Control Of Robotic Swarms Research Articles

A Low-Complexity Brain-Computer Interface for High-Complexity Robot Swarm Control.

A brain-computer interface (BCI) is a system that allows a human operator to use only mental commands in controlling end effectors that interact with the world around them. Such a system consists of a measurement device to record the human user's brain activity, which is then processed into commands that drive a system end effector. BCIs involve either invasive measurements which allow for high-complexity control but are generally infeasible, or noninvasive measurements which offer lower quality signals but are more practical to use. In general, BCI systems have not been developed that efficiently, robustly, and scalably perform high-complexity control while retaining the practicality of noninvasive measurements. Here we leverage recent results from feedback information theory to fill this gap by modeling BCIs as a communications system and deploying a human-implementable interaction algorithm for noninvasive control of a high-complexity robot swarm. We construct a scalable dictionary of robotic behaviors that can be searched simply and efficiently by a BCI user, as we demonstrate through a large-scale user study testing the feasibility of our interaction algorithm, a user test of the full BCI system on (virtual and real) robot swarms, and simulations that verify our results against theoretical models. Our results provide a proof of concept for how a large class of high-complexity effectors (even beyond robotics) can be effectively controlled by a BCI system with low-complexity and noisy inputs.

Closed-Loop Motion Control of Robotic Swarms – A Tether-Based Strategy

Swarm robots can achieve effective task execution via closed-loop motion control. However, such a goal can only be realized through accurate localization of the swarm. Past approaches have focused on addressing this issue using external sensors, static sensor networks, or through active localization—requirements that may restrict the motion of the swarm or may not be achievable in practice. We present a tether-based strategy that achieves closed-loop swarm-motion control by using a secondary team of mobile sensors. These sensors form a wireless tether that allows the swarm to indirectly sense a home base or a landmark, and to compensate for the accumulated motion errors via a closed-loop control strategy. The proposed strategy is the first to use a tether of mobile sensors that can dynamically reshape and reconnect to various points in the environment to achieve closed-loop motion control. The novelty of the strategy is in its ability to adapt to any swarm motion considered, and to be applied to swarms with limited sensing capabilities and knowledge of their environment. The performance of the proposed strategy was validated through extensive experiments.

The Robustness of Human Advantage in Swarm Leader Identification

Control of robotic swarms through control over a leader(s) has become the dominant approach to supervisory control over these largely autonomous systems. Resilience in the face of attrition is one of the primary advantages attributed to swarms yet the presence of leader(s) makes them vulnerable to decapitation. Algorithms which allow a swarm to hide its leader are a promising solution. In prior work we found that using a graph neural network, GNN, a swarm could be trained to flock following a leader. An Adversary NN trained to identify that leader (naïve condition) performed substantially better than human observers. When the swarm was trained to hide its leader (deception conditions), however, the advantage reversed with humans outperforming the Adversary. This human advantage persisted even when the swarm and Adversary were jointly trained, allowing the Adversary to adapt to the swarm’s evolving strategies for hiding its leader. The present study investigates the robustness of human leader identification by testing identifications made in the presence of medium and high levels of visual clutter. Clutter degraded human performance to some extent but human accuracy in leader identification remained well above that of the Adversary in deception conditions. Human performance even approached that for an unhidden leader under joint training. This study confirms the robustness of the human superiority effect and argues for the inclusion of humans in AI systems which may confront learned deception.

Wide and Deep Graph Neural Network With Distributed Online Learning

Graph neural networks (GNNs) are naturally distributed architectures for learning representations from network data. This renders them suitable candidates for decentralized tasks. In these scenarios, the underlying graph often changes with time due to link failures or topology variations, creating a mismatch between the graphs on which GNNs were trained and the ones on which they are tested. Online learning can be leveraged to retrain GNNs at testing time to overcome this issue. However, most online algorithms are centralized and usually offer guarantees only on convex problems, which GNNs rarely lead to. This paper develops the Wide and Deep GNN (WD-GNN), a novel architecture that can be updated with distributed online learning mechanisms. The WD-GNN consists of two components: the wide part is a linear graph filter and the deep part is a nonlinear GNN. At training time, the joint wide and deep architecture learns nonlinear representations from data. At testing time, the wide, linear part is retrained, while the deep, nonlinear one remains fixed. This often leads to a convex formulation. We further propose a distributed online learning algorithm that can be implemented in a decentralized setting. We also show the stability of the WD-GNN to changes of the underlying graph and analyze the convergence of the proposed online learning procedure. Experiments on movie recommendation, source localization and robot swarm control corroborate theoretical findings and show the potential of the WD-GNN for distributed online learning.

