Nate with an Aibo

Standard Platform League (Nao League)

2012 Robocup Champions
UT Austin Villa,
The 2012 Robocup SPL Champions.

The UT Austin Villa team, led by Prof. Peter Stone at the University of Texas at Austin, competes in the Standard Platform League (SPL). UT Austin Villa won the international SPL championship at RoboCup 2012! The team also won the 2009 US Open, the 2010 US Open, and the 2012 US Open in the SPL, and finished third in the 2013 RoboCup competition, third in the 2010 RoboCup competition, and fourth in the 2009 RoboCup competition. UT Austin Villa was previously called TT-UT Austin Villa, as it was a joint team led by Prof. Mohan Sridharan at Texas Tech University and Prof. Peter Stone at the University of Texas at Austin until the 2012 season.

In the Standard Platform League, each team competes with identical Aldebaran Nao robots. Since the hardware is identical for all teams, the focus of this league is on robotic control, humanoid motion, sensing (including computer vision), and localization algorithms, as well as on team coordination and strategic reasoning.

Our team built up its code base from scratch, starting in December 2007. We built on our experience in the previous legged league using the Sony Aibo robot to engineer our new software for easy use, extendability and debugability.

Aldebaran Nao
The Aldebaran Nao.

There are four main problems to solve in the Nao league: vision, localization, locomotion, and coordination.

Vision is the process of identifying objects on the field. This process generally includes figuring out what colors things are, recognizing the objects, and then determining the distances to the objects. Even the slightest change in lighting conditions can make vision nearly impossible.

Localization is determining where on the field the robot is. Localization relies very heavily on vision (so it can calculate its position using landmarks), as well as odometry, which tells the robot how it is moving. Localization is one of the most complex parts of our code

Locomotion is getting the robot to move. Creating stable motions for a humanoid robot such as the Nao can be quite difficult. We would like to develop a walk that is stable and fast, while also being able to change directions quickly as well. Locomotion also includes developing dynamic kicks for the Nao

Coordination is the one problem that sets RoboCup Soccer apart from many other robotics domains. Put simply, our robots need to work as a team. It is not enough for each robot to want to score a goal - if we did that, there would be just a big pile of robots trying to get the ball. Our robots need to work with each other, and against the other team, at the same time.

Videos

2013 RoboCup Highlights

Highlights of UT Austin Villa at the 2013 RoboCup Standard Platform League competition in Eindhoven, where the team took 3rd place.

2012 RoboCup Highlights

Highlights of UT Austin Villa at the 2012 RoboCup Standard Platform League competition in Mexico City, where the team took 1st place.

2010 RoboCup Highlights

Highlights of TT-UT Austin Villa at the 2010 RoboCup Standard Platform League competition in Singapore, where the team took 3rd place.

2010 RoboCup Passing Challenge

Video of TT-UT Austin Villa's entry in the passing challenge at the 2010 RoboCup Standard Platform League competition in Singapore. Austin Villa completed the challenge in the fastest time of anyone in the competition.

Learning to Score Penalty Kicks via Reinforcement Learning

The accompanying video for our ICRA 2010 paper, where we learn to score penalty kicks via a novel model-based reinforcement learning method.

2009 RoboCup Highlights

Highlights of TT-UT Austin Villa at the 2009 RoboCup Standard Platform League. TT-UT Austin Villa finished in 4th place, losing to only two teams during the tournament.

2009 US Open Highlights

Highlights of TT-UT Austin Villa at the 2009 US Open. TT-UT Austin Villa won the 2009 US Open with a finals win over UPenn (1-1 tie, 3-2 in penalty kicks).

Behavior: Simulation

This video shows the behavior of our Nao in simulation. The robot behaves as follows: First, it looks around, both to localize itself and to find the ball. Then it focuses on the ball. Next, it walks up to the ball and leans over to see if it is lined up to kick. Then it steps to the side to line up the ball, and finally kicks it into the goal.

Behavior: Physical Robot

This video shows an example behavior of our Nao. The robot walks towards the ball, while stopping every few steps to look around and localize itself. Once it reaches the ball, it leans over to see that it is there, and then walks into it, pushing it into the goal.

Kicking: Physical Robot

This video shows the Nao kicking the ball using the kick engine we developed.

Localization: Particle Filter

This video shows the localization of our Nao robot using our particle filter. The objects the robot sees are visible as it scans its head around the field. Each particle is updated based on these observations and the estimate of the robot's pose on the field is determined from these particles.

Pose Kalman Filter

We use a Kalman filter to estimate the pose of the robot (i.e. its orientation in space), because the raw accelerometer and gyros are very noisy. This video shows an example of the filtered pose compared to the pose of the robot calculated from the raw sensor values. The filtered pose is in white, the unfiltered pose is in blue.

Standing Up: Physical Robot

This video shows the Nao getting up from its back using the provided method from Aldebaran.

Team Reports

Valid CSS!
Valid XHTML 1.0!