RoboCup 2012: Robot Soccer World Cup XVI

Maintaining balance in the presence of disturbances is crucial for bipedal robots. In this paper, we focus on the lateral motion component. In order to attain disturbance rejection and to quickly recover balance, we combine three different control approaches. As a principal building block, we generate center of mass trajectories with a linear model predictive controller that takes scheduled footsteps into account. Strong disturbances generate unexpected angular momenta that can compromise stability. A second control layer extends the underlying preview controller with two recovery strategies that modify the planned CoM trajectories to dampen the rotational velocity of the robot and adapt the timing of the steps according to the expected orbital energy of CoM trajectories at support exchange. Experiments with a real Nao robot show that the system is able to recover from lateral disturbances as long as the robot does not tip over the current support leg.

The Soccer Simulation 2D League is one of the oldest competitions among the RoboCup leagues. In the simulation 2D league, the simulator enables two teams of 11 simulated autonomous agents to play a game of soccer with highly realistic rules and game play. This paper introduces the RoboCup 2012 Soccer Simulation 2D League champion team, HELIOS2012, a joint team of Fukuoka University and Osaka Prefecture University.

Inside the RoboCup Rescue Simulation League, the mission is to use robots to rescue as many victims as possible after a disaster. The research challenge is to let the robots cooperate as a team. This year in total 15 teams from 8 different countries have been active in the competitions. This paper highlights the approaches of the winners of the virtual robot competition, the infrastructure competition, and the agent competition.

In 2012, UT Austin Villa claimed Standard Platform League championships at both the US Open and RoboCup 2012 in Mexico City. This paper describes the key contributions that led to the team’s victories. First, UT Austin Villa’s code base was developed on a solid foundation with a flexible architecture that enables easy testing and debugging of code. Next, the vision code was updated this year to take advantage of the dual cameras and better processor of the new V4 Nao robots. To improve localization, a custom localization simulator allowed us to implement and test a full team solution to the challenge of both goals being the same color. The 2012 team made use of Northern Bites’ port of B-Human’s walk engine, combined with novel kicks from the walk. Finally, new behaviors and strategies take advantage of opportunities for the robot to take time to setup for a long kick, but kick very quickly when opponent robots are nearby. The combination of these contributions led to the team’s victories in 2012.

The new RoboCup Logistics League sponsored by Festo offers a competition within a simulated industrial environment. In order to solve the logistical tasks, all three Robotinos not only have to operate autonomously in a flexible, effective and robust way on their own, they should also collaborate efficiently in order to maximize the overall outcome. In this paper, the first world champion of the Logistics League, TUMsBendingUnits from the Technical University of Munich (TUM), presents their logistical system with focus on approaches concerning robot hardware modifications, software architecture, task planning and execution, multi-robot collaboration, visual perception, motion planning and execution.

Autonomous soccer-playing humanoid robots have advanced significantly in the past few years. Skill sets elementary to humans such as omnidirectional bipedal walking, path planning, and gameplay strategy have matured enough to allow for dynamic and exciting games. In this paper team CHARLI, the two-time RoboCup Humanoid AdultSize League winner, describes the design and fabrication of essential components such as the spine and mechanical structure, then overviews the increase in performance resulting from recent mechanical upgrades. Finally, we detail the custom walking controller and gameplay module changes responsible for the outstanding performance of our self-constructed lightweight full-sized humanoid platform, CHARLI-2.

One of today’s overall efforts in mobile industrial robotics is the enhancement of autonomy and flexibility considering required safety issues. The new league RoboCup@Work being carried out for the first time in Mexico City, Mexico 2012, focuses on boosting research activities in this field in order to create new, innovative ideas and concepts meeting industrial needs.

This paper introduces the new league. Furthermore, it presents the approaches of the winner team LUHbots, Leibniz Universität Hannover, Hanover, Germany, at each competition in detail.

The UT Austin Villa team, from the University of Texas at Austin, won the RoboCup 3D Simulation League in 2012 having also won the competition the previous year. This paper describes the changes and improvements made to the team between 2011 and 2012 that allowed it to repeat as champions.

