Publications

Anticipation is a skill that enables complex decision making in humans and other biological agents. We review different implementations of anticipatory behavior in robots and give an overview on anticipation in biological systems. Based on an example of anticipatory behavior in humanoid robots, we discuss decision making and anticipation in artificial agents. We show that anticipation can enable fast decisions in highly dynamic and complex situations. Our findings are supported by experimental results performed in simulation and on real robots in large scale experiments.

Conducting games in RoboCup incurs high cost in terms of effort, time, and money. The scientific outcome, however, is quite limited and often not very conclusive. Especially, analyzing and drawing conclusions about the performance of complex processes like decision making of an individual robot or the behavior on the team level poses a considerable challenge. Collecting more data during the competition games will help to analyze the performance of algorithms, identify errors and areas for improvement, and make more signiffcant statements regarding the performance of the robots. In this work we investigate the possibilities for collection of the large scale RoboCup data and its analysis. We present a system for automatic recording of synchronized videos of RoboCup games and an application for exploration and annotation of large sets of RoboCup-related data. We also present data sets collected during the competitions in 2018 and an algorithm for visual detection and tracking of robots in the RoboCup videos. A first empirical evaluation shows promising results and demonstrates how such data can be integrated and used to validate robot's behavior.

This paper introduces a method for making fast decisions in a highly dynamic situation, based on forward simulation. This approach is inspired by the decision problem within the RoboCup domain. In this environment, selecting the right action is often a challenging task. The outcome of a particular action may depend on a wide variety of environmental factors, such as the robot's position on the field or the location of obstacles. In addition, the perception is often heterogeneous, uncertain, and incomplete.In this context, we investigate forward simulation as a versatile and extensible yet simple mechanism for inference of decisions. To evaluate an action, the outcome is simulated based on the estimated state of the situation. The simulation of a single action is split into a number of simple deterministic simulations - samples - based on the uncertainties of the estimated state and of the action model. Each of the samples is then evaluated separately, and the evaluations are combined and compared with those of other actions to inform the overall decision. This allows us to effectively combine heterogeneous perceptual data, calculate a stable decision, and reason about its uncertainty. This approach is implemented for the kick selection task in the RoboCup SPL environment and is actively used in competitions. In simulated experiments we validate the new scheme and evaluate different strategies.

This paper introduces a method for making fast decisions in a highly dynamic situation, based on forward simulation. This approach is inspired by the decision problem within the RoboCup domain. In this environment, selecting the right action is often a challenging task. The outcome of a particular action may depend on a wide variety of environmental factors, such as the robot's position on the field or the location of obstacles. In addition, the perception is often heterogeneous, uncertain, and incomplete. In this context, we investigate forward simulation as a versatile and extensible yet simple mechanism for inference of decisions. The outcome of each possible action is simulated based on the estimated state of the situation. The simulation of a single action is split into a number of simple deterministic simulations -- samples -- based on the uncertainties of the estimated state and of the action model. Each of the samples is then evaluated separately, and the evaluations are combined and compared with those of other actions to inform the overall decision. This allows us to effectively combine heterogeneous perceptual data, calculate a stable decision, and reason about its uncertainty. This approach is implemented for the kick selection task in the RoboCup SPL environment and is actively used in competitions. We present analysis of real game data showing significant improvement over our previous methods.

Information about objects and their positions in an environment are necessary requirements for most tasks of a mobile autonomous robot, in particular regarding the control of behavior and navigation. This presents a special challenge for robots with a limited view angle. Autonomously soccer playing robots in the dynamic environment of the RoboCup Standard Platform League are exposed to these difficulties. Most approaches aggregate all available information in one holistic model in order to localize robots. In case of inconsistent perceptions the model either turns noisy or creates and tracks an addiotional hypothesis. To improve the localization -- and thus the behavior control -- local models have received only little attention so far. In this work the implementation of a local goal model is presented and analyzed. A multi-hypothesis particle filter is used to cope with ambiguity of goal post percepts as well as to process incomplete and uncertain sensor information. Additionally, a percept buffer supports the initialization and also facilitates the handling of sparse false measurements. On the basis of this local goal model inconsistencies can be explicitly modeled, which may be used to stabilize the location of a robot.

