Multi Agent Robotic System Design for Team NUST in RoboCup SPL

Abstract

RoboCup is amongst the most prestigious of international robotic competitions, held annually. Its Standard Platform League comprises of soccer competitions between teams of humanoid robots. NAO robot from Aldebaran Robotics is the hardware platform for RoboCup SPL; while incorporates sonar sensors, joint encoders, cameras, 25 degree of freedom, foot pressure sensors, Wifi, Embedded Linux and Atom processor. The key objective of this project is to make a multi-agent system which can contribute towards the complete software module of robots which can play soccer with each other. The robots operate fully autonomously with only local vision and highly constrained inter-robot communication, such that there is no external control by humans or computer. The robots were provides to us by RISE Lab, SMME, NUST. We incorporated the vision, localization, kinematics and artificial intelligence module inside them through different libraries like OpenCV. The project is based on design and development of a multi agent robotic system. Including four major areas of Perception, Localization, Kinematics and Architecture incorporating coordination methodologies. Perception includes detection of field lines, detecting types of corners, goal post and ball. Localization utilizes the perception output to intelligently localize the robot in the field along with other robots, ball and goal. Kinematics employs Inverse Kinematics and Path calculation to carry out a stable kick with varying direction and speed. The Architecture is used to implement an intelligent state machine to carry out the game play. The vision module is able to identify the field landmarks in a suitable amount of noise with considerable amount of accuracy. The localization module intelligently localize the robot’s position inside the field utilizing the landmarks information coming through the vision module. The kinematics module is able to kick the ball in a particular desired direction with variable speed. The artificial intelligence module is able to intelligently assign roles to the particular robot depending on the situation of the soccer field. Also the path planning algorithm is able to find the best possible path solution of making a successful goal kick. TeamNUST is the only team from South Asia to have qualified to participate in RoboCup SPL 2015 to be held in China, Hefei. This project is able to provide a basic module of the future contributors and participants of RoboCup and also able to contribute towards the 2050 goal of robots which can defeat the soccer world champions of that time.