Over the past few years, soccer-playing humanoid robots advanced significantly. Elementary skills, such as bipedal walking, visual perception, and collision avoidance have matured enough to allow for dynamic and exciting games. In this paper, team NimbRo TeenSize, the winner of the RoboCup 2012 Best Humanoid Award, presents its robotic platform and its approaches to perception and behavior control.

In this paper we describe details of our winning team Nimb-Ro@Home at the RoboCup@Home competition 2012. This year we improved the gripper design of our robots and further advanced mobile manipulation capabilities such as object perception and manipulation planning. For human-robot interaction, we propose to complement face-to-face communication between user and robot with a remote user interface for handheld PCs. We report on the use of our approaches and the performance of our robots at RoboCup 2012.

Exploration of unknown environments is an enabling task for several applications, including map building and search and rescue. It is widely recognized that several benefits can be derived from deploying multiple mobile robots in exploration, including increased robustness and efficiency. Two main issues of multirobot exploration are the

exploration strategy

employed to select the most convenient observation locations the robots should reach in a partially known environment and the

coordination method

employed to manage the interferences between the actions performed by robots. From the literature, it is difficult to assess the relative effects of these two issues on the system performance. In this paper, we contribute to filling this gap by studying a search and rescue setting in which different coordination methods and exploration strategies are implemented and their contributions to an efficient exploration of indoor environments are comparatively evaluated. Although preliminary, our experimental data lead to the following results: the role of exploration strategies dominates that of coordination methods in determining the performance of an exploring multirobot system in a highly structured indoor environment, while the situation is reversed in a less structured indoor environment.

This paper introduces an optimised method for extracting natural landmarks to improve localisation during RoboCup soccer matches. The method uses modified 1D SURF features extracted from pixels on the robot’s horizon. Consistent with the original SURF algorithm, the extracted features are robust to lighting changes, scale changes, and small changes in viewing angle or to the scene itself. Furthermore, we show that on a typical laptop 1D SURF runs more than one thousand times faster than SURF, achieving sub-millisecond performance. This makes the method suitable for visual navigation of resource constrained mobile robots. We demonstrate that by using just two stored images, it is possible to largely resolve the RoboCup SPL field end ambiguity.

Robot soccer is one of the major domains for studying the coordination of multi-robot teams. Decentralized Partially Observable Markov Decision Process (Dec-POMDP) is a recent mathematical framework which has been used to model multi-agent coordination. In this work, we model simple robot soccer as Dec-POMDP and solve it using an algorithm which is based on the approach detailed in [1]. This algorithm uses finite state controllers to represent policies and searches the policy space with genetic algorithms. We use the

TeamBots

simulation environment. We use score difference of a game as a fitness and try to estimate it by running many simulations. We show that it is possible to model a robot soccer game as a Dec-POMDP and achieve satisfactory results. The trained policy wins almost all of the games against the standard

TeamBots

teams, and a reinforcement learning based team developed elsewhere.

The RoboCup soccer simulation 2D domain is a very large testbed for the research of planning and machine learning. It has competed in the annual world championship tournaments in the past 15 years. However it is still unclear that whether more principled techniques such as decision-theoretic planning take an important role in the success for a RoboCup 2D team. In this paper, we present a novel approach based on MAXQ-OP to automated planning in the RoboCup 2D domain. It combines the benefits of a general hierarchical structure based on MAXQ value function decomposition with the power of heuristic and approximate techniques. The proposed framework provides a principled solution to programming autonomous agents in large stochastic domains. The MAXQ-OP framework has been implemented in our RoboCup 2D team, WrightEagle. The empirical results indicated that the agents developed with this framework and related techniques reached outstanding performances, showing its potential of scalability to very large domains.

The ability to detect people in domestic and unconstrained environments is crucial for every service robot. The knowledge where people are is required to perform several tasks such as navigation with dynamic obstacle avoidance and human-robot-interaction. In this paper we propose a people detection approach based on 3

d

data provided by a RGB-D camera. We introduce a novel 3

d

feature descriptor based on Local Surface Normals (LSN) which is used to learn a classifier in a supervised machine learning manner. In order to increase the systems flexibility and to detect people even under partial occlusion we introduce a top-down/bottom-up segmentation. We deployed the people detection system on a real-world service robot operating at a reasonable frame rate of 5

Hz

. The experimental results show that our approach is able to detect persons in various poses and motions such as sitting, walking, and running.

This paper reports a series of simulation competitions on domestic robots. All of these five competitions were based on a simulation platform focused on evaluating high-level functions of a domestic robot, including task planning and dialogue understanding. The object of holding these competitions is to promote research and development of service robots while avoiding limitations imposed by hardware of real robots. We also analyze the results and performances of participating teams since the competition was first held in 2009, showing that more and more terms are participating and they are performing better and better.

Humanoid robots have a large number of degrees of freedom (DoFs), therefore motor learning by such robots which explore the optimal parameters of behaviors is one of the most serious issues in humanoid robotics. In contrast, it has been suggested that humans can solve such a problem by synchronizing many body parts in the early stage of learning, and then desynchronizing their movements to optimize a behavior for a task. This is called as ”Freeze and Release.” We hypothesize that heuristic exploration through synchronization and desynchronization of DoFs accelerates motor learning of humanoid robots. In this paper, we applied this heuristic to a throwing skill learning in soccer. First, all motors related to the skill are actuated in a synchronized manner, thus the robot explores optimal timing of releasing a ball in one-dimensional search space. The DoFs are released gradually, which allows to search for the best timing to actuate the motors of all joints. The real robot experiments showed that the exploration method was fast and practical because the solution in low-dimensional subspace was approximately optimum.

This paper presents a dynamic role assignment and formation positioning system used by the 2011 RoboCup 3D simulation league champion UT Austin Villa. This positioning system was a key component in allowing the team to win all 24 games it played at the competition during which the team scored 136 goals and conceded none. The positioning system was designed to allow for decentralized coordination among physically realistic simulated humanoid soccer playing robots in the partially observable, non-deterministic, noisy, dynamic, and limited communication setting of the RoboCup 3D simulation league simulator. Although the positioning system is discussed in the context of the RoboCup 3D simulation environment, it is not domain specific and can readily be employed in other RoboCup leagues as it generalizes well to many realistic and real-world multiagent systems.

Crowd evacuation simulations are useful tools for analyzing and assessing the safety of building occupants. Agent-based simulations provide a platform for computing individual as well as and collective behaviors in crowds. During an evacuation, it is well known that trained leaders or evacuation guidance play a key role in saving human lives. In this paper, we propose an evacuation simulation system where agents are guided by evacuation orders from authorities. The simulations captured typical behaviors observed during crowd evacuation. For example, the total evacuation time was reduced when most of the agents followed the guidance, although the evacuation times of individual agents were different. When a specific agent is involved in the movement of other agents to a different destination, the evacuation takes a longer amount of time. The simulation appears to depict real-life situations well, which shows that simulations can be a useful tool to estimate evacuation situations prior to emergency evacuation drills.

Biped soccer robots have shown drastic improvements in motion skills over the past few years. Still, a lot of work needs to be done with the RoboCup Federation’s vision of 2050 in mind. One goal is creating a workflow for quickly generating reliable motions, preferably with inexpensive and accessible hardware. Our hypothesis is that using Microsoft’s Kinect sensor in combination with a modern optimization algorithm can achieve this objective. We produced four complex and inherently unstable motions and then applied three contemporary optimization algorithms (CMA-ES, xNES, PSO) to make the motions robust; we performed 900 experiments with these motions on a 3D simulated Nao robot with full physics. In this paper we describe the motion mapping technique, compare the optimization algorithms, and discuss various basis functions and their impact on the learning performance. Our conclusion is that there is a straightforward process to achieve complex and stable motions in a short period of time.

The development of high-level behavior for autonomous robots is a time-consuming task even for experts. This paper presents a Computer-Aided Software Engineering (CASE) tool, named Kouretes Statechart Editor (KSE), which enables the developer to easily specify a desired robot behavior as a statechart model utilizing a variety of base robot functionalities (vision, localization, locomotion, motion skills, communication). A statechart is a compact platform-independent formal model used widely in software engineering for designing software systems. KSE adopts the Agent Systems Engineering Methodology (ASEME) model-driven approach. Thus, KSE guides the developer through a series of design steps within a graphical environment that leads to automatic source code generation. We use KSE for developing the behavior of the Nao humanoid robots of our team Kouretes competing in the Standard Platform League of the RoboCup competition.

Decision making in complex, multi-agent and dynamic environments such as disaster spaces is a challenging problem in Artificial Intelligence. This research paper aims at developing distributed coordination and cooperation method based on reinforcement learning to enable team of homogeneous, autonomous fire fighter agents, with similar skills to accomplish complex task allocation, with emphasis on firefighting tasks in disaster space. The main contribution is applying reinforcement learning to solve the bottleneck caused by dynamicity and variety of conditions in such situations as well as improving the distributed coordination of fire fighter agent’s to extinguish fires within a disaster zone. The proposed method increases the speed of learning; it has very low memory usage and has a good scalability and robustness in the case that the number of agents and complexity of task increases. The effectiveness of the proposed method is shown through simulation results.

We developed a scene text recognition system with active vision capabilities, namely: auto-focus, adaptive aperture control and auto-zoom. Our localization system is able to delimit text regions in images with complex backgrounds, and is based on an attentional cascade, asymmetric adaboost, decision trees and Gaussian mixture models. We think that text could become a valuable source of semantic information for robots, and we aim to raise interest in it within the robotics community. Moreover, thanks to the robot’s pan-tilt-zoom camera and to the active vision behaviors, the robot can use its affordances to overcome hindrances to the performance of the perceptual task. Detrimental conditions, such as poor illumination, blur, low resolution, etc. are very hard to deal with once an image has been captured and can often be prevented. We evaluated the localization algorithm on a public dataset and one of our own with encouraging results. Furthermore, we offer an interesting experiment in active vision, which makes us consider that active sensing in general should be considered early on when addressing complex perceptual problems in embodied agents.

Computer vision systems frequently employ colour segmentation as a step of feature extraction. This is particularly crucial in an environment where important features are colour-coded, such as robot soccer. This paper describes a method for determining an appropriate colour model by measuring the compactness and separation of clusters produced by the

k

-means algorithm.

RGB

,

HSV

,

YC

b

C

r

and

CIE L*a*b*

colour models are assessed for a selection of artificial and real images, utilising an implementation of the Dunn’s-based cluster validation index. The effectiveness of the method is assessed by qualitatively comparing the relative correctness of the segmentation to the results of the cluster validation. Results demonstrate a significant variation in segmentation quality among colour spaces, and that YC

b

C

r

is the best choice for the DARwIn-OP platform tested.

In order to be able to beat the world champion human soccer team in the year 2050, soccer playing robots will need to have very robust vision systems that can cope with drastic changes in illumination conditions. However, the current vision systems are still brittle and they require exhaustive and repeated color calibration procedures to perform acceptably well. In this paper, we investigate the suitability of biologically inspired saliency-based visual attention models for developing robust vision systems for soccer playing robots while focusing on the illumination invariance aspect of the solution. The experiment results demonstrate successful and accurate detection of the ball even when the illumination conditions change continuously and dramatically.

In this paper, bag-of-features, a popular and successful approach in pattern recognition community, is used to realize place recognition based on omnidirectional vision for mobile robots by combining the real-time local visual features proposed by ourselves for omnidirectional vision and support vector machines. The panoramic images from the COLD database were used to perform experiments to determine the best algorithm parameters and the best training condition. The experimental results show that the robot can realize robust place recognition with high classification rate in real-time by using our algorithm.

The trend to have more cooperative play and the increase of game dynamics in Robocup MSL League motivates the improvement of skills for ball passing and reception. Currently the majority of the MSL teams uses ball handling devices with rollers to have more precise kicks but limiting the capability to kick a moving ball without stopping it and grabbing it. This paper addresses the problem to receive and kick a fast moving ball without having to grab it with a roller based ball handling device. Here, the main difficulty is the high latency and low rate of the measurements of the ball sensing systems, based in vision or laser scanner sensors.Our robots use a geared leg coupled to a motor that acts simultaneously as the kicking device and low level ball sensor. This paper proposes a new method to improve the capability for ball sensing in the kicker, by combining high rate measurements from the torque and energy in the motor and angular position of the kicker leg. The developed method endows the kicker device with an effective ball detection ability, validated in several game situations like in an interception to a fast pass or when chasing the ball where the relative speed from robot to ball is low. This can be used to optimize the kick instant or by the embedded kicker control system to absorb the ball energy.

Multi-robot systems are increasingly present in nowadays applications. In order to allow an effective decision making, a reliable world representation is required. In a team of robots performing a given task, it is beneficial to share information about the world. In this work, a multi-object, multi-sensor and cooperative tracking method is proposed for the Robocup Standard Platform League (SPL), where two teams of humanoid robots play soccer against each other. Each robot is equipped with two low-cost, noisy, and narrowed-field-of-view cameras and two noisy sonar sensors. In addition, they are endowed with a wireless communication hardware. The on-board computer is a low capacity processing unit (x86 500[Mhz]). The proposed tracking system uses all these hardware elements and it is distributed, in the sense that it is executed in every robot. The proposed tracking system is validated in simulations and in real experiments. Main results show an important improvement on simulated and real results when tracking mobile-objects.

Robotic vision is a challenging problem due both to the uncertain nature of real-world environments and the computational constraints of mobile platforms. Many standard computer vision algorithms have high computational requirements and are often seen as unsuitable for use in an embedded system such as the Aldebaran Nao. Many current approaches use fragile algorithms and reduced resolutions to achieve a low processing latency. We implement the Hough Transform, a standard line detection algorithm, for real-time use in the RoboCup Standard Platform League. Using assembly language instructions to expose the processor’s full potential, this project implements a real-time Hough transform for line detection on 320x240 pixel images.

The RoboCup is an initiative to promote the development of robotics in a social relevant way. The competition consists of several leagues and it would be beneficial if developments in one league could be reused in other leagues. This paper describes the development of a simulation model for a humanoid robot inside USARSim, which could be the basis of synergy between the Rescue Simulation, Soccer Simulation and @Home League. USARSim is an existing 3D simulator based on the Unreal Engine, which provides facilities for good quality rendering, physics simulation, networking, a highly versatile scripting language and a powerful visual editor. This simulator is now extended with the dynamics of a walking robot and validated for the humanoid robot Nao. On this basis many other robotic applications as benchmarked in the RoboCup initiative become possible.

Faults that occur in an autonomous robot system negatively affect its dependability. The aim of truly dependable and autonomous systems requires that one has to deal with these faults in some way. In order to be able to do this efficiently one has to have information on the nature of these faults. Very few studies on this topic have been conducted so far. In this paper we present results of a survey on faults of autonomous robots we conducted in the context of RoboCup. The major contribution of this paper is twofold. First we present an adapted fault taxonomy suitable for autonomous robots. Second we give information on the nature, the relevance and impact of faults in robot systems that are beneficial for researcher dealing with fault mitigation and management in autonomous systems.

Training robot or agent behaviors by example is an attractive alternative to directly coding them. However training complex behaviors can be challenging, particularly when it involves interactive behaviors involving multiple agents. We present a novel hierarchical learning from demonstration system which can be used to train both single-agent and scalable cooperative multiagent behaviors. The methodology applies manual task decomposition to break the complex training problem into simpler parts, then solves the problem by iteratively training each part. We discuss our application of this method to multiagent problems in the humanoid RoboCup competition, and apply the technique to the keepaway soccer problem in the RoboCup Soccer Simulator.

This paper evaluates the benefits of modeling the dynamic environment of robot soccer games as a SLAM problem. Moving objects such as other robots and the ball are not only tracked individually, but modeled in a full state and used for localization at the same time. This is described as an implementation of an efficient system capable of running in real time on limited platforms such as the humanoid robot Nao, and the system’s benefit is evaluated using real world experiments.

The development of estimation systems based on Kalman filters requires several design choices. Among others, these are the methods used for linearization, coordinate systems for measurement representations, and approximations such as how to handle multiple simultaneous observations per time step. This paper evaluates these different choices with respect to their influence on the system’s estimation quality and points out simple yet effective solutions. Camera-based localization for a humanoid robot is chosen as an example application and the localization benefit of different approaches is evaluated using real and simulated feature perceptions.