Swarm robotics |
|
|
The research of swarms in Multi-robot Systems group aims to integrate principles and theoretical background of swarm behaviours with methodology/theory describing cooperative localization of autonomous robots and principles of self-organizing adaptation leading to a flexible stand-alone system. It will enable applicability of swarm robotics in realistic outdoor scenarios of surveillance and reconnaissance. Basically, we develop principles of a decentralized relative localization of neighboring particles that are integrated to swarm behaviors with an aim to keep reciprocal visibility between neighbors. This enables to employ swarms of Micro Aerial Vehicle (MAV) outside laboratories equipped by a precise positioning system.
Besides, a concept of adaptively evolving swarm behaviors is established to decrease relative localization uncertainty. To enable multi-robot applications, theoretical principles of determining desired shapes of MAV swarms are designed based on bio-inspired methods of artificial intelligence, namely Particle Swarm Optimization and Boids models. Finally, decentralized collective decision making mechanisms are established with a theory identifying necessary assumptions of the switching between different swarm behaviors. This research is aimed at a study of observed autonomous behaviors of MAV swarms. In addition to the development of the theory of coordinated motion of swarm members, our research also concerns development of the sensory equipment needed for real world swarm flights, and subsequent integration of the observational constraints it induces on the swarm control in use. The research conducted in this stream is closely coordinated with research of multi robot systems being also realised within Multi-robot Systems group. Our state-of-the-art approaches in research of swarm robotics and description of developed methods can be found in papers chronologically listed below. |
|
Video material for RA-L 2020 paper: Autonomous Aerial Swarm in a Complex Environment without GNSS and without Communication |
|
Real-world experiments and method description
An approach for autonomous compact flocking of relatively localized Unmanned Aerial Vehicles (UAVs) in high obstacle density areas is presented in this paper. The UAV agents have no global localization system and only use on-board sensors to estimate the relative position of other agents in their local reference frame. They do not communicate any information and use a bio-inspired control law to avoid collisions with surrounding obstacles and other agents, while flocking through the environment. This is, to the best of our knowledge, the first such attempt where neither communication nor an external source of localization is used for flocking, making it the first truly distributed swarm flight in a natural environment with obstacles. Experiments from a natural forest validate the decentralized flocking behavior of the swarm and the usability of on-board relative localization for autonomous multi-UAV applications in the absence of global localization information and communication. |
Simulation experiments3D world model of the forest is publicly available at https://nasmrs.felk.cvut.cz/index.php/s/NARxc8nEzgMice6. |
Examples of research results |
|
Fully autonomous UAV swarm in a forest without GPS or communication (video featured in IEEE Robotics Video Fridays).We leveraged LIDAR-based slam, in conjunction with our specialized relative localization sensor UVDAR to perform a de-centralized, communication-free swarm flight without the units knowing their absolute locations. The swarming and obstacle avoidance control is based on a modified Boids-like algorithm, while the whole swarm is controlled by directing a selected leader unit. This video is to our knowledge first demonstration of such swarm operating in obstacle-cluttered environment outside of laboratory conditions, marking a next step in the research of real-world deployment of robotic swarms. |
|
Bio-inspired compact swarms of MAVs without communication and external localization (video featured in IEEE Robotics Video Fridays).State of the art in the field of swarm robotics lacks systems capable of absolute decentralization and is hence unable to mimic complex biological swarm systems consisting of simple units. Our research interconnects fields of swarm robotics and computer vision, and introduces novel use of a vision-based method UVDAR for mutual localization in swarm systems, allowing for absolute decentralization found among biological swarm systems. The developed methodology allows us to deploy real-world aerial swarming systems with robots directly localizing each other instead of communicating their states via a communication network, which is a typical bottleneck of current state of the art systems. |
