limit the number of decision points and branches in a plan. This raises the question of how one goes about finding optimal plans containing only a limited number of branches. In this paper, we present an any-time algorithm for optimal k-contingency planning. It is the first optimal algorithm for limited contingency planning that is not an explicit enumeration of possible contingent plans. By modelling the problem as a partially observable Markov decision process, it implements the Bellman optimality principle and prunes the solution space. We present experimental results of applying this algorithm to some simple test cases.

Meuleau, Nicolas↗

NASA NTRS2017

Dynamic Routing of Aircraft in the Presence of Adverse Weather Using a POMDP Framework

Each year weather-related airline delays result in hundreds of millions of dollars in additional fuel burn, maintenance, and lost revenue, not to mention passenger inconvenience. The current approaches for aircraft route planning in the presence of adverse weather still mainly rely on deterministic methods. In contrast, this work aims to deal with the problem using a Partially Observable Markov Decision Processes (POMDPs) framework, which allows for reasoning over uncertainty (including uncertainty in weather evolution over time) and results in solutions that are more robust to disruptions. The POMDP-based decision support system is demonstrated on several scenarios involving convective weather cells and is benchmarked against a deterministic planning system with functionality similar to those currently in use or under development.

Decision making↗

NASA NTRSDate not supplied

Adaptive Stress Testing: Using Reinforcement Learning to Find Failures in Safety-Critical Systems

Emerging applications in artificial intelligence, such as driverless cars and autonomous aircraft promise to be more efficient, cheaper to operate, and always available. However, ensuring the safety of these systems remains a major challenge to their certification and adoption. These autonomous systems are expected to routinely make safety-critical decisions where failures can have serious consequences including loss of life and property. Testing and validation techniques aim to identify and diagnose potential failures before the system is deployed. However, finding failure scenarios in autonomous systems can be very challenging due to high-dimensional and continuous state spaces, interaction with large environments over many time steps, and the rarity of failures. This talk presents Adaptive Stress Testing (AST), a simulation-based testing framework for finding the most likely path to a failure event of a safety-critical system. The key idea of AST is that stress testing can be formulated as a Partially Observable Markov Decision Process (POMDP), which enables reinforcement learning techniques to be used for finding failure events. Reinforcement learning algorithms can efficiently explore the search space and have been shown to scale to very large systems. We present applications of AST to find failures in various safety-critical systems including the aircraft collision avoidance systems, autonomous cars, and small unmanned aerial vehicles.

autonomous vehicles↗

DOE OSTI2022

Reinforcement learning for adaptive maintenance policy optimization under imperfect knowledge of the system degradation model and partial observability of system states

Maintenance policy optimization usually is faced with challenges that arise from an imperfect knowledge of system degradation models and from the partial observability of system degradation states. Here, this paper proposes a reinforcement learning method to address these two challenges for a class of maintenance problems with Markov degradation processes. The reinforcement learning approach consists of a learning component and a planning component. Using sequentially collected observations, at each step of decision-making the learning component improves the knowledge of system degradation in terms of the probability distributions of the transition rates based on sequential Bayesian inference. Using the updated transition rates, at each step of decision-making the maintenance policy optimization problem is then formulated as a partially observable Markov decision problem, and the planning component computes the optimal maintenance policy that maximizes the expected cumulative reward. The proposed method is illustrated using a numerical example with repair and inspection maintenance actions. The result shows that as more observations are collected, the learning component progressively learns the true system degradation process, and the planning component adjusts the optimal maintenance policy accordingly as well, which leads to increased reward.

42 ENGINEERING↗

NASA NTRS2007

Learning Grasp Strategies Composed of Contact Relative Motions

Of central importance to grasp synthesis algorithms are the assumptions made about the object to be grasped and the sensory information that is available. Many approaches avoid the issue of sensing entirely by assuming that complete information is available. In contrast, this paper proposes an approach to grasp synthesis expressed in terms of units of control that simultaneously change the contact configuration and sense information about the object and the relative manipulator-object pose. These units of control, known as contact relative motions (CRMs), allow the grasp synthesis problem to be recast as an optimal control problem where the goal is to find a strategy for executing CRMs that leads to a grasp in the shortest number of steps. An experiment is described that uses Robonaut, the NASA-JSC space humanoid, to show that CRMs are a viable means of synthesizing grasps. However, because of the limited amount of information that a single CRM can sense, the optimal control problem may be partially observable. This paper proposes expressing the problem as a k-order Markov Decision Process (MDP) and solving it using Reinforcement Learning. This approach is tested in a simulation of a two-contact manipulator that learns to grasp an object. Grasp strategies learned in simulation are tested on the physical Robonaut platform and found to lead to grasp configurations consistently.

Platt, Robert, Jr.↗

NASA NTRS2020

Adaptive Stress Testing: Finding Likely Failure Events with Reinforcement Learning

Finding the most likely path to a set of failure states is important to the analysis of safety-critical systems that operate over a sequence of time steps, such as aircraft collision avoidance systems and autonomous cars. In many applications such as autonomous driving, failures cannot be completely eliminated due to the complex stochastic environment in which the system operates.As a result, safety validation is not only concerned about whether a failure can occur, but also discovering which failures are most likely to occur. This article presents adaptive stress testing (AST), a framework for finding the most likely path to a failure event in simulation. We consider a general black box setting for partially observable and continuous-valued systems operating in an environment with stochastic disturbances. We formulate the problem as a Markov decision process and use reinforcement learning to optimize it. The approach is simulation-based and does not require internal knowledge of the system, making it suitable for black-box testing of large systems. We present different formulations depending on whether the state is fully observable or partially observable. In the latter case, we present a modified Monte Carlo tree search algorithm that only requires access to the pseudorandom number generator of the simulator to overcome partial observability. We also present an extension of the framework, called differential adaptive stress testing (DAST), that can find failures that occur in one system but not in another. This type of differential analysis is useful in applications such as regression testing, where we are concerned with finding areas of relative weakness compared to a baseline. We demonstrate the effectiveness of the approach on an aircraft collision avoidance application, where a prototype aircraft collision avoidance system is stress tested to find the most likely scenarios of near mid-air collision.

Verification and Validation↗

NASA NTRS2003

Hybrid Discrete-Continuous Markov Decision Processes

This paper proposes a Markov decision process (MDP) model that features both discrete and continuous state variables. We extend previous work by Boyan and Littman on the mono-dimensional time-dependent MDP to multiple dimensions. We present the principle of lazy discretization, and piecewise constant and linear approximations of the model. Having to deal with several continuous dimensions raises several new problems that require new solutions. In the (piecewise) linear case, we use techniques from partially- observable MDPs (POMDPS) to represent value functions as sets of linear functions attached to different partitions of the state space.

Feng, Zhengzhu↗