Smoothed-Particle-Hydrodynamics for the Control of Robotic Swarms, and Parametric Associations

A swarm of robots has several macroscopic properties that are extraneous to an individual robot. This is a direct consequence of the fact that a swarm is a coupled high-dimensional system that is characterized by interactions between individual robots. In a smoothed-particle-hydrodynamics-based control of a swarm of robots, the equations of motion may be parameterized by key parameters that affect the macroscopic properties of the swarm such as separation, compressibility, cohesivity, and velocity matching. In this article, we analyze this relation between the parameters in the equations of motion and the properties of the swarm. The resulting analysis enables the design of a versatile, scalable controller that may be modified and adopted to suit the desired objectives of individual applications. We also present a new interaction potential that enables controlling the robot-robot separation, beyond maintaining the robot density, and providing a swarm cohesive force that keeps the robots in the swarm together. The constructed interaction potential is conservative, and the overall swarm stability may still be established using passivity arguments in the presence of a dissipative process. In this article, we use Lyapunov methods on the total energy to establish this stability. The salient features of the proposed controller are verified using simulations and experiments. The experiments are conducted using two real physical airplanes to demonstrate the practical feasibility of the proposed controller.

Human vs. Deep Neural Network Performance at a Leader Identification Task

Control of robotic swarms through control over a leader(s) has become the dominant approach to supervisory control over these largely autonomous systems. Resilience in the face of attrition is one of the primary advantages attributed to swarms yet the presence of leader(s) makes them vulnerable to decapitation. Algorithms which allow a swarm to hide its leader are a promising solution. We present a novel approach in which neural networks, NNs, trained in a graph neural network, GNN, replace conventional controllers making them more amenable to training. Swarms and an adversary intent of finding the leader were trained and tested in 4 phases: 1-swarm to follow leader, 2-adversary to recognize leader, 3-swarm to hide leader from adversary, and 4-swarm and adversary compete to hide and recognize the leader. While the NN adversary was more successful in identifying leaders without deception, humans did better in conditions in which the swarm was trained to hide its leader from the NN adversary. The study illustrates difficulties likely to emerge in arms races between machine learners and the potential role humans may play in moderating them.

Stability of graph convolutional neural networks to stochastic perturbations

Graph convolutional neural networks (GCNNs) are nonlinear processing tools to learn representations from network data. A key property of GCNNs is their stability to graph perturbations. Current analysis considers deterministic perturbations but fails to provide relevant insights when topological changes are random. This paper investigates the stability of GCNNs to stochastic graph perturbations induced by link losses. In particular, it proves the expected output difference between the GCNN over random perturbed graphs and the GCNN over the nominal graph is upper bounded by a factor that is linear in the link loss probability. We perform the stability analysis in the graph spectral domain such that the result holds uniformly for any graph. This result also shows the role of the nonlinearity and the architecture width and depth, and allows identifying handle to improve the GCNN robustness. Numerical simulations on source localization and robot swarm control corroborate our theoretical findings.

Particle Swarm Optimization—An Adaptation for the Control of Robotic Swarms

Particle Swarm Optimization (PSO) is a numerical optimization technique based on the motion of virtual particles within a multidimensional space. The particles explore the space in an attempt to find minima or maxima to the optimization problem. The motion of the particles is linked, and the overall behavior of the particle swarm is controlled by several parameters. PSO has been proposed as a control strategy for physical swarms of robots that are localizing a source; the robots are analogous to the virtual particles. However, previous attempts to achieve this have shown that there are inherent problems. This paper addresses these problems by introducing a modified version of PSO, as well as introducing new guidelines for parameter selection. The proposed algorithm links the parameters to the velocity and acceleration of each robot, and demonstrates obstacle avoidance. Simulation results from both MATLAB and Gazebo show close agreement and demonstrate that the proposed algorithm is capable of effective control of a robotic swarm and obstacle avoidance.

A Kalman estimation model of human trust in supervisory control of robotic swarms

Trust is an important factor in the interaction between humans and automation that can mediate the reliance of human operators. In this work, we evaluate a computational model of human trust on swarm systems based on Sheridan (2019)’s modified Kalman estimation model using existing experiment data (Nam, Li, Li, Lewis, & Sycara, 2018). Results show that our Kalman Filter model outperforms existing state of the art alternatives including dynamic Bayesian networks and inverse reinforcement learning. This work is novel in that: 1) The Kalman estimator is the first computational model formulating the human trust evolution as a combination of both open-loop trust anticipation and closed-loop trust feedback. 2) The proposed model considers the operator’s cognitive time lag between perceiving and processing the system display. 3) The proposed model provides a personalized model for each individual and reaches a better level of fitness than state-of-the-art alternatives.

