Fig 1 - uploaded by Gianni A. Di Caro
Content may be subject to copyright.
Source publication
Cooperative Multi-Robot Observation of Multiple Moving Targets (CMOMMT) denotes a class of problems in which a set of autonomous mobile robots equipped with limited-range sensors are used to keep under observation a (possibly larger) set of mobile targets. Robots cooperatively plan their motion in order to maximize the time during which each target...
Context in source publication
Context 1
... probabilistic predictions about targets' motion are fed into the ILP model for the next planning stage and provide the core information to the planner to decide where to let the robots move for the next h steps. Figure 1 illustrates the overall way of operating of the system. Since the planner aims to select paths that let the robots intercept targets, it is necessary to express how good, in terms of optimizing system's monitoring and fairness performance, it would be for a robot a to be in a specific cell n at a future time t. ...
Similar publications
This article presents a miniaturized four-wheel robotic vehicle for autonomous driving research. It enables experiments in situations where testing with full-size vehicles becomes dangerous or troublesome. The article describes the physical structure of the vehicle, its actuators and sensing capabilities and the ROS-based software architecture. The...
The development of an android with convincingly lifelike appearance and behavior has been a long-standing goal in robotics, and recent years have seen great progress in many of the technologies needed to create such androids. However, it is necessary to actually integrate these technologies into a robot system in order to assess the progress that h...
Citations
... In addition to the metrics based on spatial distribution and probability just discussed, there are a few other metrics used by various works that do not fit into these two broad categories. For example, when carrying out a search for multiple targets under the Cooperative Multi-Robot Observation of Multiple Moving Targets (CMOMMT) framework, Banfi et al. (2015) introduced a metric known as the tracking fairness that measures the monitoring effort of the MRS across all targets. Should a target receive more attention from the MRS' agents, the system would have a high tracking fairness score, indicating over-monitoring of a target or too much exploitation being carried out by the system. ...
... To develop a tracking strategy for the same CMOMMT problem using deep reinforcement learning Yan, Jia, and Bai (2021) used an occupancy grid map (OGM) to characterise the environment. A tracking fairness based metric was then used, comparable to the one proposed by Banfi et al. (2015), as well as an exploration metric based on the latest observation time of all grid cells in the OGM. ...
Can a multi-robot system (MRS) be made to find and track a target that can move faster than any of its component robots? Such tasks have long been considered impossible due to the assumption that the target will always be able to outrun the individual robots. The work done in this thesis shows that this task is in fact, achievable and at its root, boils down to the exploration-exploitation dilemma---the choice a system must make between gathering more information about the environment or making use of the information currently available.
To accomplish the task of tracking a fast-moving non-evasive target, a fully decentralised search and track strategy based on the Particle Swarm Optimisation (PSO) algorithm is used. This strategy is complemented by an adaptive inter-agent repulsion behaviour, used to promote exploration, as well as an adjustable $k$-nearest neighbour communications network, used to tune the system's exploration-exploitation balance. To achieve the more challenging task of tracking an evasive target, the individual agents of the swarming MRS are endowed with a short-term memory, thereby promoting higher levels of exploitation. The two developed strategies are then validated through physical tests using a decentralised swarm of miniature ground robots.
Through both virtual and physical experimentation, an optimum level of connectivity to maximise the MRS's tracking performance is revealed. The origin of this optimum level of connectivity is further traced back to an optimum balance in the amount of exploratory and exploitative actions carried out by the system. The effect of various environmental factors and mission parameters on this optimum, such as the swarm density, number of agents used, and the movement profile of the targets, are also studied. The results presented in this thesis further emphasises the importance of attaining the correct exploration-exploitation balance when developing a swarm strategy. This optimum changes according to the task set for the system and the results presented shed some light on how to tune a swarm's exploration-exploitation dynamics to find this optimum, potentially paving the way for better swarming algorithms to be developed.