The ability to grasp objects of different size and shape is one of the most important skills of a humanoid robot. Human grasping integrates a lot of different senses. In particular, the tactile sensing is very important for a stable grasping motion. When we lift a box without knowing what is inside, we do it carefully using our tactile and proprioceptive senses to estimate the weight and thus, the force necessary to hold and to lift this box. In this paper we present an adaptive controlling mechanism which enables a robot to grasp objects of different weights. Thereby, we only use the proprioceptive sensors like positions and electric current at the joints and force sensors at the end-effectors providing the robot with tactile feedback. We implemented and tested our approach on a humanoid robot.

Strategic positioning is a decisive part of the team play within a soccer game. In most solutions the positioning techniques are treated as a constituent of a complete team play strategy. In a comprehensive overview we discuss the team play and positioning methods used within RoboCup and extract the essential requirements for player positioning. In this work, we propose an approach for strategic positioning allowing for flexible formulation of arbitrary strategies. Based on the conditions of a specific strategy, the field is subdivided in regions by a Voronoi tessellation and each region is assigned a weight. Those weights influence the calculation of the optimal robot position as well as the path. A team play strategy can be expressed by the choice of the tessellation as well as the choice of the weights. This provides a powerful abstraction layer simplifying the design of the actual play strategy. We also present an implementation of an example strategy based on this approach and analyse the performance of our approach in simulation.

The ability to grasp objects of different size and shape is one of the most important skills of a humanoid robot. There are a lot of different approaches tackling this problem; however, there is no general solution. The complexity and the skill of a possible grasping motion depend hardly on a particular robot. In this paper we analyze the kinematic and sensory grasping abilities of the humanoid robot Nao. Its kinematic constraints and hand's mechanical structure represent an interesting case of study due to lack of actuators for fingers and the limited computation power. After describing the platform and studying its capabilities, we propose some simple controllers and we present a benchmark based on some experimental data.

Numerous simulators have been developed over the years to assist robotics research in the development, testing, and evaluation. Nevertheless, there is still a big gap between the simulation and the reality. This makes it difficult to transfer methods and code. The 3D simulator | SimSpark is developed and used by a big community of AI researchers in RoboCup. But up to now there are only few applications to real robots. In this paper, we discuss the general possibilities how the SimSpark simulator can be used to support research in cognitive robotics and present applications on the humanoid robot Nao. As a result of our investigation we have developed a unified team playing both in Simulation League and Standard Platform League in RoboCup.

Automatic, full body motion generation for humanoid robots presents a formidable computational challenge. The kicking motion is one of the most important motions in a soccer game. However, at the current state the most common approaches of implementing this motion are based on key frame technique. Such solutions are inflexible and cost a lot of time to adjust robot’s position. In this paper we present an approach for adaptive control of the motions. We implemented our approach in order to solve the task of kicking the ball on a humanoid robot Nao. The approach was tested both in simulation and on a real robot.

The performance of a soccer robot is highly dependent on its motion ability. The kicking motion is one of the most important motions in a soccer game. However, automatic, full body motion generation for humanoid robots presents a formidable computational challenge. At the current state the most common approaches of implementing this motion are based on key frame technique. Such solutions are inflexible, i.e., in order to adjust the aimed direction of the kick the robot has to walk around the ball. The adjustment costs a lot of time especially if some precise adjustments have to be done, e.g., for a penalty kick. In this paper we present an approach for adaptive control of the motions. We implemented our approach in order to solve the task of kicking the ball on a humanoid robot Nao. The approach was tested both in simulation and on a real robot.

An appropriate architecture (i.e., framework) is the base of each successful heterogeneous software project. It enables a group of developers to work at the same project and to organize their solutions. From this point of view, the artificial intelligence and/or robotics related research projects are usually more complicated, since the actual result of the project is often not clear. In particular, a strong organization of the software is necessary if the project is involved in education. Obviously, there is no perfect framework which could satisfy all the needs of the developers. In this paper we present a modular software architecture designed to implement an autonomous agent. In particular, it is used to develop software which is used simultaneously at several platforms (e.g., humanoid robot, simulated agent). One of the main aspects considered in our design is a strong code modularization which allows for re-usability, transparency and easily testing. Other important aspects are real-time applicability and simple usage. This paper presents the main concepts and the particular implementation of the important parts. We also provide a qualitative comparison with other existing robotics frameworks.

