Surveillance Coverage and Vulnerability Awareness Concepts for Tactical Swarms

Currently light infantry soldiers do not have access to many of their cyber resources the moment they depart the forward operating base (FOB). Commanders with recent combat experience have reported on the dearth of computing abilities once a mission is underway [17]. To address this, our group seeks to develop a tactical, mobile cloud implemented on a swarm of semi-autonomous robots. In this paper, we propose a pattern recognition approach to network vulnerability assessment applied to a tactical swarm of robots to enhance their strategy for surveillance coverage. Our work enhances network-enabled persistent surveillance within a dynamic, mobile domain via the implementation of sensor awareness concepts. 1.0 INTRODUCTION The aim of this paper is to present preliminary results obtained with a novel methodology aimed at improving the motion and surveillance strategies of a mobile tactical swarm of robots. The problem of detecting, and maintaining target identification in realistic battlefield conditions is among the most difficult task facing the military today. For autonomous systems performing surveillance tasks, such as the tactical swarms described in this work, one of the major threats lies in the swarm’s individual robot’s lack of self-awareness. In this work we limit our definition self-awareness to inter-robot connectivity. The objective for the swarm then is to perform surveillance in as economical manner as possible. This implies successfully combining the swarm’s goal of dispersing itself over a wide area while reducing unnecessary coverage duplicity at the local node level. Depending on factors such as the local sparseness of the surveillance coverage, terrain complexity and other environmental impediments, the individual robots are at risk of losing contact with the rest of the swarm while performing their surveillance tasks. Furthermore, a robot may be lost accidentally, a lost caused for example by a mechanical failure or enemy action. In some situations, a lost can be benign to the swarm, i.e losing a robot at the external boundary of the swarm. In other Surveillance Coverage and Vulnerability Awareness Concepts for Tactical Swarms PAPER NBR 2 RTO-MP-SET-183-IST-112 UNCLASSIFIED UNCLASSIFIED circumstances the lost can be catastrophic if the lost of a robot breaks the swarm into subcomponents. The general problem addressed in this paper can be casted as a Situation Analysis (SA) problem. The aim of SA in a decision-making process is to provide and maintain a state of situation awareness for an agent observing a scene. For the purpose of the presented research, situation awareness also includes self-awareness. A critical function in the SA process is the real-time recognition of events and situations. More precisely, Situation Recognition is the action of identifying a situation to be something previously known. In the context of the present surveillance swarm monitoring problem, this entails swarm members automatically identifying situations that could pose a connectivity threat to the swarm. The proposed methodology formulates the Situation Recognition problem as a typical pattern recognition problem. The solution then is a classifier designed and trained on a set of features extracted from the swarm network. The labels associated with the set of extracted network features are obtained from extensive simulations and identify vulnerable positions for robots within the swarm under various conditions. The resulting learned configurations are then used to obtain more robust motion and surveillance strategies. 2.0 TACTICAL SWARM FOR SURVEILLANCE AND COMMUNICATION The authors consider a sensing coverage problem by a swarm of robots performing basic surveillance tasks for which each robot is responsible for maintaining acceptable global surveillance coverage for a given area even during network disturbances caused by the robot’s detection tasks. Hence, this robot swarm coverage problem is two-fold: one has to (1) maintain surveillance on a detected target of interest while (2) maintaining the swarm’s global coverage. Thus, the robot swarm must eventually be equipped with a strategy for recovering back to an equilibrium state in reaction to a perturbation. In addition, one will also require that both the original network structure and the robots’ motion strategies are robust enough to absorb small perturbations. 2.1 Problem formalization First, we consider the elements of our domain. Let R={r1, ..., rN} be the set of robots and C={c1, ..., cM} the set of clients. The set C combined with their spatial location is a configuration. We denote ρ as a robot’s unique communication range, which could be adjusted based on environmental demands. Next, E={e1, ..., eN} is the set of communication links between the robots and Gc=(R, E) is the corresponding communication graph. Let Nc, be the set of node coordinates. Two robots ri and rj are separated by a distance of dij and are connected if their distance is less than ρ. In later sections of this work, dij will denote the distance between two robots, two clients, or one robot and one client. We assume that (i) indirect links are possible through intermediate nodes acting as relays, and (ii) at least one of the nodes is connected to an external communication node such as a satellite or UAV. In other words, we assume that there exists a communication resource capability within the network to ensure that clients’ messages are handled properly through the mobile cloud via an external wide range communication relay. Surveillance Coverage and Vulnerability Awareness Concepts for Tactical Swarms RTO-MP-SET-183-IST-112 PAPER NBR 3 UNCLASSIFIED UNCLASSIFIED Figure 1: A tactical mobile cloud for communication coverage. The robot nodes ri are surrounded by light blue circle that represent their coverage. ρ is the communication range and dij is the distance between robot ri and rj. The strategy for the swarm must thus meet the objectives of (1) maintaining the global network connections (i.e., avoid loss of connectivity), (2) ensuring the clients’ coverage (i.e., no client is unconnected) and (3) maintaining a global sensing coverage (i.e., every intrusion within the network will be detected). The overall objective of the swarm of robots is thus threefold: 1. Maintain the network’s connectivity; 2. Maintain the clients’ communication coverage; 3. Maintain an acceptable level of sensing coverage. These three objectives can be expressed in term of coverage as it will be detailed in Section 3.0. The initial state is an equilibrium state in which the three objectives above are satisfied. Given the stochastic nature of the clients’ move as well as of the robots’ performances, this equilibrium state may be weakened, and two major causes of such a weakness are: a) A loss of a node (robot’s failure); b) A loss of a link (caused by a client’s move, or an obstacle to the communication link). We assume that the swarm of robots has a motion strategy to recover that state of equilibrium and ensure the clients’ coverage, the network connectivity and the network sensing coverage. We assume that both the original network structure and the motion strategy are robust enough to absorb small perturbations. But what if these perturbations grow larger? In order to cope with this possible issue, we propose a methodology for network vulnerability assessment used as an early warning mechanism that will allow each network elements (nodes or links) to evaluate the consequences its own loss. The result of this assessment would then be used to modify the motion strategy. 2.2 Motion strategies Each client knows its own location via a location finding device such as a GPS. We model the continuous behavior of clients and robots as a sequence of instances. At each time instance, each robot evaluates the current position of its clients as well as the set of neighboring robots. Based on this evaluation, each robot calculates the optimum position to provide coverage to clients as well as to maintain connectivity to at least one neighboring robot. If both goals cannot be attained, then the priority is to maintain coverage for clients. Surveillance Coverage and Vulnerability Awareness Concepts for Tactical Swarms PAPER NBR 4 RTO-MP-SET-183-IST-112 UNCLASSIFIED UNCLASSIFIED At the initial equilibrium state, the network optimally covers the set of clients (communication coverage), optimally covers the area under consideration (sensing coverage) and is fully connected (each client is able to communication with an external node relaying the communication). That means in particular that (1) each client is within the communication range of at least one robot (clients’ coverage) and (2) each robot is within the communication range of at least one other robot (network’s connectivity). Our robot motion strategy is based on the work in [2]. Each robot in the swarm moves according to spring-mass virtual physics. In particular, the motion of the i node in a swarm of N robots is as follows: