From the Publisher:
Probabilistic Reasoning in Intelligent Systems is a complete andaccessible account of the theoretical foundations and computational methods that underlie plausible reasoning under uncertainty. The author provides a coherent explication of probability as a language for reasoning with partial belief and offers a unifying perspective on other AI approaches to uncertainty, such as the Dempster-Shafer formalism, truth maintenance systems, and nonmonotonic logic. The author distinguishes syntactic and semantic approaches to uncertaintyand offers techniques, based on belief networks, that provide a mechanism for making semantics-based systems operational. Specifically, network-propagation techniques serve as a mechanism for combining the theoretical coherence of probability theory with modern demands of reasoning-systems technology: modular declarative inputs, conceptually meaningful inferences, and parallel distributed computation. Application areas include diagnosis, forecasting, image interpretation, multi-sensor fusion, decision support systems, plan recognition, planning, speech recognitionin short, almost every task requiring that conclusions be drawn from uncertain clues and incomplete information.
Probabilistic Reasoning in Intelligent Systems will be of special interest to scholars and researchers in AI, decision theory, statistics, logic, philosophy, cognitive psychology, and the management sciences. Professionals in the areas of knowledge-based systems, operations research, engineering, and statistics will find theoretical and computational tools of immediate practical use. The book can also be used as an excellent text for graduate-level courses in AI, operations research, or applied probability.

Probabilistic Reasoning in Intelligent Systems: Networks of Plausible Inference

Probability Distributions.- Linear Models for Regression.- Linear Models for Classification.- Neural Networks.- Kernel Methods.- Sparse Kernel Machines.- Graphical Models.- Mixture Models and EM.- Approximate Inference.- Sampling Methods.- Continuous Latent Variables.- Sequential Data.- Combining Models.

Pattern Recognition and Machine Learning

An Integrated Agent Architecture for Software Defined Radio. Rapid-prototype cognitive radio, CR1, was developed to apply these.The modern software defined radio has been called the heart of a cognitive radio. Cognitive radio: an integrated agent architecture for software defined radio. Http:bwrc.eecs.berkeley.eduResearchMCMACR White paper final1.pdf. The cognitive radio, built on a software-defined radio, assumes. Radio: An Integrated Agent Architecture for Software Defined Radio, Ph.D. The need for software-defined radios is underlined and the most important notions used for such. Mitola III, Cognitive radio: an integrated agent architecture for software defined radio, Ph.D. This results in the set-theoretic ontology of radio knowledge defined in the. Cognitive Radio An Integrated Agent Architecture for Software.This article first briefly reviews the basic concepts about cognitive radio CR. Cognitive Radio-An Integrated Agent Architecture for Software Defined Radio. Cognitive Radio RHMZ 2007. Software-defined radio SDR idea 1. Cognitive radio: An integrated agent architecture for software.Cognitive Radio SOFTWARE DEFINED RADIO, AND ADAPTIVE WIRELESS SYSTEMS2 Cognitive Networks. 3 Joseph Mitola III, Cognitive Radio: An Integrated Agent Architecture for Software Defined Radio Stockholm.

/pdf/cognitive-radio-an-integrated-agent-architecture-for-1v855bviue.pdf

Cognitive Radio An Integrated Agent Architecture for Software Defined Radio

Boolean satisfiability is probably the most studied of the combinatorial optimization/search problems. Significant effort has been devoted to trying to provide practical solutions to this problem for problem instances encountered in a range of applications in electronic design automation (EDA), as well as in artificial intelligence (AI). This study has culminated in the development of several SAT packages, both proprietary and in the public domain (e.g. GRASP, SATO) which find significant use in both research and industry. Most existing complete solvers are variants of the Davis-Putnam (DP) search algorithm. In this paper we describe the development of a new complete solver, Chaff which achieves significant performance gains through careful engineering of all aspects of the search-especially a particularly efficient implementation of Boolean constraint propagation (BCP) and a novel low overhead decision strategy. Chaff has been able to obtain one to two orders of magnitude performance improvement on difficult SAT benchmarks in comparison with other solvers (DP or otherwise), including GRASP and SATO.

/pdf/chaff-engineering-an-efficient-sat-solver-2ima1ekzuu.pdf

Chaff: engineering an efficient SAT solver

https://cs.ru.nl/~peterl/teaching/KeR/Theorist/reiteraij87.pdf

A theory of diagnosis from first principles

n complex concurrent critical systems, such as autonomousrobots, unmanned air vehicles, and space systems, every component is a potential point of failure. This is true not only of embedded systems but also of purely software systems such as distributed and cyber applications. Typical attempts to make such systemsmore robust and secure are both brittle and incomplete due to reliance onmanual identification of and solutions to potential failures such as by usingexception mechanisms. That is, the security is easily broken, and there aremany possible failure modes that are not handled. Failures may be rare eventsso it is less easy to test for good coverage of fault scenarios. Techniques thatexpand to handling component-level failures are very expensive to apply, yetare still quite brittle and incomplete. This is not because engineers are lazy—the sheer size and complexity of modern information systems overwhelms theattempts of engineers, and myriad methodologies, to systematically investi-gate, identify, and specify a response to all possible failures of a system.

Automatic recovery from software failure

This paper presents fast non-sampling based methods to assess the risk of trajectories for autonomous vehicles when probabilistic predictions of other agents' futures are generated by deep neural networks (DNNs). The presented methods address a wide range of representations for uncertain predictions including both Gaussian and non-Gaussian mixture models for predictions of both agent positions and controls. We show that the problem of risk assessment when Gaussian mixture models (GMMs) of agent positions are learned can be solved rapidly to arbitrary levels of accuracy with existing numerical methods. To address the problem of risk assessment for non-Gaussian mixture models of agent position, we propose finding upper bounds on risk using Chebyshev's Inequality and sums-of-squares (SOS) programming; they are both of interest as the former is much faster while the latter can be arbitrarily tight. These approaches only require statistical moments of agent positions to determine upper bounds on risk. To perform risk assessment when models are learned for agent controls as opposed to positions, we develop TreeRing, an algorithm analogous to tree search over the ring of polynomials that can be used to exactly propagate moments of control distributions into position distributions through nonlinear dynamics. The presented methods are demonstrated on realistic predictions from DNNs trained on the Argoverse and CARLA datasets and are shown to be effective for rapidly assessing the probability of low probability events.

http://www.roboticsproceedings.org/rss16/p089.pdf

Fast Risk Assessment for Autonomous Vehicles Using Learned Models of Agent Futures.

In the future webs of unmanned vehicles will act together to robustly achieve elaborate missions within uncertain environments. This web may be a distributed satellite system forming an interferometer, or may be a heterogenous set of rovers and blimps exploring Mars. We coordinate these systems by introducing a reactive model-based programming language (RMPL) that combines within a single unified representation the flexibility of embedded programming and reactive execution languages, and the deliberative reasoning power of temporal planners. To support fast mission planning as graph search, the KIRK planner compiles an RMPL program into a temporal plan network (TPN), similar to those used by temporal planners, but extended for symbolic constraints and decisions. To robustly coordinate air vehicle or rover maneuvers we combine the Kirk planning algorithm with randomized algorithms for kinodynamic path planning. Finally, we describe our Mars exploration testbed, including four RWI ATRV vehicles. 1 Model-based Programming The recent spread of advanced processing to embedded systems has created vehicles that execute complex missions with increasing levels of autonomy, in space, on land and in the air. These vehicles must respond to uncertain and often unforgiving environments, both with a fast response time and with a high assurance of first time success. The future looks to the creation of cooperative robotic networks. For example, giant space telescopes are being deployed that are composed of satellites carrying the telescope’s different optical components. These satellites act in concert to image planets around other stars, or unusual weather events on earth. In addition, the 2000 Mars Program Independent Assessment Team recommended an exploration architecture that adopts a more global view. For example, a heterogenous set of vehicles, such as orbiters, rovers and blimps might work in concert to identify and evaluate sites of greatest scientific interest. The creation of robotic networks cannot be supported by the current programming practice alone. Recent mission failures, such as the Mars Climate Orbiter and Polar Landers, highlight the challenge of creating highly capable vehicles within realistic budget limits. Due to cost constraints, spacecraft flight software teams often do not have time to think through all the plausible situations that might arise, encode the appropriate responses within their software and then validate that software with high assurance. To break through this barrier we need to invent a new programming paradigm. In this paper we advocate the creation of embedded, model-based programming languages that support the ability to specify global strategies for multivehicle coordination. First, we argue that the programmer should retain control for the overall success of a mission, by programming game plans and contingencies that in the programmer’s experience will ensure a high degree of success. The programmer should be able to program these game plans using features of the best embedded programming languages available. For example, reactive synchronous languages[5], like Esterel, Lustre and Signal, offer a rich set of constructs for interacting with sensors and actuators, for creating complex behaviors involving concurrency and preemption, and for modularizing these behaviors using all the standard encapsulation mechanisms. Model-based programming extends this style of reactive language with a minimal set of constructs neccessary to perform flexible mission coordination, while hiding its reasoning capabilities under the hood of the language’s interpreter or compiler. Second, we argue that model-based programming languages should focus on elevating the programmer’s thinking, by automating the process of reasoning about low-level system interactions. Many recent space mission failures, such as Mars Climate Orbiter and Mars Polar Lander, can be isolated to difficulties in reasoning through low-level system interactions. On the other hand, this limited form of reasoning and book keeping is the hallmark of computational methods. The interpreter or compiler of a model-based program reasons through these interactions using composable models of the system being controlled. We are developing a language, called the Reactive Model-Based Programming Language (RMPL), that supports four types of reasoning about system interactions: reasoning about contingencies, scheduling, inferring a system’s hidden state and controlling that state. This paper develops RMPL in the context of contingencies and scheduling, while [15], shows how RMPL is used to infer hidden state. Third, execution of these programs is a form of mission-level planning that searches through the options for the optimal strategy that operates within the time constraints. Vehicle movement is central to these missions, hence to achieve global optimality we unify mission-level planning with global path planning algorithms. In particular we build upon the rapidly exploring random tree (RRT) algorithm of La Valle[8]. In addition, these vehicles may operate in highly dynamic situations or situations where maneuvering is extremely tight. RMPL offers a middle ground between execution languages, like RAPS [4], and highly flexible, operator-based temporal planners,like HSTS [10]. RAPS offers the exception handling and concurrency mechanisms of embedded languages, while adding goal monitoring, nondeterministic choice and metric constraints. However, RAPS makes its decisions reactively, without addressing concerns of schedulability and threat resolution, and hence can fall into a failure state. RMPL incorporates the forward looking planning and scheduling abilities of modern temporal planners, but can substantially restrict the space of plans considered to possible threads of execution through the RMPL program. This speeds response and mitigates risk. To develop the model-based programming paradigm in the context of Mars exploration, we have developed a combination of simulation and ground-based testbeds, including a collection of four RWI ATRV rovers and a set of formation flying spacecraft, called Spheres, to be flown on space station. The paper begins by introducing a subset of RMPL that includes constructs from traditional reactive programming plus constructs for specifying contingencies and scheduling constraints. Second, we describe how Kirk, an RMPL-based planner/executive, compiles RMPL programs into temporal plan networks (TPN), which compactly represent all possible threads of execution of an RMPL program, and all resource constraints and conflicts between concurrent activities. Third, we present Kirk’s online planning algorithm for RMPL that “looks” by using network search algorithms to find threads of execution through the TPN that are temporally consistent. The result is a partially ordered temporal plan. Kirk then “leaps” by executing the plan using plan execution methods[12] developed for Remote Agent[11]. Fourth, we develop the unification of Kirk with randomized, kino-dynamic path planning algorithms. We present Kirk in the context of a simple Mars exploration mission segment. 2 Example: Cooperative Ex-

