The current chapter is devoted to <em>roughification</em>. In the most general setting, we intend the term <em>roughification</em> to refer to methods/techniques of constructing equivalence/similarity relations adequate for Pawlak-like approximations. Such techniques are fundamental in rough set theory. We propose and investigate novel roughification techniques. We show that using the proposed techniques one can often discern objects indiscernible by original similarity relations, what results in improving approximations. We also discuss applications of the proposed techniques in granulating relational databases and concept learning. The last application is particularly interesting, as it shows an approach to concept learning which is more general than approaches based solely on information and decision systems.

The current paper is devoted to belief fusion when information sources may deliver incomplete and inconsistent information. In such cases paraconsistent and commonsense reasoning techniques can be used to complete missing knowledge and disambiguate inconsistencies. We propose a novel, realistic model of distributed belief fusion and an implementation framework guaranteeing its tractability.

Research with collaborative robotic systems has much to gain by leveraging concepts and ideas from the areas of multi-agent systems and the social sciences. In this paper we propose an approach to formalizing and grounding important aspects of collaboration in a collaborative system shell for robotic systems. This is done primarily in terms of the concept of delegation, where delegation will be instantiated as a speech act. The formal characterization of the delegation speech act is based on a preformal theory of delegation proposed by Falcone and Castelfranchi. We show how the delegation speech act can in fact be used to formally ground an abstract characterization of delegation into a FIPA-compliant implementation in an agent-oriented language such as JADE, as part of a collaborative system shell for robotic systems. The collaborative system shell has been developed as a prototype and used in collaborative missions with multiple unmanned aerial vehicle systems.

In the next decades, practically viable robotic/agent systems are going to be mixed-initiative in nature. Humans will request help from such systems and such systems will request help from humans in achieving the complex mission tasks required. Pragmatically, one requires a distributed task specification language to define tasks and a suitable data structure which satisfies the specification and can be used flexibly by collaborative multi-agent/robotic systems. This paper defines such a task specification language and an abstract data structure called Task Specification Trees which has many of the requisite properties required for mixed-initiative problem solving and adjustable autonomy in a distributed context. A prototype system has been implemented for this delegation framework and has been used practically with collaborative unmanned aircraft systems.

Unmanned aircraft systems (UASs) are now becoming technologically mature enough to be integrated into civil society. An essential issue is principled mixed-initiative interaction between UASs and human operators. Two central problems are to specify the structure and requirements of complex tasks and to assign platforms to these tasks. We have previously proposed Task Specification Trees (TSTs) as a highly expressive specification language for complex multi-agent tasks that supports mixed-initiative delegation and adjustable autonomy. The main contribution of this paper is a sound and complete distributed heuristic search algorithm for allocating the individual tasks in a TST to platforms. The allocation also instantiates the parameters of the tasks such that all the constraints of the TST are satisfied. Constraints are used to model dependencies between tasks, resource usage as well as temporal and spatial requirements on complex tasks. Finally, we discuss a concrete case study with a team of unmanned aerial vehicles assisting in a challenging emergency situation.

The paper is devoted to a novel formalization of beliefs in multiagent systems. Our aim is to bridge the gap between idealized logical approaches to modeling beliefs and their actual implementations. Therefore the stages of belief acquisition, intermediate reasoning and final belief formation are isolated and analyzed. We give a novel semantics reflecting those stages and suitable for building complex belief structures in the context of incomplete and/or inconsistent information. Namely, an agent starts with constituents, i.e., sets of initial beliefs acquired by perception, expert supplied knowledge, communication with other agents and perhaps other ways. Next, the constituents are transformed into consequents according to agents’ epistemic profiles. Additionally, a uniform treatment of single agent and group beliefs is achieved. Importantly, we indicate an implementation framework ensuring tractability of reasoning about beliefs.

