Coordination of Multiple Autonomous Agents Using Naturally Generated Languages in Task Planning

Abstract

Language plays a prominent role in the activities of human beings and other intelligent creatures. One of the most important functions of languages is communication. Inspired by this, we attempt to develop a novel language for cooperation between artificial agents. The language generation problem has been studied earlier in the context of evolutionary games in computational linguistics. In this paper, we take a different approach by formulating it in the computational model of rationality in a multi-agent planning setting. This paper includes three main parts: First, we present a language generation problem that is connected to state abstraction and introduce a few of the languages' properties. Second, we give the sufficient and necessary conditions of a valid abstraction with proofs and develop an efficient algorithm to construct the languages where several words are generated naturally. The sentences composed of words can be used by agents to regulate their behaviors during task planning. Finally, we conduct several experiments to evaluate the benefits of the languages in a variety of scenarios of a path-planning domain. The empirical results demonstrate that our languages lead to reduction in communication cost and behavior restriction.common way for agents is to send a feasible plan to their teammates to be followed when cooperation is needed. The objective of this work is to construct a kind of language that can be used to specify plans, while not bringing many constraints for agents and reducing the cost of communication as much as possible.The contributions of this work have three aspects: First, we formulate a language generation problem for multi-agent systems and introduce some fundamental features of the language that appear in agent coordination; Second, we give sufficient and necessary conditions of valid abstraction, based on which an efficient algorithm is developed to generate languages that are complete and optimal. Third, we apply the algorithm to dozens of environments and compare the advantages of languages generated by our algorithm with other similar languages in a path-planning domain.The rest of the paper is organized as follows. We review the related work in Section 2. Section 3 introduces the background of multi-agent planning and presents the problem formulation of language generation. Section 4 provides the conditions of valid abstraction and describes a language generation algorithm. Section 5 implements the algorithm and evaluates the languages. Conclusions and future work are given in Section 6. Related WorkCommunication is one of the most basic and important issues in multi-agent coordination. Depending on how information is obtained, communication can be divided into implicit mechanisms, for example, pheromone [12], in which agents acquire information about their teammates through the world, and explicit mechanisms, in which agents directly transmit information through media, such as spoken languages. Explicit communication is often used in intentionally multi-agent systems, since...

A Dimensional Comparison between Evolutionary Algorithm and Deep Reinforcement Learning Methodologies for Autonomous Surface Vehicles with Water Quality Sensors

“…The EA takes advantage of bio-inspired evolutionary foundations to synthesize increasingly optimal paths within every generation. Several authors deal with high dimensional planning problems in vehicles [ 2 , 7 , 30 , 31 , 32 ]. In [ 30 ], the particular case of the Ypacaraí area coverage is treated as a Traveling Salesman Problem (TSP).…”

“…Ref. [ 32 ] also addresses the path planning via a Natural Language Generation (NGL) using an Evolutionary methodology for many agents to complete different tasks. This work also addressed the dimensionality by comparing the number of the possible states and the computation cost.…”