Approaches to cognitive architecture of autonomous intelligent agent

Behavior based AI [8,18] has questioned the need for modeling intelligent agency using generalized cognitive modules for perception and behavior generation. Behavior based AI has demonstrated successful interactions in unpredictable environments in the mobile robot domain [7, 8]. This has created a gulf between traditional approaches to modeling intelligent agency and behavior based approaches. We present an architecture for intelligent autonomous agents which we call GLAIR (Grounded Layered Architecture with Integrated Reasoning) [13, 14, 12]. GLAIR is a general multi-level architecture for autonomous cognitive agents with integrated sensory and motor capabilities. GLAIR offers an unconscious layer for modeling tasks that exhibit a close affinity between sensing and acting, i.e., behavior based AI modules, and a conscious layer for modeling tasks that exhibit delays between sensing and acting. GLAIR provides learning mechanisms that allow for autonomous agents to learn emergent behaviors and add it to their repertoire of behaviors. In this paper we will describe the principles of GLAIR and systems we have developed that demonstrate how GLAIR based agents acquire and exhibit a repertoire of behaviors at different cognitive levels.

In contrast to “conscious”, reasoned behaviors, we consider behaviors that are automatic and unreasoned to be “unconscious”. The latter are commonly found in behavior-based AI models [BroSO, Maego]. We are developing an architecture that models agents with both “conscious” and “unconscious” behaviors. Furthermore, we are interested in modeling agents that learn behaviors from their successful interactions with the world. We call these learned behaviors “emergent behaviors”. We present an architecture for intelligent autonomous agents which we call GLAIR (Grounded Layered Architecture with Integrated Reasoning) [HLS92, HCBS93, HLS93, HLCS93, LHS 1. GLAIR is a general multilevel architecture for autonomous agents with sensory and motor capabilities. GLAIR offers “unconscious” layers for modeling tasks that exhibit a close affinity between sensing and acting, i.e., behavior-based AI modules, and a “conscious” layer for modeling tasks that exhibit delays between sensing and acting, and require deliberation on the part of the agent. GLAIR provides learning mechanisms that allow for autonomous agents to learn emergent behaviors and add them to their repertoire of behaviors. In this paper we will describe the principles of GLAIR, and an application we have developed that demonstrates how GLAIR-based agents acquire and exhibit a repertoire of behaviors at different cognitive levels.

In this paper, a software architecture for multiple autonomous agents in a real-time environment is described. The software architecture for autonomous agents must effectively connect perception modules, planning modules and action modules. It is necessary to define both an interface and a data flow among different modules. The proposed software architecture connects reactive modules with deliberative modules. The architecture proved its feasibility by controlling a prototype of multiple model digital trains.

The paper describes an architecture for autonomous agents that combines the advantages of both the classical hierarchical and the layered architectures. The aim is to build more 'intelligent' autonomous agents that are capable of planning for high level goals as well as having the ability to react and respond to arising situations in a human like manner. >

In this paper, we introduce a cognitive agent architecture that can be used in the study of Human-Robot Interaction. The Cognitive Architecture for Perception-Reaction Intelligent Computer Agents (CAPRICA) is an extensible agent library built around the ideas of theory of mind, episodic memory, and embodied cognition. Existing agent research in each of these areas was used to formulate design requirements. We provide an overview of the library's design and discuss future work in progress.

This research integrates rigorous methods of reinforcement learning (RL) and control engineering with a behavioral (ethology) approach to the agent technology. The main outcome is a hybrid architecture for intelligent autonomous agents targeted to the Artificial Life like environments. The architecture adopts several biology concepts and shows that they can provide robust solutions to some areas. The resulting agents perform from primitive behaviors, simple goal directed behaviors, to complex planning. The agents are fully autonomous through environment feedback evaluating internal agent state and motivates the agent to perform behaviors that return the agent towards optimal conditions. This principle is typical to animals. Learning and control is realized by multiple RL controllers working in a hierarchy of Semi Markov Decision Processes (SMDP). Used model free Q(lambda) learning works online, the agents gain experiences during interaction with the environment. The decomposition of the root SMDP into hierarchy is automated as opposed to the conventional methods that are manual. The agents assess utility of the behavior and provide rewards to RL controller as opposed to the conventional RL methods where the rewards-situations map is defined by the designer upfront. The resulting learning algorithm converges to a recursively optimal solution with probability 1. Agent behavior is continuously optimized according to the distance from the agentpsilas optimal conditions.

This paper integrates rigorous methods of reinforcement learning (RL) and control engineering with a behavioral (ethology) approach to the agent technology. The main outcome is a hybrid architecture for intelligent autonomous agents targeted to the Artificial Life like environments. The architecture adopts several biology concepts and shows that they can provide robust solutions to some areas. The resulting agents perform from primitive behaviors, simple goal directed behaviors, to complex planning. The agents are fully autonomous through environment feedback evaluating internal agent state and motivate the agent to perform behaviors that return the agent towards optimal conditions. This principle is typical to animals. Learning and control is realized by multiple RL controllers working in a hierarchy of Semi Markov Decision Processes (SMDP). Used model free Q(λ ) learning works online, the agents gain experiences during interaction with the environment. The decomposition of the root SMDP into hierarchy is automated as opposed to the conventional methods that are manual. The agents assess utility of the behavior and provide rewards to RL controller as opposed to the conventional RL methods where the rewards situations map is defined by the designer upfront. Agent behavior is continuously optimized according to the distance from the agent’s optimal conditions.

