Multiple-category concept tasks may be divided into three stages: (a) specification of dimensions and values, (&) dimension selection, and (c) associative learning. Although previous research has indicated that most of the total learning time was spent in associative learning, the relative importance of the stages was expected to be contingent on stimulus-complexity variables. The effects of number (2 or 3) of values per dimension and the number (1 or 2) of irrelevant dimensions on dimension selection and associative learning were studied in a 2 X 2 factorial design using 44 5"s. The results supported the hypothesis that difficulty in dimension selection is primarily a function of the number of irrelevant dimensions, while dfficulty in associative learning is dependent on the number of values per dimension.
This paper outlines a novel method for Cooperative Behavioral Control of distributed heterogeneous autonomous systems, emulating the methods in which humans collaborate. This method allows autonomous systems to collaborate on tasks and mission goals, similar to how humans interact, and is effective and efficient for real-time resource allocation. The proposed method fundamentally reduces the required communication bandwidths by significantly decreasing the amount of data necessary for real-time information exchange between cooperating agents. This is done by creating a swarm to estimate the beliefs of the collective, and not on physical states which is usually done by classical approaches. In sharing core beliefs, a collective of heterogeneous agents can plan as an individual, inherently and naturally deconflicting the notion of cost and optimality.
An important problem for intelligent autonomous mobile systems/agents is the ability to predict the motions of other objects/agents. This has natural extensions to cooperative behavior control, where mobile agents avoid each other by predicting the other's motion. In this paper, we have formulated a spatial probability distribution for moving objects with respect to First-Order predictions, which take into account mobility characteristics and how they relate to probable motion. This is a novel method since the most common approach uses Kalman Filters to estimate future states based upon observed previous states only, assuming a geospatial 2-D Gaussian distribution with monolithic variances in both the normal and tangential directions of motion. Unlike prior approaches, our methodology takes into consideration specific dynamic constraints (e.g., Ackermann Steering), and probable decision making capabilities of the mover. By adding higher levels of fidelity to prediction models, more accurate and precise object tracking, avoidance, or engagement can be accomplished with already developed techniques.
In Artificial Intelligence (AI), utility functions are used to compare the relative goodness of an AI system making one decision over another. These utility functions, along with their coefficients and parameters, comprise a set of beliefs. The term "belief" is used to describe how autonomous systems associate quantified confidences in the existence of things that make up their worldly knowledge; hence, decisions made by an AI system are governed by the states of its belief. Just as in control system theory where the states are quantities used to estimate and describe the behaviors of electrical and/or mechanical systems, belief states are quantities that are used to describe and estimate decision making characteristics.For evolving AI systems (e.g., ones that change their belief states over time through learning) and heterogeneous systems (e.g., ones that do not share a common notion of a global belief), it is important to develop mechanisms that will allow these systems to converge their beliefs, where they can determine how each other "thinks." This is important since it will allow each agent in a cooperating collective to make decisions that optimally solves their individual needs, along with their collective needs, without requiring explicit communication or a mediator. In this paper, methodologies are presented that facilitate belief convergence. The significance of this study is that we demonstrate how a state observer, such as a Kalman filter, can be used by each agent in a collective to estimate neighboring belief states. This is done with no a priori information of their neighbors' belief states, and by only comparing each individual's estimate for the required effort of the whole collective to perform a group plan.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.