Frank Guerin

A Survey of the Ontogeny of Tool Use: From Sensorimotor Experience to Planning

In this paper, we review current knowledge on tool use development in infants in order to provide relevant information to cognitive developmental roboticists seeking to design artificial systems that develop tool use abilities. This information covers: 1) sketching developmental pathways leading to tool use competences; 2) the characterization of learning and test situations; 3) the crystallization of seven mechanisms underlying the developmental process; and 4) the formulation of a number of challenges and recommendations for designing artificial systems that exhibit tool use abilities in complex contexts.

Denotational semantics for agent communication language

The dilemma encountered in the design of an agent communication language (ACL) for an open society is that it should be based on externally observable phenomena yet it should capture something of the intuitions behind the high level abstractions typically found in internal mental states. Our solution treats an ACL message as a declarative statement that is given a procedural interpretation by a denotational semantics. This defines a speech act as a function between states. These states are social states which store public information including expressed mental attitudes and control variables. Expressed mental attitudes are externally observable and capture the conventional public meaning of communication. The variables control the flow of conversation in a protocol. We conclude firstly that since the denotational semantics is based on externally observable phenomena, it is possible to verify compliance and prove properties of protocols. Secondly, since the semantics is more expressive than behavioural specifications, it lays the foundation for high-level communication between intelligent agents.

Development of Object and Grasping Knowledge by Robot Exploration

We describe a bootstrapping cognitive robot system that-mainly based on pure exploration-acquires rich object representations and associated object-specific grasp affordances. Such bootstrapping becomes possible by combining innate competences and behaviors by which the system gradually enriches its internal representations, and thereby develops an increasingly mature interpretation of the world and its ability to act within it. We compare the systems prior competences and developmental progress with human innate competences and developmental stages of infants.

Guaranteeing Properties for E-commerce Systems

Agents in an electronic commerce system act on behalf of potentially competing individuals and organisations. It is proposed that such systems will be used in scenarios where legally binding contracts are made or money is exchanged by the agents on behalf of their owners. Agent owners may be reluctant to delegate tasks involving uncertain and possibly detrimental outcomes to an agent without assurances about the systems properties. It may be a requirement, for example, that an agent cannot profit from lying to its peers. This paper demonstrates how solutions from game theory together with computing theories can be used to publicly specify rules and prove desirable properties for agent systems. This has the potential to increase the range of applications in which agent owners may be willing to delegate to their embedded counterparts.

Protocols and intentional specifications of multi-party agent conversions for brokerage and auctions

Conversations involving three or more agents often occur in multi-agent systems, for example in brokering and auctions. For developing agents in open systems, it is important that the interactions in such conversations have a precise and unambiguous meaning. We address this issue by generalising a protocol-based semantic framework for expressing the semantics of Agent Communication Languages. The generalisations involve exploiting mechanistic aspects of the interaction (conversation identifiers), greater flexibility in the space of possible replies, and a richer representation of protocol states. We define intentional specifications for some brokerage and auction protocols, including event-based clocks to determine the ordering of events. We conclude that this approach to specifying multi-party protocols leads to clearer interfaces for open systems and easier re-use, with a potentially significant impact on standardisation efforts.

Verifying Dominant Strategy Equilibria in Auctions

Future agent mediated eCommerce will involve open systems of agents interoperating between different institutions, where different auction protocols may be in use. We argue that in order to achieve this agents will need a method to automatically verify the properties of a previously unseen auction protocol; for example, they may wish to verify that it is fair and robust to deception. We are therefore interested in the problem of automatically verifying the game-theoretic properties of a given auction mechanism, especially the property of strategyproofness. In this paper we show how the Alloy model checker can be used to automatically verify such properties. We illustrate the approach via two examples: a simple two player Vickrey auction and a quantity restricted multi-unit auction using the Vickrey-Clarke-Groves mechanism.

Proving properties of open agent systems

In an open agent system the constituent agents are developed and owned by different individuals or organisations who may have conflicting interests. Hence the internals (i.e. program and state) of agents are not public and so notions of trust and deception are relevant. It is proposed that such systems will be used in scenarios where legally binding contracts are made and money is exchanged by the agents on behalf of their owners. Not surprisingly, agent owners can be expected to be reluctant to delegate tasks involving potentially detrimental outcomes to an agent unless they can be assured that the system has certain desirable properties. It may be a requirement, for example, that an agent cannot profit from lying to its peers. Solutions from game theory and economics [1] allow us to design mechanisms for interactions (a set of public rules governing an interaction) which provide incentives for participants to behave as we desire, for example to tell the truth; mechanisms can be designed to have properties such as individual rationality, incentive compatibility and stability. Having chosen a suitable mechanism, we can implement it for an agent system as a protocol and prove that these properties hold for the agent system.

Realising Common Knowledge Assumptions in Agent Auctions

Game theory is popular in agent systems for designing auctions with desirable properties. However, many of these properties will only hold if the game and its properties are common knowledge among the agents. For example, in an auction where truthful bidding is an equilibrium strategy, unless this is common knowledge, it may not be rational for an agent to bid truthfully. It is currently not clear how this state of common knowledge can be achieved, especially in open agent societies where agents may encounter previously unseen auction specifications. We need a method for communicating the rules of the game to the agents, and the agents need to be able to determine its properties. We present a machine-readable language in which the rules of the game can be written. We show that it is not feasible for an agent to determine the properties of any arbitrary specification, unless information about the properties is communicated and/or certain restrictions are placed on the specification. We look at two special cases where common knowledge is achievable: auctions with identical players where the two highest bidders determine the price, and Groves mechanisms with a restriction on the pricing rule.

A semantic framework for specifying agent communication languages

The paper presents a hybrid semantic framework for the specification of agent communication languages (ACL), unifying intentional semantics with protocol based approaches to provide a more expressive framework. A richer state description based on publicly expressed mental attitudes is introduced. The state description also allows agents to be assigned certain roles in a protocol. Together, this moves to a more workflow-oriented approach, in which conversations are guided by protocols but the effect of individual communicative acts is highly sensitive to the context in which it occurs. We evaluate the usefulness of this framework as a standard for ACLs.

Learning and using context on a humanoid robot using latent dirichlet allocation

In this work, we model context in terms of a set of concepts grounded in a robots sensorimotor interactions with the environment. For this end, we treat context as a latent variable in Latent Dirichlet Allocation, which is widely used in computational linguistics for modeling topics in texts. The flexibility of our approach allows many-to-many relationships between objects and contexts, as well as between scenes and contexts. We use a concept web representation of the perceptions of the robot as a basis for context analysis. The detected contexts of the scene can be used for several cognitive problems. Our results demonstrate that the robot can use learned contexts to improve object recognition and planning.