On how swarm robotics can be used to describe particle system’s deformation

In previous works, we have described time evolution of a two-dimensional particle lattice, subject to deformation, without the use of Newton’s law. According to our experience, in control of robotic swarm, the new position of a particle is determined by the spatial position of its neighbours; therefore, we have used an interaction law based on the spatial position of the particles themselves. The tool that we have realized reproduced some behaviour of deformable bodies both according to the standard Cauchy model and the second gradient theory. In this paper, we try to stress what is still under investigation, like the relationship describing the interaction rule and its physical meaning; moreover, we shall describe as some solutions do not agree with the behaviour of the classical solution coming out from differential equations.

A distributed robot swarm control for dynamic region coverage

We propose a new distributed method for coverage of a moving deformable convex region with a team of robots in a communication network. Robots execute a distributed self-deployment strategy based on Centroidal Voronoi Tessellations (CVT) to cover the region evenly while preventing collisions. The main contribution is the addition of a feedforward action to overcome the well-known slow convergence issue of the basic CVT algorithms. This action is derived by each robot from the information about the region that floods through the network from a few selected leaders. The method allows to quickly adapt to the fastly changing working area in spite of the light communication requirements, and it is well suited for large teams of expendable robots.

Evolution of swarm robotics systems with novelty search

Novelty search is a recent artificial evolution technique that challenges traditional evolutionary approaches. In novelty search, solutions are rewarded based on their novelty, rather than their quality with respect to a predefined objective. The lack of a predefined objective precludes premature convergence caused by a deceptive fitness function. In this paper, we apply novelty search combined with NEAT to the evolution of neural controllers for homogeneous swarms of robots. Our empirical study is conducted in simulation, and we use a common swarm robotics task—aggregation, and a more challenging task—sharing of an energy recharging station. Our results show that novelty search is unaffected by deception, is notably effective in bootstrapping evolution, can find solutions with lower complexity than fitness-based evolution, and can find a broad diversity of solutions for the same task. Even in non-deceptive setups, novelty search achieves solution qualities similar to those obtained in traditional fitness-based evolution. Our study also encompasses variants of novelty search that work in concert with fitness-based evolution to combine the exploratory character of novelty search with the exploitatory character of objective-based evolution. We show that these variants can further improve the performance of novelty search. Overall, our study shows that novelty search is a promising alternative for the evolution of controllers for robotic swarms.

Formation control of robotic swarm using bounded artificial forces.

Formation control of multirobot systems has drawn significant attention in the recent years. This paper presents a potential field control algorithm, navigating a swarm of robots into a predefined 2D shape while avoiding intermember collisions. The algorithm applies in both stationary and moving targets formation. We define the bounded artificial forces in the form of exponential functions, so that the behavior of the swarm drove by the forces can be adjusted via selecting proper control parameters. The theoretical analysis of the swarm behavior proves the stability and convergence properties of the algorithm. We further make certain modifications upon the forces to improve the robustness of the swarm behavior in the presence of realistic implementation considerations. The considerations include obstacle avoidance, local minima, and deformation of the shape. Finally, detailed simulation results validate the efficiency of the proposed algorithm, and the direction of possible futrue work is discussed in the conclusions.

Temporal Logic Planning and Control of Robotic Swarms by Hierarchical Abstractions

<para xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> We develop a hierarchical framework for planning and control of arbitrarily large groups (swarms) of fully actuated robots with polyhedral velocity bounds moving in polygonal environments with polygonal obstacles. At the first level of hierarchy, we aggregate the high-dimensional control system of the swarm into a small-dimensional control system capturing its essential features. These features describe the position of the swarm in the world and its size. At the second level, we reduce the problem of controlling the essential features of the swarm to a model-checking problem. In the obtained hierarchical framework, high-level specifications given in natural language, such as linear temporal logic formulas over linear predicates in the essential features, are automatically mapped to provably correct robot control laws. For the particular case of an abstraction based on centroid and variance, we show that swarm cohesion, interrobot collision avoidance, and environment containment can also be specified and automatically guaranteed in our framework. The obtained communication architecture is centralized. </para>

Hormone-Inspired Self-Organization and Distributed Control of Robotic Swarms

The control of robot swarming in a distributed manner is a difficult problem because global behaviors must emerge as a result of many local actions. This paper uses a bio-inspired control method ca...