The problem of herding a flock birds is posed in graph theoretic framework. A novel algorithm, called the n-wavefront Is developed for enabling single unmanned aerial vehicle to herd desired point. technique applied diverting approaching an airport away from protected zone around airport. algorithm demonstrated simulation and compared with existing strategies using graph-based metrics.

Distributed (federated) learning has become a popular paradigm in recent years. In this scenario private data is stored among several machines (possibly cellular or mobile devices). These collaboratively solve distributed optimization problem, using data, to learn predictive models. The aggressive use of problems involving sensitive resulted privacy concerns. paper we present synchronous stochastic gradient descent based algorithm that introduces via obfuscation client-server model. We prove...

This paper considers a distributed multi-agent optimization problem, with the global objective consisting of sum local functions agents. The agents solve problem using computation and communication between adjacent in network. We present two randomized iterative algorithms for optimization. To improve privacy, our add "structured" randomization to information exchanged prove deterministic correctness (in every execution) proposed despite being perturbed by noise non-zero mean. that special...

This paper is concerned with robotic herding of a swarm birds by another adversarial agent, referred to as the pursuer. The objective to prevent from entering specified volume space, such air space around an airport. n-Wavefront algorithm was introduced authors in a prior enable using unmanned aerial vehicle. In this paper, the performance and stability characteristics are analyzed tools linear nonlinear stability theory, aim proving its identifying permissible optimum values...

This paper considers a distributed multi-agent optimization problem, with the global objective consisting of sum local functions agents. The agents solve problem using computation and communication between adjacent in network. We present two randomized iterative algorithms for optimization. To improve privacy, our add "structured" randomization to information exchanged prove deterministic correctness (in every execution) proposed despite being perturbed by noise non-zero mean. that special...

Unmanned Aerial Systems (UAS) have great potential to aid in search and situation assessment. Here, we present a UAV swarm system performing target adversarial environment. It utilizes Ant-Colony (ACO) Artificial Potential Function (APF) based decentralised strategy. ACO meta-heuristic forms the higher level guidance algorithm APF provide global local path planning. Uncertainty maps are used represent probable locations. The is scalable shown be robust agent loss. Its distributed nature...

We present a distributed solution to optimizing convex function composed of several non-convex functions. Each is privately stored with an agent while the agents communicate neighbors form network. show that coupled consensus and projected gradient descent algorithm proposed in [1] can optimize sum functions under additional assumption on Lipschitzness. further discuss applications this analysis improving privacy optimization.

This paper considers a distributed reinforcement learning problem in the presence of Byzantine agents. The system consists central coordinating authority called "master agent" and multiple computational entities "worker agents". master agent is assumed to be reliable, while, small fraction workers can (malicious) adversaries. are interested cooperatively maximize convex combination honest (non-malicious) worker agents' long-term returns through communication between A actor-critic algorithm...

We propose a distributed algorithm to compute an equilibrium in aggregate games where players communicate over fixed undirected network. Our exploits correlated perturbation obfuscate information shared the prove that our does not reveal private of honest-but-curious adversary who monitors several nodes In contrast with differential privacy based algorithms, method sacrifice accuracy computation provide guarantees.

Continual data collection and widespread deployment of machine learning algorithms, particularly the distributed variants, have raised new privacy challenges. In a scenario, dataset is stored among several machines they solve optimization problem to collectively learn underlying model. We present secure multi-party computation inspired preserving algorithm for optimizing convex function consisting possibly non-convex functions. Each individual objective privately with an agent while agents...

We present a distributed optimization protocol that preserves statistical privacy of agents' local cost functions against passive adversary corrupts some agents in the network. The is composition "zero-sum" obfuscation obfuscates functions, and standard non-private method. show our protects up to t arbitrary as long communication network has (t+1 )-vertex connectivity. sum therefore ensures accuracy computed solution.

Availability of both massive datasets and computing resources have made machine learning predictive analytics extremely pervasive. In this work we present a synchronous algorithm architecture for distributed optimization motivated by privacy requirements posed applications in learning. We an the recently proposed multi-parameter-server architecture. consider group parameter servers that learn model based on randomized gradients received from clients. Clients are computational entities with...

This paper studies a system of linear equations, denoted as $Ax = b$, which is horizontally partitioned (rows in $A$ and $b$) stored over network $m$ devices connected fixed directed graph. We design fast distributed algorithm for solving such that additionally, protects the privacy local data against an honest-but-curious adversary corrupts at most $τ$ nodes network. First, we present TITAN, privaTe fInite Time Average coNsensus algorithm, general average consensus problem graphs, while...

Unmanned Aerial Systems (UAS) possess tremendous capabilities to perform search, tracking and situation assessment in hazardous environments. However, it sometimes becomes critical include a human component offset the limited reasoning of agents. This paper presents Human-in-Loop (HiL) control architecture for distributed cooperative target search. Hierarchical autonomy provides tight over system, enhancing operational reliability, search efficiency effectiveness. Human feedback also...

We propose a distributed algorithm to compute an equilibrium in aggregate games where players communicate over fixed undirected network. Our exploits correlated perturbation obfuscate information shared the prove that our does not reveal private of honest-but-curious adversary who monitors several nodes In contrast with differential privacy based algorithms, method sacrifice accuracy computation provide guarantees.

We present a distributed optimization protocol that preserves statistical privacy of agents' local cost functions against passive adversary corrupts some agents in the network. The is composition ``{\em zero-sum}" obfuscation obfuscates functions, and standard non-private method. show our protects up to $t$ arbitrary as long communication network has $(t+1)$-vertex connectivity. sum therefore ensures accuracy computed solution.