In this paper we analyze the kinematic and sensory grasping abilities of the humanoid robot Nao. Its kinematic constraints and hand's mechanical structure represent an interesting case of study due to lack of actuators for fingers and the limited computation power. After describing the platform and studying its capabilities, we propose some simple controllers and we present a benchmark based on some experimental data.

Many tasks of a mobile robot, e.g., navigation, require the knowledge of the positions of the objects in the surrounding environment. This task is especially challenging for the robots which perception is based on a directed visual system, e.g., a camera with a limited view angle. The incomplete and noisy sensor information leads to the uncertainty in the robots belief of the world. An appropriate model of the world may enable the robot to make plans and to realize complex behavior. The state of the art modeling methods use often only a small part of the available information. In particular the redundant information remain unused. In this work we investigate methods to exploit effectively the redundant information in order to get a better model of the world. In the first part we discuss a number of possibilities to use of specific properties of the objects to estimate the parameters of the camera matrix. In the second part we present a constraint based approach for the world modeling.

In this paper we present a novel approach using constraint based techniques for world modeling, i.e. self localization and object modeling. Within the last years, we have seen a reduction of landmarks such as beacons or colored goals within the RoboCup domain. Using other features as line information becomes more important. Using such sensor data is tricky, especially when the resulting position belief is stretched over a larger area. Constraints can overcome this limitations, as they have several advantages: they can represent large distributions and are easy to store and to communicate to other robots. Propagation of several constraints can be computationally cheap. Even high dimensional belief functions can be used. We will describe a sample implementation and show experimental results.

The most of the vision based self-localization methods are using landmarks to estimate the position of the robot. The results of those methods depend highly on the precision of the perceptual information provided by the vision system. Depending on situation, some landmarks provide more certain measurements than others. We present a general criterion to predict the sensitivity of measurements concerning errors. In addition an algorithm is presented for automatic selection of optimal landmarks. This enables the robot to choose actively those landmarks that provide the highest certainty, which leads to better localization results. We demonstrate the performance and accuracy of the algorithm through a series of experiments.

In this paper we present a novel approach using constraint based techniques for world modeling, i.e. self localization and object modeling. Within the last years, we have seen a reduction of landmarks as beacons, colored goals, within the RoboCup domain. Using other features as line information becomes more important. Using such sensor data is tricky, especially when the resulting position belief is stretched over a larger area. Constraints can overcome this limitations, as they have several advantages: They can represent large distributions and are easy to store and to communicate to other robots. Propagation of a several constraints can be computationally cheap. Even high dimensional belief functions can be used. We will describe a sample implementation and show experimental results.

Mobile autonomous robotics is a very young and complex field of research. Only in recent decades have robots become able to explore, to move, navigate and to interact with their environment. Since the world is uncertain and since robots can only gain partial information about it, probabilistic navigation algorithms have become very popular whenever a robot has to localize itself or surrounding objects. Furthermore, cooperative exploration and localization approaches have become very relevant lately, as robots begin to act not just alone but in groups. Within this thesis a new approach using the concept of spatial percept-relations for cooperative environment modeling is presented and evaluated. As a second contribution, constraint based localization techniques will be introduced for having a robot or a group of robots efficiently localized and to model their environment.

Modeling the environment is crucial for a mobile robot. Common approaches use Bayesian filters like particle filters, Kalman filters and their extended forms. We present an alternative and supplementing approach using constraint techniques based on spatial constraints between object positions. This yields several advantages: a) the agent can choose from a variety of belief functions, b) the computational complexity is decreased by efficient algorithms. The focus of the paper are constraint propagation techniques under the special requirements of navigation tasks.

