Reasoning about reasoning in a meta-level architecture

In this paper we discuss reasoning about reasoning in a multiple agent scenario. We consider agents that are perfect reasoners, loyal, and that can take advantage of both the knowledge and ignorance of other agents. The knowledge representation formalism we use is (full) first order predicate calculus, where different agents are represented by different theories, and reasoning about reasoning is realized via a meta-level representation of knowledge and reasoning. The framework we provide is pretty general: we illustrate it by showing a machine checked solution to the three wisemen puzzle. The agents' knowledge is organized into units: the agent's own knowledge about the world and its knowledge about other agents are units containing object-level knowledge; a unit containing meta-level knowledge embodies the reasoning about reasoning and realizes the link among units. In the paper we illustrate the meta-level architecture we propose for problem solving in a multi-agent scenario; we discuss our approach in relation to the modal one and we compare it with other meta-level architectures based on logic. Finally, we look at a class of applications that can be effectively modeled by exploiting the meta-level approach to reasoning about knowledge and reasoning.     

Formalising a Value-Passing Calculus in HOL

Milner's value-passing calculus for describing and reasoning about communicating systems is formalised in the HOL proof assistant. Based on a previously defined mechanisation of pure CCS (no data communication, only synchronisation) in HOL, value-passing agents are given behavioural semantics by translating them into pure agents. An interactive proof environment is derived that supports both reasoning about the value-passing calculus and verification of value-passing specifications, which are defined over an infinite value domain. Received September 1997 / Accepted in revised form July 1999     

An embodied approach to arthropod animation

We describe a system for dynamically animating the locomotive behaviour of arthropods (insects, spiders and numerous other species) in real‐time, facilitating realistic and autonomous traversal across an arbitrary environment. By combining a decentralised reactive behavioural model with a hybrid approach to motion that utilises the comparative advantages of physical simulation and kinematic control, our system is capable of automatically generating complex organic motion over a wide range of surface features, independent of structural complexity. The reactive embodiment of the creature, combined with the physical simulation of the virtual world enables the formation of emergent behaviours that are entirely based on circumstance, including rigid‐body interaction, grip recovery and adaptive wall climbing. Copyright © 2012 John Wiley & Sons, Ltd.     

A logic-based environment for reactive agents in intelligent CAD systems

This paper is an evolution of a previous article by the authors[1] (Bento, J. P., Feijó, B., Lloyd-Smith, D., Computers and Structures, 1977, 63(5), 1015–1032) motivated by the need to provide computational support to an agent-based implementation of design processes. It presents a new programming environment to support the development of CAD systems based on a hybrid agent architecture in which the symbolic reasoning is carried out by first-order logic. The reactive behaviour of the agents can be achieved through a number of characteristics proposed for the object-oriented environment. This environment is also a general proposal for representing engineering design knowledge in which logic is integrated into an object-oriented paradigm. © 1998 Published by Elsevier Science Limited. All rights reserved.     

Factored reasoning for monitoring dynamic team and goal formation

We study the problem of monitoring goals, team structure and state of agents, in dynamic systems where teams and goals change over time. The setting for our study is an asymmetric urban warfare environment in which uncoordinated or loosely coordinated units may attempt to attack an important target. The task is to detect a threat such as an ambush, as early as possible. We attempt to provide decision-makers with early warnings, by simultaneously monitoring the positions of units, the teams to which they belong, and the goals of units. The hope is that we can detect situations in which teams of units simultaneously make movements headed towards a target, and we can detect their goal before they get to the target. By reasoning about teams, we may be able to detect threats sooner than if we reasoned about units individually. We develop a model in which the state space is decomposed into individual units’ positions, team assignments and team goals. When a unit belongs to a team it adopts the team’s goal. An individual unit’s movement depends only on its own goal, but different units interact as they form teams and adopt new goals. We present an algorithm that simultaneously tracks the positions of units, the team structure and team goals. Goals are inferred from two sources: individual units’ behavior, which provides information about their goals, and communications by units, which provides evidence about team formation. Our algorithm reasons globally about interactions between units and team formation, and locally about individual units’ behavior. We show that our algorithm performs well at the task, scaling to twenty units. It performs significantly better than several alternative algorithms: standard particle filtering, standard factored particle filtering, and an algorithm that performs all reasoning locally within the units.     

Motor planning: An architecture for specifying and controlling the behaviour of virtual actors

As virtual worlds become increasingly complex, task level interaction with virtual actors becomes correspondingly important. The control problem simply becomes unmanageable if we try to interact with synthetic agents at the wrong level of abstraction. However, it is not sufficient merely to implement a set of behaviours for a virtual actor; we require some mechanism for selecting and sequencing motor skills appropriate to the current behavioural goals and the states of other objects and actors in the virtual environment. In this paper we will describe a mechanism for linking perception and action to generate routine behaviours in a process we call motor planning. We present our implementation of the skill network, in which motor skills are the nodes and the arcs represent inhibitory and excitatory connections, including extensions to this architecture based on recent work in robotics. We characterize the domain of motor planning, i.e. what kinds of behaviour can it account for, and when will it fail? We close with a discussion of the limits of our current implementation and work that remains to be done.     

Self-calibration of a vision-based sensor network

When a network of vision-based sensors is emplaced in an environment for applications such as surveillance or monitoring the spatial relationships between the sensing units must be inferred or computed for self-calibration purposes. In this paper we describe a technique to solve one aspect of this self-calibration problem: automatically determining the topology and connectivity information of a network of cameras based on a statistical analysis of observed motion in the environment. While the technique can use labels from reliable cameras systems, the algorithm is powerful enough to function using ambiguous tracking data. The method requires no prior knowledge of the relative locations of the cameras and operates under very weak environmental assumptions. Our approach stochastically samples plausible agent trajectories based on a delay model that allows for transitions to and from sources and sinks in the environment. The technique demonstrates considerable robustness both to sensor error and non-trivial patterns of agent motion. The output of the method is a Markov model describing the behavior of agents in the system and the underlying traffic patterns. The concept is demonstrated with simulation data for systems containing up to 10 agents and verified with experiments conducted on a six camera sensor network.     

Learning to assign degrees of belief in relational domains

A recurrent problem in the development of reasoning agents is how to assign degrees of beliefs to uncertain events in a complex environment. The standard knowledge representation framework imposes a sharp separation between learning and reasoning; the agent starts by acquiring a “model” of its environment, represented into an expressive language, and then uses this model to quantify the likelihood of various queries. Yet, even for simple queries, the problem of evaluating probabilities from a general purpose representation is computationally prohibitive. In contrast, this study embarks on the learning to reason (L2R) framework that aims at eliciting degrees of belief in an inductive manner. The agent is viewed as an anytime reasoner that iteratively improves its performance in light of the knowledge induced from its mistakes. Indeed, by coupling exponentiated gradient strategies in learning and weighted model counting techniques in reasoning, the L2R framework is shown to provide efficient solutions to relational probabilistic reasoning problems that are provably intractable in the classical paradigm.     

IMPROVING THE SCOPE OF INTELLIGENT TUTORING BY ADAPTING A CASE-BASED METHODOLOGY THROUGH A DISTRIBUTED ARCHITECTURE

This paper describes an architecture for distributed case-based tutoring, called DICABTU, which provides an environment that facilitates cooperation among independent agents working together to provide highly individualized instruction. The fusion of these agents through a blackboard platform creates a distributed learning environment in which the most competent agents are called up to assist a student during a tutoring session. Following a curriculum derived from a node-based knowledge network, case-based reasoning is used to compose lessons at various levels of knowledge, to generate teaching materials, and to solve problems.     

A framework for interactive responsive animation

The availability of high‐performance 3D workstations has increased the range of application for interactive real‐time animation. In these applications the user can directly interact with the objects in the animation and direct the evolution of their motion, rather than simply watching a pre‐computed animation sequence. Interactive real‐time animation has fast‐growing applications in virtual reality, scientific visualization, medical training and distant learning. Traditional approaches to computer animation have been based on the animator having complete control over all aspects of the motion. In interactive animation the user can interact with any of the objects, which changes the current motion path or behaviour in real time. The objects in the animation must be capable of reacting to the user's actions and not simply replay a canned motion sequence. This paper presents a framework for interactive animation that allows the animator to specify the reactions of objects to events generated by other objects and the user. This framework is based on the concept of relations that describe how an object reacts to the influence of a dynamic environment. Each relation specifies one motion primitive triggered by either its enabling condition or the state of the environment. A collection of the relations is structured through several hierarchical layers to produce responsive behaviours and their variations. This framework is illustrated by several room‐based dancing examples that are modelled by relations. Copyright © 2000 John Wiley & Sons, Ltd.     

基于多自主智能体的群体动画创作

群体动画一直是计算机动画界一个具有挑战性的研究方向，提出了一个基于多自主智能体的群体动画创作框架：群体中的各角色作为自主智能体，能感知环境信息，产生意图，规划行为，最后通过运动系统产生运动来完成行为和实现意图，与传统的角色运动生成机理不同，首先采用运动捕获系统建立基本运动库，然后通过运动编辑技术对基本运动进行处理以最终得到角色运动，应用本技术，动画师只需“拍摄”角色群体的运动就能创作群体动画，极大地提高了制作效率。     

Planning interactive task for intelligent characters

Motion planning is an important problem in character animation and interactive simulation. However, few planning methods have considered domain‐specific knowledge that governs the agent's behaviors, and none of them is capable of planning the interactive task in which the agent interacts with the objects in the virtual environment. This paper presents a novel method to plan the interactive task based on Q‐learning for intelligent characters. The approach can be described as a three‐phase framework: data preprocessing phase, controller learning phase, and motion‐synthesis phase. In the data preprocessing phase, we abstract the motion clips as high‐level behaviors and construct the interactive behavior graph (IBG) to define the interactive capabilities of the agent in terms of interactive features. For the controller training phase, with IBG, Q‐learning algorithm is employed to train the control policy in the discrete domain with interactive features. In the motion‐synthesis phase, the optimal motion sequences can be generated by following the policy to accomplish the interactive task finally. The experimental results demonstrate that the uniform framework can generate reasonable and realistic motion sequences to plan interactive task in complex environment. Copyright © 2012 John Wiley & Sons, Ltd.     