Description logics refer to a family of formalisms concentrated around concepts, roles and individuals. They are used in many multiagent and Semantic Web applications as a foundation for specifying knowledge bases and reasoning about them. Among them, one of the most important logics is <em>SROIQ</em><img src=\"http://www.springerlink.com/jsMath/fonts/cmsy10/alpha/120/char53.png\" /><img src=\"http://www.springerlink.com/jsMath/fonts/cmsy10/alpha/120/char52.png\" /><img src=\"http://www.springerlink.com/jsMath/fonts/cmsy10/alpha/120/char4F.png\" /><img src=\"http://www.springerlink.com/jsMath/fonts/cmsy10/alpha/120/char49.png\" /><img src=\"http://www.springerlink.com/jsMath/fonts/cmsy10/alpha/120/char51.png\" />, providing the logical foundation for the OWL 2 Web Ontology Language recommended by W3C in October 2009. In the current paper we address the problem of inconsistent knowledge. Inconsistencies may naturally appear in the considered application domains, for example as a result of fusing knowledge from distributed sources. We introduce a number of paraconsistent semantics for <em>SROIQ</em><img src=\"http://www.springerlink.com/jsMath/fonts/cmsy10/alpha/120/char53.png\" /><img src=\"http://www.springerlink.com/jsMath/fonts/cmsy10/alpha/120/char52.png\" /><img src=\"http://www.springerlink.com/jsMath/fonts/cmsy10/alpha/120/char4F.png\" /><img src=\"http://www.springerlink.com/jsMath/fonts/cmsy10/alpha/120/char49.png\" /><img src=\"http://www.springerlink.com/jsMath/fonts/cmsy10/alpha/120/char51.png\" />, including three-valued and four-valued semantics. The four-valued semantics reflects the well-known approach introduced in [5,4] and is considered here for comparison reasons only. We also study the relationship between the semantics and paraconsistent reasoning in <em>SROIQ</em><img src=\"http://www.springerlink.com/jsMath/fonts/cmsy10/alpha/120/char53.png\" /><img src=\"http://www.springerlink.com/jsMath/fonts/cmsy10/alpha/120/char52.png\" /><img src=\"http://www.springerlink.com/jsMath/fonts/cmsy10/alpha/120/char4F.png\" /><img src=\"http://www.springerlink.com/jsMath/fonts/cmsy10/alpha/120/char49.png\" /><img src=\"http://www.springerlink.com/jsMath/fonts/cmsy10/alpha/120/char51.png\" /> through a translation into the traditional two-valued semantics. Such a translation allows one to use existing tools and reasoners to deal with inconsistent knowledge.

The starting point of this research is the multimodal approach to modeling multiagent systems, especially Beliefs, Goals and Intention systems. Such an approach is suitable for specifying and verifying many subtle aspects of agents’ informational and motivational attitudes.However, in this chapter we make a shift in a perspective. More precisely, we propose the method of embedding multimodal approaches into a form of approximate reasoning suitable for modeling perception, namely a <em>similarity-based approximate reasoning</em>. We argue that this formalism allows one to both keep the intuitive semantics compatible with that of multimodal logics as well as to model and implement phenomena occurring at the perception level.

Collaborative robotic systems have much to gain by leveraging results from the area of multi-agent systems and in particular agent-oriented software engineering. Agent-oriented software engineering has much to gain by using collaborative robotic systems as a testbed. In this article, we propose and specify a formally grounded generic collaborative system shell for robotic systems and human operated ground control systems. Collaboration is formalized in terms of the concept of delegation and delegation is instantiated as a speech act. Task Specification Trees are introduced as both a formal and pragmatic characterization of tasks and tasks are recursively delegated through a delegation process implemented in the collaborative system shell. The delegation speech act is formally grounded in the implementation using Task Specification Trees, task allocation via auctions and distributed constraint problem solving. The system is implemented as a prototype on Unmanned Aerial Vehicle systems and a case study targeting emergency service applications is presented.

When unmanned aerial vehicles (UAVs) are used to survey distant targets, it is important to transmit sensor information back to a base station. As this communication often requires high uninterrupted bandwidth, the surveying UAV often needs afree line-of-sight to the base station, which can be problematic in urban or mountainous areas. Communication ranges may also belimited, especially for smaller UAVs. Though both problems can be solved through the use of relay chains consisting of one or more intermediate relay UAVs, this leads to a new problem: Where should relays be placed for optimum performance? We present two new algorithms capable of generating such relay chains, one being a dual ascent algorithm and the other a modification of the Bellman-Ford algorithm. As the priorities between the numberof hops in the relay chain and the cost of the chain may vary, wecalculate chains of different lengths and costs and let the ground operator choose between them. Several different formulations for edge costs are presented. In our test cases, both algorithms are substantially faster than an optimized version of the original Bellman-Ford algorithm, which is used for comparison.

The study of frameworks and formalisms for reasoning about action and change [67, 58, 61, 65, 70, 3, 57] has been central to the knowledge representation field almost from the inception of Artificial Intelligence as a general field of research [52, 56]. The phrase “Temporal Action Logics” represents a class of logics for reasoning about action and change that evolved from Sandewall’s book on Features and Fluents [61] and owes much to this ambitious project. There are essentially three major parts to Sandewall’s work. He first developed a narrative-based logical framework for specifying agent behavior in terms of action scenarios. The logical framework is state-based and uses explicit time structures. He then developed a formal framework for assessing the correctness (soundness and completeness) of logics for reasoning about action and change relative to a set of well-defined intended conclusions, where reasoning problems were classified according to their ontological or epistemological characteristics. Finally, he proposed a number of logics defined semantically in terms of definitions of preferential entailment1 and assessed their correctness using his assessment framework.

Fuzzy logics are one of the most frequent approaches to model uncertainty and vagueness. In the case of fuzzy modeling, degrees of belief and disbelief sum up to 1, which causes problems in modeling the lack of knowledge and inconsistency. Therefore, so called paraconsistent intuitionistic fuzzy sets have been introduced, where the degrees of belief and disbelief are not required to sum up to 1. The situation when this sum is smaller than 1 reflects the lack of knowledge and its value greater than 1 models inconsistency. In many applications there is a strong need to guide and interpret fuzzy-like reasoning using qualitative approaches. To achieve this goal in the presence of uncertainty, lack of knowledge and inconsistency, we provide a framework for qualitative interpretation of the results of fuzzy-like reasoning by labeling numbers with words, like <em>true, false, inconsistent, unknown</em>, reflecting truth values of a suitable, usually finitely valued logical formalism.

This paper extends the basic rough set formalism introduced by Pawlak [1] to a rule-based knowledge representation language, called Rough Datalog, where rough sets are represented by predicates and described by finite sets of rules. The rules allow us to express background knowledge involving rough concepts and to reason in such a knowledge base. The semantics of the new language is based on a four-valued logic, where in addition to the usual values True and False, we also have the values Boundary, representing uncertainty, and Unknown corresponding to the lack of information. The semantics of our language is based on a truth ordering different from the one used in the well-known Belnap logic [2, 3] and we show why Belnap logic does not properly reflect natural intuitions related to our approach. The declarative semantics and operational semantics of the language are described. Finally, the paper outlines a query language for reasoning about rough concepts.

Robots are autonomous agents whose actions are performed in the real world during a period of time. There are a number of general constraints on such actions, for example that the same action can not have two separate instances during overlapping time intervals, or restrictions that are due to which state variables affect the action or are affected by it. Each process in the robot's cognitive system that is to request the initiation of an action must respect those restrictions. In this article we describe the design and formal characterization of a separate process, called an action coordinator, that manages these restrictions.