Presented topic is from area of development of artificial creatures and proposes new architecture of autonomous agent. The work builds on a research of the latest approaches to Artificial Life, realized by the Department of Cybernetics, CTU in Prague in the last twenty years. This architecture design combines knowledge from Artificial Intelligence (AI), Ethology, Artificial Life (ALife) and Intelligent Robotics. From the field of classical AI, the fusion of reinforcement learning, planning and artificial neural network into one more complex control system was used here. The main principle of its function is inspired by the field of Ethology, this means that life of given agent tries to be similar to life of an animal in the Nature, where animal learns relatively autonomously from simpler principles towards the more complex ones. The architecture supports on-line learning of all knowledge from the scratch, while the core principle is in hierarchical Reinforcement Learning (RL), this action hierarchy is created autonomously based solely on agents interaction with an environment. The main key idea behind this approach is in original implementation of a domain independent hierarchical planner. Our planner is able to operate with behaviors learned by the RL. It means that an autonomously gained hierarchy of actions can be used not only by action selection mechanisms based on the reinforcement learning, but also by a planning system. This gives the agent ability to utilize high-level deliberative problem solving based solely on his experiences. In order to deal with higher-level control rather than a sensory system, the life of agent was simulated in a virtual environment.

Exploiting the large language models, autonomous agents have been widely developed for tackling decision-making and complex tasks in interactive environments. With the development of LLMs, multimodal agents have emerged with the ability of understanding not only text, but also images, sounds and videos. Therefore, multimodal agents are naturally applicable to generating recommendation information. However, few solutions focus on solving the intelligent recommendation problems on mobile phones, particularly in actively collecting and analyzing users preferences. In this paper, we propose an architecture for autonomous recommendation agents on cellphones, which can actively collect and understand multimodal data without restrictions on application layouts and graphical interface. We also introduce detail methods and workflows of the proposed agent. To this end, we capture a complete view of the proposed system.

A study on mimic reactive architecture for autonomous agents

This paper presents a three level architecture for an autonomous agent and its application to the navigation problem in an unknown environment. The architecture is structured in three levels, called, reactive, instinctive and cognitive. The reactive level is based on a feed-forward symbolic neural network. The instinctive level is based on a set of predefined behaviors selected by a fuzzy classifier according to the perceived situation. Finally, the cognitive level is supported by symbolic production rules that determine the global behavior of the agent. In this sense, the three levels are responsible by behaviors of increasing complexity. The main characteristics of the architecture are: heterogeneity, hierarchic assembly, behavior-oriented design and biological plausibility. Some examples are also presented, that show the behavior robustness of the proposed architecture in a simulated environment.

Autonomous robots will have to have the capability to make decisions on their own to varying degrees. In this chapter, I will make the plea for developing moral capabilities deeply integrated into the control architectures of such autonomous agents, for I shall argue that any ordinary decision-making situation from daily life can be turned into a morally charged decision-making situation.

Autonomous agents (AAs) are designed to embody the natural intelligence by incorporating cognitive mechanisms that are applied to evaluate stimuli from an emotional perspective. Computational models of emotions (CMEs) implement mechanisms of human information processing in order to provide AAs for a capability to assign emotional values to perceived stimuli and implement emotion-driven behaviors. However, a major challenge in the design of CMEs is how cognitive information is projected from the architecture of AAs. This article presents a cognitive model for CMEs based on appraisal theory aimed at modeling AAs' interactions between cognitive and affective processes. The proposed scheme explains the influence of AAs' cognition on emotions by fuzzy membership functions associated to appraisal dimensions. The computational simulation is designed in the context of an integrative framework to facilitate the development of CMEs, which are capable of interacting with cognitive components of AAs. This article presents a case study and experiment that demonstrate the functionality of the proposed models.

A new architecture for autonomous agents is proposed. The architecture integrates the symbolic and the behavioral processing of data coming from the robot sensors. The integration is based on the introduction of a conceptual space representation that links the subconceptual level, which is a repository of behavioral modules, with the symbolic level, in which rich symbolic descriptions of the agent environment take place.

A generic architecture for autonomous agents is presented. In common with other current proposals the agent is capable of reacting to and reasoning about events which occur in its environment, executing actions and plans in order to achieve goals in its environment, and communicating with other agents. The work described here proposes certain advances on other systems, notably the ability to reason about and make decisions under uncertainty, including decisions about competing beliefs and alternative actions. The framework is grounded in a non-classical decision model, the ‘domino’ model. This is formalized to ensure continuity with classical decision theory and avoid ad hoc features. The domino model is embodied in a well-defined knowledge representation language, R2L, which explicitly supports the central concepts of decisions and plans, and associated constructs of goals, arguments, commitments, obligations and constraints. The availability of such a language provides a sound basis for building knowledge-based agents for practical applications. A major issue for such applications, however, is how to ensure their safe operation. This is a central issue whether the agents are used in an advisory role (e.g. decision support systems) or an autonomous one (e.g. in a robot). Techniques for explicit management of safety are described and some broader theoretical implications are discussed.