A Mobile Agent Cloning Controller for Servicing Networked Robots W.Wilfred Godfrey 1,a and Shivashankar B. Nair 1,b 1 Indian Institute of Technology Guwahati, India a w.godfrey@iitg.ernet.in, b sbnair@iitg.ernet.in Abstract. In this paper we describe how mobile agents that form service providers in a robotic network stigmergically sense the available bandwidth and control their respective cloning rates so as to reduce service times. Apart from a life-time, these agents also have a Cloning Resource which is charged by rewards gained as and when they service the robots. This novel concept of resource embeds an adaptive demand based cloning behavior. The results portrayed in this paper clearly show that the use of such a resource inhibits excessive cloning which otherwise could degrade the efficiency and speed of the service. Keywords: Mobile Agents, Networked Robots, Cloning, Population Control. 1. Introduction Mobile agents are programs which carry their execution state along with data from one node to another within a network. Based on the application, they can be programmed to perform a variety of tasks. In addition to being mobile and autonomous, these agents also have the ability to clone. Creating multiple clones can, in many scenarios, improve robustness, fault tolerance and efficiency [1]. Since clones possess the same capabilities as their originals, they allow for a parallel and hence faster dissemination of the information contained within them. Unregulated cloning can however lead to the creation of a large number of such agents, which in turn can consume the network bandwidth and other related resources resulting in a rapid degradation in performance. It is thus essential to exercise an appropriate and near optimal control on their population lest such a clonal expansion cause the over-populated network to come to a stand-still. In addition, for certain dynamic environments, it may also be essential to increase their population based on demand [2]. In real systems, a centralized mechanism will naturally consume a considerable part of the bandwidth in periodically communicating with the agents (or nodes) within the network, thus degrading performance of the system. Though decentralized or distributed mechanisms [3, 4] have been suggested, they are aimed at a homogenous set of agents and are not on-demand based mechanisms. Decentralized control of a population of heterogeneous agents still remains a challenging task. In this paper, we describe a novel decentralized mechanism to control selective agent populations to cater to a distributed networked robotic application. Armed with a resource for cloning, agents within this network, use stigmergic [5] means to approximate the agent population and autonomously decide whether or not to clone. 2. Networked Robots: Application Scenario Godfrey and Nair [2] have proposed an architecture for networked multi-mobile-robot systems which uses mobile agents to service the robots. They have also suggested the use of a Pheromone- Conscientious (PherCon) migration strategy [6] for mobile agents residing in a robotic node. Each robotic node periodically senses and updates its list of neighbours. In addition, the robotic node also contains a platform that can host mobile agents. Mobile agents carrying services populate the network. A service could be some useful information (such as source code, rules, etc.) required by a robotic node. The network allows users to tether robots without the need to embed, a priori, programs for all tasks they can perform. When a robot is issued a 2011 International Conference on Future Information Technology IPCSIT vol.13 (2011) © (2011) IACSIT Press, Singapore 91