In this paper we present a novel approach using constraint based techniques for world modeling, i.e. self localization and object modeling. Within the last years, we have seen a reduction of landmarks as beacons, colored goals, within the RoboCup domain. Using other fea- tures as line information becomes more important. Using such sensor data is tricky, especially when the resulting position belief is stretched over a larger area. Constraints can overcome this limitations, as they have several advantages: They can represent large distributions and are easy to store and to communicate to other robots. Propagation of a several constraints can be computationally cheap. Even high dimensional belief functions can be used. We will describe a sample implementation and show experimental results.

Common approaches for robot navigation use Bayesian filters like particle filters, Kalman filters and their extended forms. We present an alternative and supplementing approach using constraint techniques based on spatial constraints between object positions. This yields several advantages. The robot can choose from a variety of belief functions, and the computational complexity is decreased by efficient algorithms. The paper investigates constraint propagation techniques under the special requirements of navigation tasks. Sensor data are noisy, but a lot of redundancies can be exploited to improve the quality of the result. We introduce two quality measures: The ambiguity measure for constraint sets defines the precision, while inconsistencies are measured by the inconsistency measure. The measures can be used for evaluating the available data and for computing best fitting hypothesis. A constraint propagation algorithm is presented.

In this paper we will present an application for constraint based methods to self localize within the RoboCup domain. During a robotic soccer game, robots of a team need to know where they and their team mates are on the Field, therefore they need to localize themselves. For self localization, constraint based methods can be an effective alternative to classic Bayesian approaches as Kalman filters or Monte-Carlo methods. In this paper we will present, how constraint based techniques can be applied to a humanoid robot. Therefore we will implement constraint based methods in a humanoid robot "NAO" and see how the constraint based approach works within the Standard Platform League.

In this paper we introduce a state-estimation method that uses a short-term memory to calculate the current state. A common way to solve state estimation problems is to use implementations of the Bayesian algorithm like Kalman filters or particle filters. When implementing a Bayesian filter several problems can arise. One difficulty is to obtain error models for the sensors and for the state transitions. The other difficulty is to find an adequate compromise between the accuracy of the belief probability distribution and the computational costs that are needed to update it. In this paper we show how a short-term memory of perceptions and actions can be used to calculate the state. In contrast to the Bayesian filter, this method does not need an internal representation of the state which is updated by the sensor and motion information. It is shown that this is especially useful when information of sparse sensors (sensors with non-unique measurements with respect of the state) has to be integrated.

Robots perceiving its environment using cameras usually need a good representation of how the camera is aligned to the body and how the camera is rotated relative to the ground. This is especially important for bearing-based distance measurement. In this paper we show how to use reference objects to improve vision-based distance measurements to objects of unknown size. Several methods for different kinds of reference objects are introduced. These are objects of known size (like a ball), objects extending over the horizon (like goals and beacons), and objects with known shape on the ground (like field lines). We give a detailed description how to determine the rotation of the robot's camera relative to the ground, provide an error-estimation for all methods and describe the experiments we performed on an Aibo robot.

Robots perceiving their environment using cameras usually need a good representation of how the camera is aligned to the body and how the camera is rotated relative to the ground. This is especially important for bearing-based distance measurements. In this paper we show how to use reference objects to improve vision-based distance measurements to objects of unknown size. Several methods for different kinds of reference objects are introduced. These are objects of known size (like a ball), objects extending over the horizon (like goals and beacons), and objects with known shape on the ground (like field lines). We give a detailed description how to determine the rotation of the robot's camera relative to the ground, provide an error-estimation for all methods and describe the experiments we performed on an Aibo robot.

Robots perceiving their environment using cameras usually need a good representation of how the camera is aligned to the body and how the camera is rotated relative to the ground. This is especially important for bearing-based distance measurements. In this paper we show how to use reference objects to improve vision-based distance measurements to objects of unknown size. Several methods for different kinds of reference objects are introduced. These are objects of known size (like a ball), objects extending over the horizon (like goals and beacons), and objects with known shape on the ground (like field lines). We give a detailed description how to determine the rotation of the robot's camera relative to the ground, provide an error-estimation for all methods and describe the experiments we performed on an Aibo robot.