The introduction of games for benchmarking intelligent systems has a long tradition in AI (Artificial Intelligence) research. Alan Turing was one of the first to mention that a computer can be considered as intelligent if it is able to play chess. Today AI benchmarks are designed to capture difficulties that humans deal with every day. They are carried out on robots with unreliable sensors and actuators or on agents integrated in digital environments that simulate aspects of the real world. One example is given by the annually held RoboCup competitions, where robots compete in a football game but also fight for the rescue of civilians in a simulated large-scale disaster simulation. Besides these scientific events, another environment, also challenging AI, originates from the commercial computer game market. Computer games are nowadays known for their impressive graphics and sound effects. However, the latest generation of game engines shows clearly that the trend leads towards more realistic physics simulations, agent centered perception, and complex player interactions due to the rapidly increasing degrees of freedom that digital characters obtain. This new freedom requests another quality of the player’s environment, a quality of ambient intelligence that appears both plausible and in real time. This intelligence has, for example, to control more than $40$ facial muscles of digital characters while they interact with humans, but also to control a team of digital characters for the support of human players. This article emphasizes the current difference between AI systems and digital characters in commercial computer games and emphasizes the advantages that arise if shrinking the gap between them. We sketch some methods currently utilized in RoboCup and relates them to methods found in commercial computer games. We show how methods from RoboCup might contribute to game AI and improve both the performance and plausibility of its digital characters. Furthermore, we describe state-of-the-art game engines and discuss the challenge but also opportunity they are offering to AI research.

This chapter proposes a framework for specifying, constructing, and managing aparticular class of approximate knowledge structures for use with intelligent artifacts rangingfrom simpler devices such as personal digital assistants (PDAs) to more complex ones suchas unmanned aerial vehicles (UAVs). This chapter introduces the notion of an approximationtransducer, which takes approximate relations as input and generates a (possibly moreabstract) approximate relation as output by combining the approximate input relations witha crisp local logical theory representing dependencies between input and output relations.Approximation transducers can be combined to produce approximation trees, which representcomplex approximate knowledge structures characterized by the properties of elaborationtolerance, groundedness in the application domain, modularity, and context dependency.Approximation trees are grounded through the use of primitive concepts generated with supervisedlearning techniques. Changes in definitions of primitive concepts or in the locallogical theories used by transducers result in changes in the knowledge stored in approximationtrees by increasing or decreasing precision in the knowledge qualitatively. Intuitionsand techniques from rough set theory are used to define approximate relations where eachhas an upper and a lower approximation. The constituent components in a rough set havecorrespondences in a logical language used to relate crisp and approximate knowledge. Theinference mechanism associated with the use of approximation trees is based on a generalizationof deductive databases that we call rough relational databases. Approximation trees andqueries to them are characterized in terms of rough relational databases and queries to them.By placing certain syntactic restrictions on the local theories used in transducers, the computationalprocesses used in the query/answering and generation mechanism for approximationtrees remain in PTIME.

Developing agents for simulation environments is usually the responsibility of computer experts. However, as domain experts have superior knowledge of the intended agent behavior, it is desirable to have domain experts directly specifying behavior. In this paper we describe a system which allows non-computer experts to specify the behavior of agents for the RoboCup domain. An agent designer is presented with a Graphical User Interface with which he can specify behaviors and activation conditions for behaviors in a layered behavior-based system. To support the testing and debugging process we are also developing interfaces that show, in real-time, the world from the agents perspective and the state of its reasoning process.

In this paper we present an AI programming organised around the RoboCup soccer simulation system. The course participants create a number of software agents that form a team, and participate in a tournament at the end of the course. The use of a challenging and interesting task, and the incentive of having a tournament has made the course quite successful, both in term of enthusiasm of the students and of knowledge acquired. In the paper we describe the structure of the course, discuss in what respect we think the course has met its aim, and the opinions of the students about the course.

Sandewall has recently proposed a systematic approach to the representation of knowledge about dynamical systems that includes a general framework in which to assess the range of applicability of existing and new logics for action and change and to provide a means of studying whether and in what sense the logics of action and change are relevant for intelligent agents. As part of the framework, a number of logics of preferential entailment are introduced and assessed for particular classes of action scenario descriptions. This paper provides syntactic characterizations of several of these relations of preferential entailment in terms of standard FOPC and circumscription axioms. The intent is to simplify the process of comparison with existing formalisms which use more traditional techniques and to provide a basis for studying the feasibility of compiling particular classes of problems into logic programs.