PUBLICATIONS & PAPERS

Inspection method based on multi-agent auction for graphic-like maps

 

Abstract: Nature teaches us how the collaboration between the members of a herd is an important aspect to ensure their survival as a group. Collaboration requires communication, but it may happen that it could not be established for some reason. In this context, members of the herd have to make decisions on its own, trying to do the best for the group. Inspired by this principle, we have designed a method to  overcome the communication difficulties of the environment. The team members compete using their own made decisions, this competition affects on the best possible way to the team benefit.

Download it

 

Selective Method Based on Auctions for Map Inspection by Robotic Teams

 

Abstract: In the inspection of a known environment by a team of robots, communication problems may exists between members of the team, even, due to the hostile environment these members can be damaged. In this paper, a redundant, robust and fault tolerant method to cover a known environment using a multi-agent system and where the communications are not guaranteed is presented. Through a simple auction system for cooperation and coordination, the aim of this method is to provide an effective way to solve communication or hardware failures problems in the inspection task of a known environment. We have conducted several experiments in order to verify and validate the proposed approach. The results are commented and compared to other methods.

Download it from Springer. Also, you can see a video showing how it works.

 

Study of a Multi-Robot Collaborative Task through Reinforcement Learning

 

Abstract: A open issue in multi-robots systems is coordinating the collaboration between several agents to obtain a common goal. The most popular solutions use complex systems, several types of sensors and complicated controls systems. This paper describes a general approach for coordinating the movement of objects by using reinforcement learning. Thus, the method proposes a framework in which two robots are able to work together in order to achieve a common goal. We use simple robots without any kind of internal sensors and they only obtain information from a central camera. The main objective of this paper is to define and to verify a method based on reinforcement learning for multi-robot systems, which learn to coordinate their actions for achieving common goal.

Download it from Springer

 

Evolutionary Controllers for Snake Robots Basic Movements

  • J. Pereda, J. de Lope and D. Maravall “Evolutionary Controllers for Snake Robots Basic Movements”, 2nd International Workshop on Hybrid Artificial Intelligence Systems; HAIS (2007).

 

Abstract: A method to generate movements in a snake robot using proportional-integral-derivate controllers (PID) and adjust the constants values to be natural is proposed. Specifically, the method is applied to adjust the movement of a snake robot to natural postures defining a simplify PID controller and adjusting the constants values of the controller. Our approach is based on proportional-integral-derivate controllers, using genetics algorithms to solve the problem. In this paper we explain how adjust the restrictions that must be accomplished for generate a natural movement and make an exhaustive study about snake robots, proportional-integral-derivate controllers and genetics algorithms.

Download it from Springer

 

Comparative Analysis of Artificial Neural Network Training Methods for Inverse Kinematics Learning

  • J. Pereda, J. de Lope and D. Maravall “Comparative Analysis of Artificial Neural Network Training Methods for Inverse Kinematics Learning”, 2nd International Workshop on Hybrid Artificial Intelligence Systems; HAIS (2007).
 

Abstract: A method for obtaining an approximate solution to the inverse kinematics of a articulated chain is proposed in this paper. Specifically, the method is applied to determine the joint positions of a humanoid robot in locomotion tasks, defining the successive stable robot configurations needed to achieve the final foot position in each step. Our approach is based on the postural scheme method, using artificial neural networks to solve the problem. In this paper we define the restrictions that must be accomplished by the networks and make an exhaustive study about the learning algorithms, transfer functions, training sets composition, data normalization and artificial neural network topologies.

Download it from Springer