Prithviraj Dasgupta

A Multiagent Swarming System for Distributed Automatic Target Recognition Using Unmanned Aerial Vehicles

Over the past few years, automatic target recognition (ATR) has emerged as an essential image analysis tool to identify objects from temporally and spatially disjoint possibly noisy image data. For many current applications, ATR is performed by unmanned aerial vehicles (UAVs) that fly within a reconnaissance area to collect image data through sensors and upload the data to a central ground control station for analyzing and identifying potential targets. The centralized approach to ATR introduces several problems, including scalability with the number of UAVs, network delays in communicating with the central location, and the susceptibility of the system to malicious attacks on the central location. These challenges can be addressed by using a distributed system for performing ATR. In this paper, we describe a multiagent-based prototype system that uses swarming techniques inspired from insect colonies to perform ATR using UAVs in a distributed manner within simulated scenarios. We assume that UAVs are constrained in the resources available onboard and in their capabilities for performing ATR due to payload limitations. Our focus in this paper is on the coordination aspects between UAVs to efficiently decide how they are to act by using a swarming mechanism. We describe algorithms for the different operations performed by the UAVs in the system and for different swarming strategies, which are embedded within software agents located on the UAVs. We provide empirical simulations of our system within a simulated area of interest to determine its behavior in different scenarios with varying operational constraints. Our experimental results indicate that swarming strategies for distributed ATR perform favorably compared with centralized ATR strategies.

Reciprocal resource sharing in P2P environments

Peer-to-peer (P2P) systems enable users to share resources in a networked environment without worrying about issues such as scalability and load balancing. Unlike exchange of goods in a traditional market, resource exchange in P2P networks does not involve monetary transactions. This makes P2P systems vulnerable to problems including the free-rider problem that enables users to acquire resources without contributing anything, collusion between groups of users to incorrectly promote or malign other users, and zero-cost identity that enables nodes to obliterate unfavorable history without incurring any expenditure. Previous research addresses these issues using user-reputation, referrals, and shared history based techniques. Here, we describe a multi-agent based reciprocity mechanism where each users agent makes the decision to share a resource with a requesting user based on the amount of resources previously provided by the requesting user to the providing user and globally in the system. A robust reputation mechanism is proposed to avoid the differential exploitations by the free-riders and to prevent collusion. Experimental results on a simulated P2P network addresses the problems identified above and shows that users adopting the reciprocative mechanism outperform users that do not share resources in the P2P network. Hence, our proposed reciprocative mechanism effectively suppresses free-riding.

ModRED: Hardware design and reconfiguration planning for a high dexterity modular self-reconfigurable robot for extra-terrestrial exploration

Abstract This paper presents a homogeneous modular robot system design based on four per-module degrees of freedom (DOF), including a prismatic DOF to increase the versatility of its reconfiguration and locomotion capabilities. The ModRED (Modular Robot for Exploration and Discovery) modules are developed with rotary-plate genderless single sided docking mechanisms (RoGenSiD) that allow chain-type configurations and lead towards hybrid-type configurations. Various locomotion gaits are simulated through the Webots robot simulator and implemented in the real ModRED system. This work also addresses the problem of dynamic reconfiguration in a modular self-reconfigurable robot (MSR). The self-reconfiguration problem is modeled as an instance of the graph-based coalition formation problem. We formulate the problem as a linear program that finds the “best” partition or coalition structure among a set of ModRED modules. The technique is verified experimentally for a variety of settings on an accurately simulated model of the ModRED robot within the Webots robot simulator. Our experimental results show that our technique can find the best partition with a reasonably low computational overhead.

Auction-based multi-robot task allocation in COMSTAR

Over the past few years, swarm based systems have emerged as an attractive paradigm for building large scale distributed systems composed of numerous independent but coordinating units. In our previous work, we have developed a protoype system called COMSTAR (Cooperative Multi-agent Systems for automatic TArget Recognition) using a swarm of unmanned aerial vehicles(UAVs) that is capable of identifying targets in software simulations of reconnaissance operations. Experimental results from the simulations of the COMSTAR system show that task selection among the UAVs is a crucial operation that determines the overall efficiency of the system. Previously described techniques for task selection among swarm units use a centralized server such as a ground control station to coordinate the activities of the swarm units. However, such systems are not truly distributed since the behavior of the swarm units is predominantly directed by the centralized servers task allocation algorithm. In this paper we focus on the problem of distributed task selection in a swarmed system where each swarm unit decides on the tasks it will execute by sharing information and coordinating its actions with other swarm units without the intervention of a centralized ground control station supervising its activities. Specifically, we build our task selection algorithm on an auction-based algorithm for task selection in robotic swarms described by Kalra et al. We report experimental results in a simulated environment with 18 robots and 20 tasks and compare the performance of our auction-based algorithm with other heuristic-based task selection strategies in multi-agent swarms. Our simulation results show that the auction-based algorithm improves the task completion times by 30-60% and reduces the communication overhead by as much as 90% with respect to other heuristic-based strategies maintaining similar performance in load balancing.