... In their CMOMMT task Banfi et al. (2015), introduced a metric known as the tracking fairness that measures the monitoring effort of the MRS across all targets. Should a target receive more attention from the MRS's agents, the system would have a high tracking fairness score, indicating over-monitoring of a target or too much exploitation being carried out by the system. ...
... To develop a tracking strategy for the same CMOMMT problem using deep reinforcement learning Yan et al. (2021) used an occupancy grid map (OGM) to characterize the environment. A tracking fairness based metric was then used, comparable to the one proposed by Banfi et al. (2015), as well as an exploration metric based on the latest observation time of all grid cells in the OGM. ...
Multi-agent systems and multi-robot systems have been recognized as unique solutions to complex dynamic tasks distributed in space. Their effectiveness in accomplishing these tasks rests upon the design of cooperative control strategies, which is acknowledged to be challenging and nontrivial. In particular, the effectiveness of these strategies has been shown to be related to the so-called exploration–exploitation dilemma: i.e., the existence of a distinct balance between exploitative actions and exploratory ones while the system is operating. Recent results point to the need for a dynamic exploration–exploitation balance to unlock high levels of flexibility, adaptivity, and swarm intelligence. This important point is especially apparent when dealing with fast-changing environments. Problems involving dynamic environments have been dealt with by different scientific communities using theory, simulations, as well as large-scale experiments. Such results spread across a range of disciplines can hinder one’s ability to understand and manage the intricacies of the exploration–exploitation challenge. In this review, we summarize and categorize the methods used to control the level of exploration and exploitation carried out by an multi-agent systems. Lastly, we discuss the critical need for suitable metrics and benchmark problems to quantitatively assess and compare the levels of exploration and exploitation, as well as the overall performance of a system with a given cooperative control algorithm.
... In Refs. [26,27], the authors present a novel optimization model for CMOMMT scenarios which features fairness of observation among different targets as an additional objective.The authors in Ref. [28] extend the conventional CMOMMT problem with limited sensing range and the moving targets are un-directional. In Ref. [29], the authors incorporate a multi-hop clustering and a dual-pheromone ant-colony model to optimize the target detection and tracking problem. ...
This paper presents a novel distributed algorithm for a moving targets search with a team of cooperative unmanned aerial vehicles (UAVs). UAVs sense targets using on-board sensors and the information can be shared with teammates within a communication range. Based on local and shared information, the UAV swarm tries to maximize its average observation rate on targets. Unlike traditional approaches that treat the impact from different sources separately, our framework characterizes the impact of moving targets and collaborating UAVs on the moving decision for each UAV with a unified metric called observation profit. Based on this metric, we develop a profit-driven adaptive moving targets search algorithm for a swarm of UAVs. The simulation results validate the effectiveness of our framework in terms of both observation rate and its adaptiveness.
... An early version of this work has appeared in Banfi et al. (2015). This paper significantly extends those preliminary results with a revised ILP model (now allowing to specify an arbitrary replanning time), a more in-depth discussion of the CMFMT performance metrics and of their correlation with the ILP model, new simulation results and analyses, comparisons with other methods, scenarios with obstacles, and experiments with real robots. ...
... Finally, note that a variation of the above ILP model could in principle be able to take collision avoidance constraints Fig. 2 The operational model of the robot system. Image taken from Banfi et al. (2015) into account, but at the expenses of an increase in the model complexity; see Yu and LaValle (2016), Banfi et al. (2017). ...
Cooperative Multi-Robot Observation of Multiple Moving Targets (CMOMMT) denotes a class of problems in which a set of autonomous mobile robots equipped with limited-range sensors keep under observation a (possibly larger) set of mobile targets. In the existing literature, it is common to let the robots cooperatively plan their motion in order to maximize the average targets’ detection rate, defined as the percentage of mission steps in which a target is observed by at least one robot. We present a novel optimization model for CMOMMT scenarios which features fairness of observation among different targets as an additional objective. The proposed integer linear formulation exploits available knowledge about the expected motion patterns of the targets, represented as a probabilistic occupancy maps estimated in a Bayesian framework. An empirical analysis of the model is performed in simulation, considering multiple scenarios to study the effects of the amount of robots and of the prediction accuracy for the mobility of the targets. Both centralized and distributed implementations are presented and compared to each other evaluating the impact of multi-hop communications and limited information sharing. The proposed solutions are also compared to two algorithms selected from the literature. The model is finally validated on a real team of ground robots in a limited set of scenarios.
