PABLO SAMPAIO

The Gravitational Strategy for the Timed Patrolling

A great number of techniques were already applied to the non-adversarial variation of the multiag... more A great number of techniques were already applied to the non-adversarial variation of the multiagent patrolling problem. Experiments suggest that, for general graphs, all those approaches are inferior to a strategy based on the travelling salesman problem (TSP), in which agents are distributed equidistantly along the TSP-cycle. This approach, however, is neither optimal nor scalable. In this article, we present

CinBalada: A multiagent rhythm factory

Journal of the Brazilian Computer Society, Sep 1, 2008

CinBalada is a system for automatic creation of polyphonic rhythmic performances by mixing elemen... more CinBalada is a system for automatic creation of polyphonic rhythmic performances by mixing elements from different musical styles. This system is based on agents that act as musicians playing percussion instruments in a drum circle. Each agent has to choose from a database the rhythm pattern of its instrument that satisfies the "rhythmic role" assigned to him in order to produce a collectivelyconsistent rhythmic performance. A rhythmic role is a concept that we proposed here with the objective of representing culture-specific rules for creation of polyphonic performances.

Download

Decentralized Strategies Based on Node Marks for Multi-Robot Patrolling on Weighted Graphs

In the multi-robot patrolling problem in graphs, each robot continually visits the nodes in order... more In the multi-robot patrolling problem in graphs, each robot continually visits the nodes in order to minimize a collective performance metric. Many decentralized strategies were proposed. However, the more general case of graphs with arbitrary weights (representing distances or costs) was approached mostly by centralized algorithms or algorithms with some kind of global information. In this work, we propose three decentralized techniques that use only local information – data written on the nodes – and can be implemented in simple robots. They extend three previous decentralized strategies byincorporating the edge weights in the decision process. We present a simulation study that show that the extended techniques increased the performances in many settings, but not in all.

New Patrolling Strategies with Short-Range Perception

In a Multi-Agent Patrolling Problem (MAPP), a team of agents must optimize a collective metric by... more In a Multi-Agent Patrolling Problem (MAPP), a team of agents must optimize a collective metric by moving through a given environment model. MAPPs have a wide range of applications, like continuous surveillance, floor cleaning, and detecting gas leakage. Many heuristic strategies for MAPPs were proposed in the literature using different techniques. In the k-range strategies, autonomous agents communicate indirectly by reading and writing information in the surrounding nodes that are located up to k edges away from the agent. From these, the 0-range strategies (where agents can only sense its current node) are especially suitable for the design of robust autonomous robots with simple hardware. In this work, we propose a new method to convert 1-range strategies (the most common case) into 0-range strategies. We evaluated the new strategies in simulations comparing them to the original 1-range versions. We also built a real low-cost robot that runs the 0-range strategies.

Reducing the Range of Perception in Multi-agent Patrolling Strategies

Journal of Intelligent and Robotic Systems, Oct 19, 2017

Multi-Agent Patrolling Problems consist in moving agents throughout a graph in order to optimize ... more Multi-Agent Patrolling Problems consist in moving agents throughout a graph in order to optimize a collective performance metric. Some strategies from the literature tackle this problem by dispatching decentralized autonomous agents that coordinate themselves merely by sensing and writing information in the nodes. In this work, they are called k-range local strategies, were k indicates the range, in number of edges, of the agents' sensing capabilities. The 1-range strategies (where agents can sense up to its neighbor nodes) are certainly the most common case in the literature. And only few 0-range strategies (where agents can only sense its current node) were found, although this type of strategy has the advantage of requiring simpler hardware, when applied in the design of real robots. In this work, we propose two higher-level procedures to reduce the perception range of 1-range strategies to 0: the Zr Method and the EZr Method. Applying both methods in 1-range strategies found in the literature, we created twenty new 0-range strategies, which were evaluated in a simulation experiment described and analyzed here. We also developed a prototype of a low-cost patrolling robot that is able to run the 0-range strategies proposed in this work.

Download

Multi-Robot Navigation and Exploration in Graphs with Entrance-Dependent Identification of the Edges

2019 Latin American Robotics Symposium (LARS), 2019 Brazilian Symposium on Robotics (SBR) and 2019 Workshop on Robotics in Education (WRE)

We propose two multi-robot algorithms to enable navigation in graph environments where numerical ... more We propose two multi-robot algorithms to enable navigation in graph environments where numerical ports identify edges incidents to each node. This identification, however, varies dynamically according to the entrance edge used by the robot to reach the node. A new exploration algorithm called ECEP writes information on the nodes that can later be used by the new navigation algorithm NCEP to properly navigate to any given neighbor. Both algorithms require minimal memory space from the agent and from each node. Simulations showed that it requires an average exploration time O(G/k), for k robots. Both were also successfully implemented on robots with simple hardware.