Coherent swarming of UAVs with minimum computational and communication requirements.The presented algorithm enables a swarm of MAVs to maintain its coherence and perform compact motion in complex environments while avoiding obstacles in a decentralized way. Feasible and collision free control inputs are computed onboard of each vehicle using only limited sensory information without any requirement for external infrastructure, such as global navigation systems. The proposed method robustly manages to incomplete sensory information and it is highly scalable, since increasing number of MAVs even improves the required coherence behavior. Statistical tests in various robotic simulators and environments were conducted to analyze the algorithm performance in its different configurations. The tests also sought to verify the algorithm's reliability, taking into account limitations of designed systems of mutual localization of swarm members, which is a crucial tool required by multi-robot systems and especially by swarms sharing the same workspace. |
MAV group forming a distributed sensory array (flying adaptive “antennas”).A realistic simulation of a compact self-stabilized MAV group forming a distributed sensory array to illustrate the proposed concept of flying adaptive “antennas”. The flexible sensory array adapts its shape to move through an environment with obstacles. The realistic simulations (including MAV dynamics and interaction with environment) were prepared in Gazebo using the multi-MAV control system successfully deployed by our team in the MBZIRC competition. |
Decentralized self-organizing swarming of UAVs based on extended Boids model.The video demonstrates a system proposed for stabilization of a swarm of unmanned and fully autonomous helicopters, using an expanded swarming model Boids. Its main focus lies in a proposal of robust and decentralized swarming behavior suited for complex environments with high density of obstacles, and its relatively straightforward integration to a swarm of real helicopters. Corresponding constraints of multi-robot systems working in real time had to be considered. The capability of the swarm of unmanned helicopters to cluster and navigate in complex environments was verified in various simulations and real experiments. |
Simulation verification of a swarm of UAVs using an extended Boids model.Realistic simulation verification of a swarm of unmanned aerial vehicles in a forest-like environment based on extended Boids model. Obstacles are predefined prior the flight in the controller. The reactive control of the UAVs relies solely on its and others known position. The swarm is flying with constant altitude 2m. |
Swarms of micro aerial vehicles in a former strip mine.Micro aerial vehicles stabilized relatively to their neighbors within a formation or a swarm. Robots employ an onboard visual localization for their relative stabilization. No external localization system, such as Vicon or GPS is used. Experiments are also conducted in challenging outdoor environment of former pit. |
Swarm of self-stabilized unmanned helicopters in indoor environment.A stabilization and control technique developed for steering swarms of unmanned micro aerial vehicles. The approach based on a visual relative localization of swarm particles is designed for utilization of multi-robot teams in real-world dynamic environments. The core of the swarming behaviour is inspired by Reynold's BOID model proposed for 2D simulations of schooling behaviour of fish. The proposed method aspires to be an enabling technique for deployment of swarms of micro areal vehicles outside laboratories that are equipped with precise positioning systems. |
Formation of micro aerial vehicles using relative visual localization.Formation of relatively stabilized micro aerial vehicles. Robots employ an onboard visual localization for their stabilization in a changing formation. No external localization system, such as Vicon is used. |
3D simulation of swarm movement using the escape behavior method.This movie presents an investigation of swarm control dealing with an escape behavior, which is important functionality in application with human-swarm coexistence. The escape behavior algorithm was extended for the swarm purposes. The movement strategies originally developed for holonomic point particles were replaced with dynamic models of UAVs. Examples of the swarm movement under the rules of the escape behavior using the dynamic models of UAVs are shown in the movie. |
Simulation of swarm movement using the escape behavior method.Utilization of Fish School Search (FSS) based method for searching in 3D environment. The FSS based algorithm is modified for control a swarm of quadrocopters and it respects motion constraints and limits of the visual relative localization. In the experiments, MAVs cooperatively search for locations with lowest intensity of a signal transmitted from four transmitters distributed in the environment in different altitude. The intensity of the signal is simulated in the experiment based on known locations and transmission power of virtual transmitters. The experiment was realized in the GRASP Laboratory of University Of Pennsylvania within joint project of Czech Technical University in Prague and the University Of Pennsylvania. |
Selected publications:Journal articles:
WoS conference papers:
Student works
|
|
Gallery:
