In this paper, we present our design and experiments on a planar biped robot under the control of a pure sensor-driven controller. This design has some special mechanical features, for example small curved feet allowing rolling action and a properly positioned center of mass, that facilitate fast walking through exploitation of the robot’s natural dynamics. Our sensor-driven controller is built with biologically inspired sensor- and motor-neuron models, and does not employ any kind of position or trajectory tracking control algorithm. Instead, it allows our biped robot to exploit its own natural dynamics during critical stages of its walking gait cycle. Due to the interaction between the sensor-driven neuronal controller and the properly designed mechanics of the robot, the biped robot can realize stable dynamic walking gaits in a large domain of the neuronal parameters. In addition, this structure allows the use of a policy

This paper gives a brief overview of the work done in the past year. A detailed description of the entire system used in the competition---including its underlying architecture---can be found at [2]

Abstract — Robot learning is a growing area of research at the intersection of robotics and machine learning. The main contributions of this article include a review of how machine learning has been used on Sony AIBO robots and at RoboCup with a focus on the four-legged league during the years 1998– 2004. The review shows that the application oriented use of machine learning in the four-legged league was still conservative and restricted to a few well-known and easy to use methods such as standard decision trees, evolutionary hill climbing, and support vector machines. Method oriented spin-off studies emerged more frequently and increasingly addressed new and advanced machine learning techniques. Further the article presents some details about the growing impact of machine learning in the software system developed by authors ’ robot soccer team—the NUbots.

www.sim.informatik.tu-darmstadt.de

Abstract. This paper describes the hardware and software design of the kidsize humanoid robot systems of the Darmstadt Dribblers in 2007. The robots are used as a vehicle for research in control of locomotion and behavior of autonomous humanoid robots and robot teams with many degrees of freedom and many actuated joints. The Humanoid League of RoboCup provides an ideal testbed for such aspects of dynamics in motion and autonomous behavior as the problem of generating and maintaining statically or dynamically stable bipedal locomotion is predominant for all types of vision guided motions during a soccer game. A modular software architecture as well as further technologies have been developed for efficient and effective implementation and test of modules for sensing, planning, behavior, and actions of humanoid robots. 1

The locomotion of a quadruped robot was examined with the emphasis on the creation, optimization and merging of motions and gaits. A Fourier Series Expansion was performed on the controller commands used in the robot gait. It was found that omitting higher order terms has very little, at times even positive e#ect on the robot's performance yet yields a greatly reduced parameter set describing the motion. The parameter reduction is desirable for creating, optimizing and adapting motions. Furthermore, the acquired frequency space representation allows for simple creation of new motions by merging of existing ones and also for creating transitions from one type of motion to another.

The design considerations for a small, relatively fast walking, autonomous humanoid robot are presented. The robot must be energy efficient but also produce sufficient torque to reach greater speeds. On the basis of previous investigations into gait optimization for multilegged systems, dynamical modeling and nonlinear optimization tools are used for design optimization and choosing the motor size and gear ratios. Gait trajectories for relatively fast steps and different prototypes were calculated. The design decisions are described for the humanoid robot with 6 degrees-of-freedom (DoF) in each leg and 2 DoF in each arm based on numerical results and preliminary investigations with a 4 DoF test robot.

Abstract. Developing and testing the key modules of autonomous humanoid soccer robots (e.g., for vision, localization, and behavior control) in software-in-the-loop (SIL) experiments, requires real-time simulation of the main motion and sensing properties. These include humanoid robot kinematics and dynamics, the interaction with the environment, and sensor simulation, especially the camera properties. To deal with an increasing number of humanoid robots per team the simulation algorithms must be very efficient. In this paper, the simulator framework MuRoSimF (Multi-Robot-Simulation-Framework) is presented which allows the flexible and transparent integration of different simulation algorithms with the same robot model. These include several algorithms for simulation of humanoid robot motion kinematics and dynamics (with O(n) runtime complexity), collision handling, and camera simulation including lens distortion. A simulator for teams of humanoid robots based on MuRoSimF is presented. A unique feature of this simulator is the scalability of the level of detail and complexity which can be chosen individually for each simulated robot and tailored to the requirements of a specific SIL test. Performance measurements are given for real-time simulation on a moderate laptop computer of up to six humanoid robots with 21 degrees of freedom, each equipped with an articulated camera. 1

This paper discusses the design concept and system development of a small and relatively fast walking, autonomous humanoid robot with 17 degrees-offreedom (DoF). The selection of motor size and gear ratios is based on numerical optimization of detailed multibody dynamics and optimal control corresponding to fast steps of the robot with an envisioned target speed of more than 0.5 m/s. In this paper the design considerations based on numerical optimal control studies and the mechanical realization of the robot are presented including first investigations on the achievable performance of a decentralized, microcontrollerbased control architecture.

Abstract Evolutionary Robotics (ER) is a promising methodology, intended for the autonomous development of robots, in which their behaviors are obtained as a consequence of the structural coupling between robot and environment. It is essential that there be a great amount of interaction to generate complex behaviors. Thus, nowadays, it is common to use simulation to speed up the learning process; however simulations are achieved from arbitrary off-line designs, rather than from the result of embodied cognitive processes. According to the reality gap problem, controllers evolved in simulation usually do not allow the same behavior to arise once transferred to the real robot. Some preliminary approaches for combining simulation and reality exist in the ER literature; nonetheless, there is no satisfactory solution available. In this work we discuss recent advances in neuroscience as a motivation for the use of environmentally adapted simulations, which can be achieved through the co-evolution of robot behavior and simulator. We present an algorithm in which only the differences between the behavior fitness obtained in reality versus that obtained in simulations are used as feedback for adapting a simulation. The proposed algorithm is experimentally validated by showing the successful development and continuous transference to reality of two complex low-level behaviors with Sony AIBO 1 robots: gait optimization and ball-kicking behavior.