Improving peer-to-peer resource discovery using mobile agent based referrals

A peer-to-peer(P2P) system consists of a decentralized, distributed network of nodes that is capable of sharing resources and services without centralized supervision. A major functionality in P2P networks is locating resources or services present on remote nodes. Traditional techniques for resource discovery include blind searches among the nodes using query flooding, or, positioning resources strategically to enable rapid lookup using distributed hash tables. In this paper, we propose a mobile agent based referral service that directs resources queries intelligently towards the nodes possessing the resource. The mobile agents adaptively learn paths called trails within the P2P network to enable rapid location of resources. Preliminary results of our algorithm demonstrate that the agent based technique performs favorably with existing P2P resource discovery protocols.

Multi-Attribute Dynamic Pricing for Online Markets Using Intelligent Agents

Intelligent agents called pricebots provide a convienient mechanism for implementing automated dynamic pricing algorithms for sellers in an online economy. Pricebots enable an online seller to dynamically calculate a competitive price for a product in response to variations in market parameters such as competitorsý prices and consumersý purchase preferences. Previous research on pricebot mediated pricing makes certain simplifying assumptions of online markets such as providing sellers with complete knowledge of market parameters to facililate calculations by the dynamic pricing algorithm, and, considering product price as the only attribute that determines consumersý purchase decision. In this paper, we address the problem of dynamic pricing in a competitive online economy where a product is differentiated by buyers and sellers on multiple attributes, and, sellers possess limited knowledge about market parameters. A seller uses a collaborative filtering algorithm to determine temporal consumersý purchase preferences followed by a dynamic pricing algorithm to determine a competitive price for the product. Simulation results using our market model show that collaborative filtering enabled dynamic pricing techniques compare favorably against other dynamic pricing algorithms. Collaborative filtering enables sellers to rapidly identify temporal customer preferences and improve sellersý profits.

Multi-Robot Task Allocation for Performing Cooperative Foraging Tasks in an Initially Unknown Environment

We consider the problem of multi-robot task allocation (MRTA) by a set of robots that are deployed in an initially unknown environment to perform foraging tasks. The location of each task has to be discovered by each robot through searching in the environment. Each task also requires multiple robots to share the task’s execution to complete the task. We discuss an emergent, distributed MRTA algorithm and describe a set of heuristics that can be used by the MRTA algorithm to select the order of the tasks so that the tasks are performed in an efficient manner. The heuristics are evaluated using simulated robots on the Webots simulator to analyze their relative performance.

Dynamic Area Coverage using Faulty Multi-Agent Swarms

We consider the problem of distributed coverage of an unknown two-dimensional environment using a swarm of mobile mini-robots. In contrast to previous approaches for robotic area coverage, we assume that each robot (agent) in our system is limited in its communication range and memory capacity. Agents are also susceptible to sensor noise while communicating with other agents, and, can be subject to transient or permanent failures. The objective of the agents is to cover the entire environment while reducing the coverage time and the redundancy in the area covered by different agents. First, we describe our distributed coverage algorithm where each agent uses a local heuristic based on Manhattan distances and the information gained from other agents at each step to decide on its next action (movement). We then describe and analyze the fault model of our agents and show that the local heuristic used by the agents deteriorate linearly as the communication noise increases. Finally, we verify the performance of our system empirically within a simulated environment and show that the system is able to scale efficiently in the number of robots and size of the environment, and, determine the effect of communication faults and robot failures on the system performance.

Dynamic Pricing for Time-Limited Goods in a Supplier-Driven Electronic Marketplace

Existing e-commerce systems employ a pull model of marketing where buyers, possibly through agents, search the e-market for suppliers offering the product of their choice. In contrast, the push model where suppliers’ agents approach buyers with their products, has been relatively less investigated. Push strategies are particularly appropriate for commodities that have a short shelf-life and, therefore, an elastic demand curve, allowing suppliers to exploit unexpected supply. The speed and low cost of e-commerce makes it particularly suited to the push paradigm. In this paper, we consider time-limited goods in a supplier driven marketplace that employs the push model of marketing. When constrained by a strict deadline to sell the good, the supplier uses a mobile sales agent that visits every buyer and estimates the short run demand curve of the good. At every buyer, the sales agent also employs a heuristic technique called the Maximum Returns Algorithm to recalculate the price of the good, so that the supplier can obtain the best possible gross returns from trading with the buyers. On the other hand, when the deadline to sell is not stringent, the sales agent negotiates the exchange at a point that improves both the buyer’s utility and the supplier’s profit, as compared to the exchange point without negotiation.

Hardware Design and Testing of ModRED: A Modular Self-Reconfigurable Robot System

Unstructured environments are challenging for conventional robots, and modular self-reconfigurable robots (MSRs) can be deployed to overcome this challenge. The goal of the current work was to develop a flexible, cost effective multi-module robot system capable of self-reconfiguration and achieving various gaits in unstructured environments. This paper discusses the communication aspects of the Modular Robot for Exploration and Discovery (ModRED) robot system from a hardware perspective. To ensure enhanced flexibility and local autonomy as well as better reconfiguration, each robot module is built with four independent degrees of freedom, and a novel docking interface provides interconnection of modules. The prototyping effort is described with emphasis on the implementation of inter-module communication. The electronic hardware layout and control system are described, and the communication system is outlined. Finally, some preliminary testing of the developed prototype is presented.