Project

General

Profile

Atividade #625

Patrocínio NI

Added by Carla Cosenza about 3 years ago. Updated almost 3 years ago.

Status:
Fechada
Priority:
Alta
Assignee:
Target version:
Start date:
03/15/2018
Due date:

Description

O Onias me disse que o senhor tinha uma competição da NI cujo premio era uma viagem para um congresso

History

#1 Updated by Luiz Renault Leite Rodrigues about 3 years ago

http://www.ni.com/research-grants/

Dê uma olhada neste link.

A inovação é a inteligência implementada em LabVIEW que levou ao campeonato Latino-Americano. Agora pretende-se ganhar a RoboCUP.

#2 Updated by Carla Cosenza about 3 years ago

Estes foram os textos escritos para o patrocínio.

Patrocínio NI

Research Project Title:

RoboIME: Autonomous Robot Soccer Team

Research Project Description (700 words max):

RoboIME is a team that participates in the Small Size League (SSL) competition, category B, of the RoboCup federation and sponsored by the IEEE Robotics. It is expected that by 2050, a team of fully autonomous humanoide robots will be in conditions to play and win against the FIFA World Cup champion. In order to reach this objective, the RoboCup federation holds anual competitions in several areas of robotics. These competitions stimulate the development and research of robotics. There areas include the robot’s physical stability, movement and decision making processes. The SSL competition focuses on developing a team of robots that can cooperate and play soccer autonomously. In specific, the B category has the objective of developing six robots that will play in a field of size 6 by 9 meters. The robots in this category are not humanoids but rather cylinders. They can have a radius of up to 9cm and a maximum height of 15cm.
This league is focused on the development of artificial intelligence in a highly dynamic environment and therefore works with softwares that are pre-made to control the vision. In order to know the location of the robots, four cameras are located on top of the field, which capture the position and orientation of the robots and the ball. Each robot has a specific pattern on its top, which is used to know its orientation and identify it. These numbers are then sent to both teams, processed by their respective AI and acted upon by their robots. 
Our AI software is made exclusively using LabView. During last year’s RoboCup, we noticed that we are the only team that uses LabView for all the AI. LabView has helped us advance our software immensely. With its help, we went from last place in the Latin American Robotics Competition (LARC) in 2016 to champions of the same competition in the following year. And this year, we intend to develop our AI even more in order to become champions of the category B of the RoboCup.  

Research Experimental Technique Description (1000 words max):

As stated before, our AI software receives the information from the cameras from a standardized software of the competition. Our processing system includes three major modules: filtering and defining game states, decision making, and communication. In the first module, the filtering and defining game states, all the information is processed in order to make it as accurate as possible. Since four cameras are used to locate the objects in the field, it is possible to receive four different positions of the same object. To determine the position to be used, all the information is passed through a Kalman filter. During this phase, the state of the game is also determined using the pre-established rules and the past commands of the referee. With the processed information, the software then starts to use it to decide the future actions of the robots. To help the decision process, each robot is given a personality according to its position and the game state. Our current AI has four personalities: attacker, striker, defender and goalie. The attacker must go after the ball and kick it to the enemy goal or the striker, making a pass. The striker must always be ready to receive a rebound ball or a pass from the attacker. The defenders create a barrier between the ball and our goal, preventing the ball from passing while the goalie kicks it away from our goal area in case it does pass through the barrier. The decision of which personality each robot will assume is done through finite state machines. This year we are implementing a statistical approach in order to later implement machine learning. After the decisions are made, the new data is sent to the robots using our communication module, which uses a protobuf developed in LabView by our members. 
There is also a lot of experiment and development done off the field. In order to have a good control of the robot during the game, many mathematical models are used to select the best control loop for the electronic system. The maintenance of the robots is done using a program developed in LabView that makes each module function independently in order to check if it is performing correctly. 
In the decision making process, a new algorithm is tested in comparison to the AI without it. The new AI and the old one play against each other on the simulator in order to compare the performances of each of them. This makes it easier to notice the difference caused by the new algorithm.

#3 Updated by Luiz Renault Leite Rodrigues about 3 years ago

Tentei colocar em negrito minhas sugestões.

Patrocínio NI

Research Project Title:

RoboIME: Autonomous Robot Soccer Team (mudar para algo relacionado a inteligência artificial e labview).

Research Project Description (700 words max):

RoboIME is a research group that participates in the Small Size League (SSL) competition of the RoboCup federation, sponsored by the IEEE Robotics. One of the main goals of the research is that by 2050, a team of fully autonomous humanoide robots will be able to play and win against the FIFA World Cup champion. In order to reach this objective, the RoboCup federation holds anual competitions in several areas of robotics. These competitions stimulate the development and research of robotics including the robot’s physical stability, movement and decision making processes. The SSL competition focuses on developing a team of robots that can cooperate and play soccer autonomously. The B category, ** in which we participate,** has the objective of developing six robots that will play in a field of size 6 by 9 meters. The robots in this category are not humanoids but rather cylinders*, with * a radius of up to 9cm and a maximum height of 15cm.

This league is focused on the development of artificial intelligence in a highly dynamic environment ** **. In order to know the location of the robots, four cameras are located on top of the field, which capture the position and orientation of the robots and the ball. Each robot has a specific pattern on its top, which is used to know its orientation and identify it. These numbers are then sent to both teams, processed by their respective AI and acted upon by their robots.

Our AI software is made exclusively using LabView. During last year’s RoboCup, we noticed that we are the only team that uses LabView for all the AI. LabView has helped us advance our software immensely. With its help, we went from last place in the Latin American Robotics Competition (LARC) in 2016 to champions of the same competition in the following year. This year, we intend to improve our AI even more in order to become champions of the category B of the RoboCup.

