Plamen V. Petrov

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.

An intelligent-agent based decision support system for a complex command and control application

The authors present an architectural overview of an agent based decision support environment. We have selected the domain of AWACS Command and Control, in which human controllers need to make critical decisions under strict timing constraints in a dynamically changing environment. The decision support training environment is based on distributed simulation, tightly coupled with an intelligent agent infrastructure. The agents apply heuristics based algorithms to provide decision support to the human controllers.

A multi-agent UAV swarm for automatic target recognition

We address the problem of automatic target recognition (ATR) using a multi-agent swarm of unmanned aerial vehicles(UAVs) deployed within a reconnaissance area. Traditionally, ATR is performed by UAVs that fly within the reconnaissance area to collect image data through sensors and upload the data to a central base 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, susceptibility of the system to malicious attacks on the central location. In this paper, we describe a multi-agent system of UAVs to perform ATR. We assume that each UAV has limited computational capabilities and target identification can be performed by several UAVs that combine their resources including their computational capabilities. The UAVs employ a swarming algorithm implemented through software agents to congregate at and identify potential targets, and, a gossiping mechanism to disseminate information within the swarm.

A language support environment for complex distributed real-time applications

Engineering of complex distributed real-time applications is one of the hardest tasks faced by the software profession today. All aspects of the process, from design to implementation, are made more difficult by the interaction of behavioral and platform constraints. Providing tools for this task is likewise not without major challenges. In this paper, we discuss a tool suite at New Jersey Institute of Technologys Real-Time Computing Lab which supports the development of complex distributed real-time applications in a suitable high-level language (CRL). The suites component tools include a compiler, a transformer-optimizer, an allocator-migrator, schedulability analyzers, a debugger-monitor, a kernel, and a (simulated) network manager. The overall engineering approach supported by the suite is to provide as simple and natural an integrated development paradigm as possible. The suite tools address complexity due to distribution, scheduling, allocation and other sources in an integrated manner (largely) transparent to the developer. To reflect the needs of propagation of functional and non-functional requirements throughout the development process, a number of robust code transformation and communication mechanisms have been incorporated into the suite. To facilitate practical use of the suite, the developed programs compile-transform to a safe subset of C++ with appropriate libraries and runtime support.

Collective agents interpolative integral (CAII) for asymmetric threat detection

This paper presents a reasoning system that pools the judgments from a set of inference agents with information from heterogeneous sources to generate a consensus opinion that reduces uncertainty and improves knowledge quality. The system, called Collective Agents Interpolation Integral (CAII), addresses a high level data fusion problem by combining, in a mathematically sound manner, multi-models of inference in knowledge intensive multi agent architecture. Two major issues are addressed in CAII. One is the ability of the inference mechanisms to deal with hybrid data inputs from multiple information sources and map the diverse data sets to a uniform representation in an objective space of reasoning and integration. The other is the ability of the system architecture to allow the continuous and discrete outputs of a diverse set of inference agents to interact, cooperate, and integrate.

Statistical Fusion of Unmanned Aerial Vehicle Observations for Aided Target Recognition

The use of computers and sensors to detect and classify targets, often called aided target recognition (ATR), is an important component of military and civilian surveillance. Carrying it out from unmanned aerial vehicles is expensive in terms of both manpower and hardware. In this paper, we discuss the creation of a distributed ATR (DATR) method which replaces a single monolithic approach to ATR with a more robust multi-agent method. We present software and algorithms which have been developed for the purpose of testing and proving the validity of DATR, as well as some of the implementation possibilities and steps which we have taken to further prove the validity of the approach. Our examples and experiments with the normal light-weight ATR algorithms for our agents, combined with a fusion method based on belief calculus to demonstrate what DATR would perform like, show the validity of the approach

