Notions of similarity and distance play an important role in informatics. Different disciplines have developed their own treatment of related measures. We consider this problem under the viewpoint of Case Based Reasoning. While distance and similarity can be considered to be formally equivalent, there exist some differences concerning their intuitive use which have impact to the composition of global measures from local ones.

Similarity and Distance in Case Based Reasoning

The problem of defining robot behaviors to completely address a large and complex set of situations is very challenging. We present an approach for robot's action selection in the robot soccer domain using Case-Based Reasoning techniques. A case represents a snapshot of the game at time t and the actions the robot should perform in that situation. We basically focus our work on the retrieval and reuse steps of the system, presenting the similarity functions and a planning process to adapt the current problem to a case. We present first results of the performance of the system under simulation and the analysis of the parameters used in the approach.

/pdf/retrieving-and-reusing-game-plays-for-robot-soccer-2ub2p6gkgo.pdf

Retrieving and reusing game plays for robot soccer

In this paper, we describe a novel self-localization algorithm. Self-localization methods are required for lowering the computational cost and handling vague sensor data. Thus, we propose to use only the uniform distribution to represent probability distributions in Monte Carlo localization, and name this method a uniform Monte Carlo localization (Uniform MCL). We manifest the low computational cost and robustness of Uniform MCL in the environment of RoboCup Sony legged robot league.

Uniform Monte Carlo localization - fast and robust self-localization method for mobile robots

Self localization is necessary for mobile robot navigation. The conventional method such as geometric reconstruction from landmark observations is generally time-consuming and prone to errors. This paper proposes a method which constructs a decision tree and prediction trees of the landmark appearance that enable a mobile robot with a limited visual angle to observe efficiently and make decisions without global positioning in the environment. By constructing these trees based on information criterion, the robot can accomplish the given task efficiently. The validity of the method is shown with a four legged robot.

/pdf/observation-strategy-for-decision-making-based-on-1iy6ue55xp.pdf

Observation strategy for decision making based on information criterion

Designing the decision-making engine of a team of robots is a challenging task, not only due to the complexity of the environment where the robots usually perform their task, which include uncertainty, dynamism and imprecision, but also because the coordination of the team must be included in this design, The robots must be aware of other robots' actions to cooperate and to successfully achieve their common goal. Besides, decisions must be made in real-time and with limited computational resources.
This thesis contributes a novel case-based approach for action selection and coordination in joint multi-robot tasks in real environments. This approach has been applied and evaluated in the representative domain of robot soccer, although the ideas presented are applicable to domains such as disaster rescue operations, exploration of unknown environments and underwater surveillance, among others.
The retrieval process proposes a case to reuse, evaluating the candidate cases through different measures to overcome the real world characteristics, including the adversarial component which is a key ingredient in the robot soccer domain. Unlike classical case-based reasoning engines, the case reuse consists in the execution of a set of actions through a team of robots. Therefore, from the multi-robot perspective, the system has to include a mechanism for deciding who does what and how. In this thesis, we propose a multi-robot architecture along with a coordination mechanism to address these issues.
We have validated the approach experimentally both in a simulated environment and with real robots. The results showed that our approach achieves the expected goals of the thesis, i.e. designing the behavior of a cooperative team of robots. Moreover, the experimentation also showed the advantages of using collaborative strategies in contrast to individualistic ones, where the adversarial component plays an important role.

/pdf/action-selection-in-cooperative-robot-soccer-using-case-7fmlwzyirj.pdf

Action selection in cooperative robot soccer using case-based reasoning

In this article we present a case-based approach for the selflocalization of autonomous robots based on local visual information of landmarks. The goal is to determine the position and the orientation of the robot sufficiently enough, despite some strongly incorrect visual information. Our approach to solve this problem makes use of case-based reasoning methods.

/pdf/fault-tolerant-self-localization-by-case-based-reasoning-2gyjzw1030.pdf

Steffen Brüggert

Papers

Fault-Tolerant Self Localization by Case-Based Reasoning