RoboCup Standard Platform League
RoboCup is an international initiative that fosters research in robotics and artificial intelligence. The RoboCup Standard Platform League (SPL) is a soccer league in which all teams compete with identical robots. The robots operate fully autonomously, i.e. there is no external control, neither by humans nor by computers.
I have been involved with the RoboCup competition from 2002 to 2008, using both the Sony legged AIBO and the Aldebaran humanoid Nao robots. During this time our team has consistently placed among the top in the world.
The majority of my RoboCup related research has focused on the design and development of the single-robot behavior architecture that is responsible for high-level decision making. I have also assisted in the development of vision, localization, world modeling and motion components. Some of my independent research projects related to RoboCup include:
- Autonomous Walk Learning: Developing fast gaits for legged robots is a difficult task that requires optimizing parameters in a highly irregular, multidimensional space. In the past, walk optimization for quadruped robots, namely the Sony AIBO robot, was done by hand-tuning the parameterized gaits. In addition to requiring a lot of time and human expertise, this process produced sub-optimal results. We present an evolutionary algorithm for autonomous gait learning that has achieved a 20% increase in the speed of our robots over the best hand-turned walks. [PDF]
- Kick Modeling: For legged robots, such as the Sony AIBO, designing effective motions and determining their effects is a challenging problem that is traditionally approached through a generate and test methodology (i.e. trial and error). We present a method for learning the effects of kicking motions by automatically acquiring models of the kicks in terms of key values that describe their effects on the ball's trajectory. These models are then used to select the most promising kick in a given state of the world. We demonstrate that a robot that takes into account the learned predicted effects of its actions performs significantly better than its counterpart. [PDF]
- Observation Models: We present a general-purpose framework for updating a robot's observation model within the context of planning and execution. Traditional plan execution relies on monitoring plan step transitions through accurate state observations obtained from sensory data. In order to gather meaningful state data from sensors, tedious and time-consuming calibration methods are often required. To address this problem we introduce Reverse Monitoring, a process of learning an observation model through the use of plans composed of scripted actions. The automatically acquired observation models allow the robot to adapt to changes in the environment and robustly execute arbitrary plans. [PDF]
- TA for CMRoboBits: Creating an Intelligent AIBO Robot: This undergraduate robotics course introduces students to the components required to build a complete autonomous robot system. As a teaching assistant, my responsibilities included the leading the weekly lab sessions, design and grading of homework assignments and the presentation of 3-4 lectures over the course of the semester.
Humanoid Game Commentators
Research in robot soccer, in particular within RoboCup, has rapidly progressed since its beginnings in the mid 1990s. Robots successfully compete in teams, perceiving a challenging dynamic environment, making decisions, cooperating with teammates, and acting to achieve concrete objectives. Although the robots play the game, humans perform all the other functions associated with the competition, including commentating and refereeing. One challenging question is whether robots can also be used to perform these functions.
Using two Sony humanoid QRIO robots, we developed a team of two commentators for RoboCup AIBO Four-Legged League (now the Standard Platform League). The QRIO robots stand on the side lines of the field, autonomously observe the game, wirelessly listen to a "game controller," and coordinate their announcements with each other. The project was successfully demonstrated at RoboCup 2006 in Bremen, Germany.
Manuela Veloso, Nicholas Armstrong-Crews, Sonia Chernova, Elisabeth Crawford, Colin McMillen, Maayan Roth and Douglas Vail. A Team of Humanoid Game Commentators. In Proceedings of the International Conference on Humanoid Robots, Genova, Italy, 2006. [PDF]
Hands-Free Interaction for Human-Robot Teams
Robots have long assisted humans with repetitive, arduous, and especially dangerous tasks. The goal of this project was to design a robust human-robot interaction interface that can operate under highly variable conditions in complex, real-world environments. We developed a robotic system that is user-independent and robust to changes in lighting, variable terrain and sensor noise. The key contributions of this work include a rugged physical platform, active sensing to achieve fiducial-free person detection and tracking, intuitive robot control through the use of gesture and voice commands, and user feedback via speech synthesis. The result is a robust autonomous and hands-free robot capable of following the commands of a human, both indoors and out.
This work was performed at the iRobot Corporation in collaboration with Chris Jones, Odest Chadwicke Jenkins, Nathan Koenig, and Matthew Loper.
Engaging Long-Term Human-Robot Interaction
The Emotionally GrOunded (EGO) robot architecture developed by Sony aims to provide complex, natural and engaging long-term interaction between humans and robots. The goal of this project was to use the latest findings in cognitive psychology to extend the EGO architecture to include a mechanism for acquiring routine behaviors from repeated deliberative plans. Routine behavior is defined as the habitual performance of an established procedure or task. In humans, routine behavior occurs in a reactive manner, sometimes without awareness, and requires far less focused attention than conscious and highly supervised task execution, or deliberative behavior. We demonstrate a similar adaptational mechanism for robotic systems and present a method for learning robotic tasks and transferring their execution from deliberative to routine control. This work was applied to the fully autonomous Sony QRIO humanoid robot.
This work was performed at the Sony Intelligence Dynamics Laboratories in Tokyo, Japan in collaboration with Toshitada Doi, Masahiro Fujita, Hideki Shimomura, and Ronald Arkin.
Sonia Chernova and Ronald C. Arkin. From Deliberative to Routine Behaviors: a Cognitively-inspired Action Selection Mechanism for Routine Behavior Capture. Adaptive Behavior Journal, Vol. 15, No. 2, pages 199-216, 2007. [PDF]