Patrulha temporal: taxonomia, métricas e novas soluções

Universidade Federal de Pernambuco, Mar 6, 2013

Dedico, primeiramente, à minha esposa amada, Priscila, à Laura, minha filha preciosa, e aos próxi... more Dedico, primeiramente, à minha esposa amada, Priscila, à Laura, minha filha preciosa, e aos próximos filhos que pretendemos ter. À minha mãe Valdeluce, in memoriam, mulher sábia e virtuosa (agradeço a Deus pelos anos que a tive comigo). Foi ela quem me deu o incentivo final para entrar no doutorado e assumir, de vez, minha vocação para a carreira acadêmica. Na verdade, ela fez muito mais do que isso, mas não sou capaz de expressar tudo em palavras. Ao meu pai Ivaldo, que junto com a minha mãe, lutou pela minha formação, me dando as melhores condições possíveis para estudar e crescer profissionalmente. Aos meus irmãos, que sempre torceram por mim. E a todos os demais familiares que também me incentivaram. Agradeço ao Deus Todo-Poderoso criador dos céus e da Terra por ter me permitido e me dado forças para concluir este trabalho. Aos meus orientadores, Geber Ramalho e Patrícia Tedesco, pela orientação, paciência e por terem me proporcionado essa enriquecedora experiência de trabalhar com eles. "The mere formulation of a problem is far more essential than its solution.

Download

Reducing the Range of Perception in Multi-agent Patrolling Strategies

Journal of Intelligent & Robotic Systems, 2017

Multi-Agent Patrolling Problems consist in moving agents throughout a graph in order to optimize ... more Multi-Agent Patrolling Problems consist in moving agents throughout a graph in order to optimize a collective performance metric. Some strategies from the literature tackle this problem by dispatching decentralized autonomous agents that coordinate themselves merely by sensing and writing information in the nodes. In this work, they are called k-range local strategies, were k indicates the range, in number of edges, of the agents' sensing capabilities. The 1-range strategies (where agents can sense up to its neighbor nodes) are certainly the most common case in the literature. And only few 0-range strategies (where agents can only sense its current node) were found, although this type of strategy has the advantage of requiring simpler hardware, when applied in the design of real robots. In this work, we propose two higher-level procedures to reduce the perception range of 1-range strategies to 0: the Zr Method and the EZr Method. Applying both methods in 1-range strategies found in the literature, we created twenty new 0-range strategies, which were evaluated in a simulation experiment described and analyzed here. We also developed a prototype of a low-cost patrolling robot that is able to run the 0-range strategies proposed in this work.

Download

New Patrolling Strategies with Short-Range Perception

2016 XIII Latin American Robotics Symposium and IV Brazilian Robotics Symposium (LARS/SBR), 2016

In a Multi-Agent Patrolling Problem (MAPP), a team of agents must optimize a collective metric by... more In a Multi-Agent Patrolling Problem (MAPP), a team of agents must optimize a collective metric by moving through a given environment model. MAPPs have a wide range of applications, like continuous surveillance, floor cleaning, and detecting gas leakage. Many heuristic strategies for MAPPs were proposed in the literature using different techniques. In the k-range strategies, autonomous agents communicate indirectly by reading and writing information in the surrounding nodes that are located up to k edges away from the agent. From these, the 0-range strategies (where agents can only sense its current node) are especially suitable for the design of robust autonomous robots with simple hardware. In this work, we propose a new method to convert 1-range strategies (the most common case) into 0-range strategies. We evaluated the new strategies in simulations comparing them to the original 1-range versions. We also built a real low-cost robot that runs the 0-range strategies.

The Gravitational Strategy for the Timed Patrolling

2010 22nd IEEE International Conference on Tools with Artificial Intelligence, 2010

A great number of techniques were already applied to the non-adversarial variation of the multiag... more A great number of techniques were already applied to the non-adversarial variation of the multiagent patrolling problem. Experiments suggest that, for general graphs, all those approaches are inferior to a strategy based on the travelling salesman problem (TSP), in which agents are distributed equidistantly along the TSP-cycle. This approach, however, is neither optimal nor scalable. In this article, we present

CinBalada: a multiagent rhythm factory

Journal of the Brazilian Computer Society, 2008

CinBalada is a system for automatic creation of polyphonic rhythmic performances by mixing elemen... more CinBalada is a system for automatic creation of polyphonic rhythmic performances by mixing elements from different musical styles. This system is based on agents that act as musicians playing percussion instruments in a drum circle. Each agent has to choose from a database the rhythm pattern of its instrument that satisfies the "rhythmic role" assigned to him in order to produce a collectivelyconsistent rhythmic performance. A rhythmic role is a concept that we proposed here with the objective of representing culture-specific rules for creation of polyphonic performances.

Download

CInBalada: um Laboratório Multiagente de Geração de Ritmos de Percussão

Proceedings of the X …, 2005

Resumo. Neste trabalho apresentamos o sistema CInBalada, onde agentes percussionistas inteligente... more

Download

Decentralized Strategies Based on Node Marks for Multi-Robot Patrolling on Weighted Graphs

2019 Latin American Robotics Symposium (LARS), 2019 Brazilian Symposium on Robotics (SBR) and 2019 Workshop on Robotics in Education (WRE)

In the multi-robot patrolling problem in graphs, each robot continually visits the nodes in order... more In the multi-robot patrolling problem in graphs, each robot continually visits the nodes in order to minimize a collective performance metric. Many decentralized strategies were proposed. However, the more general case of graphs with arbitrary weights (representing distances or costs) was approached mostly by centralized algorithms or algorithms with some kind of global information. In this work, we propose three decentralized techniques that use only local information – data written on the nodes – and can be implemented in simple robots. They extend three previous decentralized strategies byincorporating the edge weights in the decision process. We present a simulation study that show that the extended techniques increased the performances in many settings, but not in all.

Uploads

Papers by PABLO SAMPAIO

Log In