... In some real world situations, like wildlife research, search and detection, and crowded and social movement [11], one or more moving entities need to be continuously maintained under observation or surveillance by other moving entities. Currently, the observers entities are Unnamed Aerial Vehicles (UAV), Unnamed Ground Vehicles (UGV), Unnamed Underwater Vehicles (UUV) [2], equipped with sensing, processing, and communication skills. Generally, these vehicles have limited resources, like low-range visibility, communication difficulties and many entities to be observed for few observer entities. ...
... Robot formation [65] Model-predictive control C [66] Clustering Dc [67] Integer linear program C [68], [69] Un Force vectors [76] UAV Differential game [77] < 1 UGV Task-scheduling C [78] Negotiation and auctions D In case of platforms with local processing power, each robot iteratively executes four actions ( Fig. 2): taking observations and processing the data locally; exchanging information with other robots (and the ground station); planning its motion; and generating control actions to execute its physical motion according to the plan. Once the robots have moved, the current location information and the new observation from the sensor initiate a new iteration of this processing cycle. ...
... The work compares K-means clustering and hill-climbing algorithms, which are scalable in degree of decentralization, for achieving the objective of CMOMMT. The expected motion patterns of the targets can be exploited to observe each target for an equal amount of time [67]. ...
The deployment of multiple robots for achieving a common goal helps to improve the performance, efficiency, and/or robustness in a variety of tasks. In particular, the observation of moving targets is an important multirobot application that still exhibits numerous open challenges, including the effective coordination of the robots. This paper reviews control techniques for cooperative mobile robots monitoring multiple targets. The simultaneous movement of robots and targets makes this problem particularly interesting, and our review systematically addresses this cooperative multirobot problem for the first time. We classify and critically discuss the control techniques: cooperative multirobot observation of multiple moving targets, cooperative search, acquisition, and track, cooperative tracking, and multirobot pursuit evasion. We also identify the five major elements that characterize this problem, namely, the coordination method, the environment, the target, the robot and its sensor(s). These elements are used to systematically analyze the control techniques. The majority of the studied work is based on simulation and laboratory studies, which may not accurately reflect real-world operational conditions. Importantly, while our systematic analysis is focused on multitarget observation, our proposed classification is useful also for related multirobot applications.
Intelligent robotic systems are becoming ever more present in our lives across a multitude of domains such as industry, transportation, agriculture, security, healthcare and even education. Such systems enable humans to focus on the interesting and sophisticated tasks while robots accomplish tasks that are either too tedious, routine or potentially dangerous for humans to do. Recent advances in perception technologies and accompanying hardware, mainly attributed to rapid advancements in the deep-learning ecosystem, enable the deployment of robotic systems equipped with onboard sensors as well as the computational power to perform autonomous reasoning and decision making online. While there has been significant progress in expanding the capabilities of single and multi-robot systems during the last decades across a multitude of domains and applications, there are still many promising areas for research that can advance the state of cooperative searching systems that employ multiple robots. In this article, several prospective avenues of research in teamwork cooperation with considerable potential for advancement of multi-robot search systems will be visited and discussed. In previous works we have shown that multi-agent search tasks can greatly benefit from intelligent cooperation between team members and can achieve performance close to the theoretical optimum. The techniques applied can be used in a variety of domains including planning against adversarial opponents, control of forest fires and coordinating search-and-rescue missions. The state-of-the-art on methods of multi-robot search across several selected domains of application is explained, highlighting the pros and cons of each method, providing an up-to-date view on the current state of the domains and their future challenges.