/pdf/model-based-reactive-programming-of-cooperative-vehicles-for-i98a113mff.pdf

Model-based Reactive Programming of Cooperative Vehicles for Mars Exploration

Multi-Modal Particle Filtering for Hybrid Systems with Autonomous Mode Transitions

Autonomous systems operating in real-world environments must plan, schedule, and execute missions while robustly adapting to uncertainty and disturbance. One way to mitigate the effect of uncertainty and disturbance is to dynamically schedule the plan online, through dispatchable execution. Dispatchable execution increases the efficiency of plan execution by introducing (1) a compiler that reduces a plan to a dispatchable form that enables real-time scheduling, and (2) a temporal plan dispatcher that schedules start times of activities (or controllable events) dynamically in response to disturbances. Previous work addresses efficient dispatchable execution of plans described as Simple Temporal Problems (STPs). While STPs have proven useful for many applications, Temporal Constraint Satisfaction Problems (TCSPs) provide a more rich language by introducing disjunctive constraints. However, previous approaches to dispatchable execution of disjunctive temporal plans are intractable for moderatelysized problems. The key contribution of this paper is an efficient algorithm for compiling and dynamically scheduling TCSPs. We present an incremental algorithm that compiles a TCSP to a compact representation, encoding the solution set in terms of the differences among solutions. We empirically demonstrate that this novel encoding reduces the space to encode the solution set by up to three orders of magnitude compared to prior art, and supports fast dynamic scheduling.

/pdf/fast-dynamic-scheduling-of-disjunctive-temporal-constraint-1chrxm1o3d.pdf

Brian C. Williams

Papers

Automatic recovery from software failure

Fast Risk Assessment for Autonomous Vehicles Using Learned Models of Agent Futures.

Model-based Reactive Programming of Cooperative Vehicles for Mars Exploration

Multi-Modal Particle Filtering for Hybrid Systems with Autonomous Mode Transitions

Fast Dynamic Scheduling of Disjunctive Temporal Constraint Networks through Incremental Compilation