Video enhancement effectiveness for target detection

Unmanned aerial vehicles (UAVs) capture real-time video data of military targets while keeping the warfighter at a safe distance. This keeps soldiers out of harms way while they perform intelligence, surveillance and reconnaissance (ISR) and close-air support troops in contact (CAS-TIC) situations. The military also wants to use UAV video to achieve force multiplication. One method of achieving effective force multiplication involves fielding numerous UAVs with cameras and having multiple videos processed simultaneously by a single operator. However, monitoring multiple video streams is difficult for operators when the videos are of low quality. To address this challenge, we researched several promising video enhancement algorithms that focus on improving video quality. In this paper, we discuss our video enhancement suite and provide examples of video enhancement capabilities, focusing on stabilization, dehazing, and denoising. We provide results that show the effects of our enhancement algorithms on target detection and tracking algorithms. These results indicate that there is potential to assist the operator in identifying and tracking relevant targets with aided target recognition even on difficult video, increasing the force multiplier effect of UAVs. This work also forms the basis for human factors research into the effects of enhancement algorithms on ISR missions.

Semantic bifurcated importance field visualization

While there are many good ways to map sensual reality to two dimensional displays, mapping non-physical and possibilistic information can be challenging. The advent of faster-than-real-time systems allow the predictive and possibilistic exploration of important factors that can affect the decision maker. Visualizing a compressed picture of the past and possible factors can assist the decision maker summarizing information in a cognitive based model thereby reducing clutter and perhaps related decision times. Our proposed semantic bifurcated importance field visualization uses saccadic eye motion models to partition the display into a possibilistic and sensed data vertically and spatial and semantic data horizontally. Saccadic eye movement precedes and prepares decision makers before nearly every directed action. Cognitive models for saccadic eye movement show that people prefer lateral to vertical saccadic movement. Studies have suggested that saccades may be coupled to momentary problem solving strategies. Also, the central 1.5 degrees of the visual field represents 100 times greater resolution that then peripheral field so concentrating factors can reduce unnecessary saccades. By packing information according to saccadic models, we can relate important decision factors reduce factor dimensionality and present the dense summary dimensions of semantic and importance. Inter and intra ballistics of the SBIFV provide important clues on how semantic packing assists in decision making. Future directions of SBIFV are to make the visualization reactive and conformal to saccades specializing targets to ballistics, such as dynamically filtering and highlighting verbal targets for left saccades and spatial targets for right saccades.

Taking Intelligent Agents to the Battlefield

The battlefield is a place of violence ruled by uncertainty. Timely knowledge of whats happening around a soldier can mean the difference between life and death. The goals of an enhanced mobile infantry are becoming a reality through innovative progress by the U.S. Armys 21st Century Land Warrior (LW) program. However, the current system does not provide a head up display capability like that provided by todays avionics. When the soldier employs the weapon, he should see objects easily distinguishable as friendly or not, as well as enemy locations. The Eyekon project is an intelligent agent-based decision support system hosted on the soldiers wearable computer. Eyekon will use the LW network to provide a perspective view in the weapon sight. This will naturally draw the warrior to the most desirable target. There are many performance and human factors issues to address before the concept can be used in lethal situations.

Using XML for Interprocess Communications in a Space Situational Awareness and Control Application

As various militaries progress toward a network-centric environment, their dependency upon space assets is expected to grow exponentially. It is important for the commander to be aware of the dispositions and movements in this fourth medium. The success of terrestrial forces requires comprehensive space situational awareness (SSA). 21st Century Systems, Inc. is essentially developing a SSA and decision support application that, with proper inputs, will provide an integrated space picture. We call this concept, SituSpace. The visualization uses two and three dimensional depictions to provide the watchstander with rapid situational awareness. SituSpace employs intelligent software agents to provide timely alerts and cogent recommendations. The intelligent agents apply the user’s ROE, situation, resources, etc. to rapidly derive actionable recommendations to accelerate the decision loop. The SituSpace concept gains synergy from combining two complementary ideas: using a metalanguage as the medium of storage and interprocess communication and a modular publish-subscribe software architecture.