Building blocks for goal-directed motion

An approach for the integration of intelligent goal directed motion within an interactive real-time computer animation environment is sketched. The approach is based on a cause-and-effect model for intelligent goal direction motion which gives rise to types of motion primitives: one is the application of a so called multi-track recording paradigm for direction motion manipulation; the second is an algorithm to mimic a control system for goal directed motion of joints in a multi-link structure. The techniques described here are features of the procedure based animation system which is described in Reference 1.     

Genetic programming for articulated figure motion

We present an approach to articulated figure motion in which motion tasks are defined in terms of goals and ratings. The agents are dynamically-controlled robots whose behaviour is determined by robotic controller programs. The controller programs for the robots are evaluated at each time step to yield torque values which drive the dynamic simulation of the motion. We use the AI technique of genetic programming (GP) to automatically derive control programs for the agents which achieve the goals. This type of motion specification is an alternative to key framing which allows a highly automated, learning-based approach to generation of motion. This method of motion control is very general (it can be applied to any type of motion), yet it allows for specifications of the types of specific motion which are desired for a high quality animation. We show that complex, specific, physically plausible and aesthetically appealing motion can be generated using these methods.     

Generation of augmented video sequences combining behavioral animation and multi‐object tracking

In this paper we present a novel approach to generate augmented video sequences in real‐time, involving interactions between virtual and real agents in real scenarios. On the one hand, real agent motion is estimated by means of a multi‐object tracking algorithm, which determines real objects' position over the scenario for each time step. On the other hand, virtual agents are provided with behavior models considering their interaction with the environment and with other agents. The resulting framework allows to generate video sequences involving behavior‐based virtual agents that react to real agent behavior and has applications in education, simulation, and in the game and movie industries. We show the performance of the proposed approach in an indoor and outdoor scenario simulating human and vehicle agents. Copyright © 2009 John Wiley & Sons, Ltd.     

基于动态描述逻辑DDL的动作理论

基于一阶谓词逻辑或高阶逻辑的动作理论与采用命题语言的动作理论之间存在一个关于描述和推理能力的鸿沟;作为描述逻辑的动态扩展,动态描述逻辑DDL为基于描述逻辑的动作刻画和推理提供了一种途径.系统地研究了基于DDL的动作表示和推理问题.首先,在应用描述逻辑对静态领域知识进行刻画的基础上,引入带参数的原子动作定义式和带参数的复...     

Multi levels semantic architecture for multimodal interaction

This paper presents a semantic architecture for solving multimodal interaction. Our architecture is based on multi agent systems where agents are purely semantic using ontologies and inference system. Multi levels concepts and behavioural models are taken into account to bring a fast high level reasoning on a big amount of percepts and low level actions. We apply this architecture to make a system aware of different situations in a network like tracking object behaviours of the environment. As a proof of concept, we apply our architecture to an assistant robot helping blind or disabled people to cross a road in a virtual reality environment.     

A Computational Framework for Practical Social Reasoning

This article describes a framework for practical social reasoning designed to be used for analysis, specification, and implementation of the social layer of agent reasoning in multiagent systems. Our framework, called the expectation strategy behavior (ESB) framework, is based on (i) using sets of update rules for social beliefs tied to observations (so‐called expectations), (ii) bounding the amount of reasoning to be performed over these rules by defining a reasoning strategy, and (iii) influencing the agent's decision‐making logic by means of behaviors conditioned on the truth status of current and future social beliefs. We introduce the foundations of ESB conceptually and present a formal framework and an actual implementation of a reasoning engine, which is specifically combined with a general (belief–desire–intention‐based) practical reasoning programming system. We illustrate the generality of ESB through select case studies, which show that it is able to represent and implement different typical styles of social reasoning. The broad coverage of existing social reasoning methods, the modularity that derives from its declarative nature, and its focus on practical implementation make ESB a useful tool for building advanced socially reasoning agents.     

COSAR: hybrid reasoning for context-aware activity recognition

Human activity recognition is a challenging problem for context-aware systems and applications. Research in this field has mainly adopted techniques based on supervised learning algorithms, but these systems suffer from scalability issues with respect to the number of considered activities and contextual data. In this paper, we propose a solution based on the use of ontologies and ontological reasoning combined with statistical inferencing. Structured symbolic knowledge about the environment surrounding the user allows the recognition system to infer which activities among the candidates identified by statistical methods are more likely to be the actual activity that the user is performing. Ontological reasoning is also integrated with statistical methods to recognize complex activities that cannot be derived by statistical methods alone. The effectiveness of the proposed technique is supported by experiments with a complete implementation of the system using commercially available sensors and an Android-based handheld device as the host for the main activity recognition module.