Research Experimental Technique Description (1000 words max):

** The ** AI software receives the information from the cameras from a standardized software of the competition. Our processing system includes three major modules: raw data filtering and game state estimation, decision making, machine learning, and communication. In the first module, all the information is processed in order to make it as accurate as possible and provide good estimates of the robot positions, velocity and in game events. Since four cameras are used to locate the objects in the field, it is possible to receive four different noisy positions of the same object. To determine the position to be used, all the information is passed through a Kalman filter. During this phase, the state of the game is also determined using the pre-established rules and the ** ** commands of the referee.

With the processed information, the software then starts to use it to decide the future actions of the robots. To help the decision process, each robot is given a personality according ** to ** the game state. Our current AI has four personalities: attacker, striker, defender and goalie. The attacker must go after the ball and kick it to the enemy goal or the striker, making a pass. The striker must always be ready to receive a rebound ball or a pass from the attacker. The defenders create a barrier between the ball and our goal, preventing the ball from passing while the goalie kicks it away from our goal area in case it does pass through the barrier. The decision of which personality each robot will assume is done through finite state machines. This year we are implementing a stochastic approach in order to improve machine learning. After the decisions are made, the new commands are sent to the robots using our communication module, which uses a custom protocol also developed in LabView.

There is also a lot of experiment and development done off the field. In order to have a good control of the robot during the game, many mathematical models are used to select the best control strategies for the electronic system. The testing and maintenance of the robots are done using a program developed in LabView that makes each module function independently in order to check if it is performing correctly, following industry trends.

In the decision making process, a new algorithm is tested making the new AI and the old one play against each other on a simulator in order to compare the performance of each of them. This makes it easier to notice the difference caused by the new algorithm and to decide for the next steps.

#4 Updated by Carla Cosenza about 3 years ago

Os textos ficaram assim:

Patrocínio NI

Research Project Title:

RoboIME: AI for Robot Soccer

Research Project Description (700 words max):

RoboIME is a research group that participates in the Small Size League (SSL) competition of the RoboCup federation and sponsored by the IEEE Robotics. One of the main goals of the research is that by 2050, a team of fully autonomous humanoide robots will be able to play and win against the FIFA World Cup champion. In order to reach this objective, the RoboCup federation holds anual competitions in several areas of robotics. These competitions stimulate the development and research of robotics including the robot’s physical stability, movement and decision making processes. The SSL competition focuses on developing a team of robots that can cooperate and play soccer autonomously. The B category, in which we participate, has the objective of developing six robots that will play in a field of size 6 by 9 meters. The robots in this category are not humanoids but rather cylinders with a radius of up to 9cm and a maximum height of 15cm.

This league is focused on the development of artificial intelligence in a highly dynamic environment. In order to know the location of the robots, four cameras are located on top of the field, which capture the position and orientation of the robots and the ball. Each robot has a specific pattern on its top, which is used to know its orientation and identify it. These numbers are then sent to both teams, processed by their respective AI and acted upon by their robots.

Our AI software is made exclusively using LabView. During last year’s RoboCup, we noticed that we are the only team that uses LabView for all the AI. LabView has helped us advance our software immensely. With its help, we went from last place in the Latin American Robotics Competition (LARC) in 2016 to champions of the same competition in the following year. And this year, we intend to develop our AI even more in order to become champions of the category B of the RoboCup.

Research Experimental Technique Description (1000 words max):

The AI software receives the information from the cameras from a standardized software of the competition. Our processing system includes three major modules: raw data filtering and game state estimation, decision making, machine learning and communication. In the first module, all the information is processed in order to make it as accurate as possible and provide good estimates of the robot positions, velocity and in game events. Since four cameras are used to locate the objects in the field, it is possible to receive four different noisy positions of the same object. To determine the position to be used, all the information is passed through a Kalman filter. During this phase, the state of the game is also determined using the pre-established rules and the commands of the referee.

With the processed information, the software then starts to use it to decide the future actions of the robots. To help the decision process, each robot is given a personality according to the game state. Our current AI has four personalities: attacker, striker, defender and goalie. The attacker must go after the ball and kick it to the enemy goal or the striker, making a pass. The striker must always be ready to receive a rebound ball or a pass from the attacker. The defenders create a barrier between the ball and our goal, preventing the ball from passing while the goalie kicks it away from our goal area in case it does pass through the barrier. The decision of which personality each robot will assume is done through finite state machines. This year we are implementing a stochastic approach in order to improve machine learning. After the decisions are made, the new data is sent to the robots using our communication module, which uses a custom protocol also developed in LabView.

There is also a lot of experiment and development done off the field. In order to have a good control of the robot during the game, many mathematical models are used to select the best control strategies for the electronic system. The testing and maintenance of the robots is done using a program developed in LabView that makes each module function independently in order to check if it is performing correctly.

In the decision making process, a new algorithm is tested making the new AI and the old one play against each other on the simulator in order to compare the performances of each of them. This makes it easier to notice the difference caused by the new algorithm.

#5 Updated by Carla Cosenza about 3 years ago

Outras opções de título: AI for Robot Soccer with LabView/ Developing AI for Robot Soccer with LabView/ Robot Soccer AI in LabView

#6 Updated by Luiz Renault Leite Rodrigues about 3 years ago

Foi enviada a proposta?

#7 Updated by Carla Cosenza about 3 years ago

Foi enviado sim Capitão!

#8 Updated by Carla Cosenza almost 3 years ago

  • Status changed from Em andamento to Resolvida

#9 Updated by Carla Cosenza almost 3 years ago

  • Status changed from Resolvida to Fechada

Also available in: Atom PDF

Go to top
Add picture from clipboard (Maximum size: 500 MB)