This paper presents an interval propagation algorithm for variables in single-loop linkages. Given allowed intervals of values for all variables, the algorithm provides, for every variable, the exact interval of values for which the linkage can actually be assembled. We show further how this algorithm can be integrated in a branch-and bound search scheme, in order to solve the position analysis of general multi-loop linkages. Experimental results are included, comparing the method’s performance with that of previous techniques given for the same task.

In this paper, we confront the problem of applying reinforcement learning to agents that perceive the environment through many sensors and that can perform parallel actions using many actuators as is the case in complex autonomous robots. We argue that reinforcement learning can only be successfully applied to this case if strong assumptions are made on the characteristics of the environment in which the learning is performed, so that the relevant sensor readings and motor commands can be readily identified. The introduction of such assumptions leads to strongly-biased learning systems that can eventually lose the generality of traditional reinforcement-learning algorithms. In this line, we observe that, in realistic situations, the reward received by the robot depends only on a reduced subset of all the executed actions and that only a reduced subset of the sensor inputs (possibly different in each situation and for each action) are relevant to predict the reward. We formalize this property in the so called categorizability assumption and we present an algorithm that takes advantage of the categorizability of the environment, allowing a decrease in the learning time with respect to existing reinforcement-learning algorithms. Results of the application of the algorithm to a couple of simulated realistic-] robotic problems (landmark-based navigation and the six-legged robot gait generation) are reported to validate our approach and to compare it to existing flat and generalization-based reinforcement-learning approaches.

Some robotic tasks require an accurate control to follow the desired trajectory in the presence of unforeseen external disturbances and system parameters variations. In this case conventional control techniques such as PID must be constantly readjusted and a compromise solution must be adopted. This problem can be avoided using a learning process that automatically learns the appropriate control law and adapts to ongoing system variations. But a drawback of many learning systems is that they are not effective for non-toy problems. In this paper we present the results obtained with a categorization and learning algorithm able to perform efficient generalization of the observed situations, and learn accurate control policies in a short time without any previous knowledge of the plant.

We present a reactive controller that is able to displace a legged robot along an arbitrary trajectory with a high degree of accuracy. We designed the different modules of our controller so that they can deal with arbitrary leg configurations. In this way, any leg movement necessary to overcome unexpected terrain irregularities can be correctly compensated by the controller, while still following the trajectory commanded by the user. Since we move the robot as a reaction to leg movements while stepping, the speed of the robot is automatically adjusted to the terrain profile; the more obstacles in the terrain, the more leg movements necessary to overcome them, and the slower the movement of the robot. We prove that, as the terrain becomes simpler, so does the gait generated by our controller, automatically converging to the tripod gait when the terrain becomes flat. This is achieved without requiring a map of the terrain and, thus, our controller can be used by robots with minimum computational and sensing capabilities. The results we report using different legged robots and in different environments prove the adequacy of our approach.

This paper presents a new approach to detect salient regions in an image. A number of units is used to respond to the most salient regions, adapting their response to the size and range of colors present in each region. The system can be directly used on a sequence of images, continuously adapting its output with time. A quality estimation of each unit allows selecting the most relevant regions present in the image at any time. The output of this process can be used as a visual attention mechanism to determine what parts of the image deserve to be processed further.

This paper presents a new approach to detect salient regions in an image. A number of units is used to respond to the most salient regions, adapting their response to the size and range of colors present in each region. The system can be directly used on a sequence of images, continuously adapting its output with time. A quality estimation of each unit allows to select the most relevant regions present in the image at any time. Experiments performed on test images showing a robust behavior of the system are presented. The output of this process can be used in a visual landmark-based navigation system to determine what parts of the image should be explored to find the most useful landmarks.

Visual navigation in unstructured, previously unknown, environments is investigated in the project ARGOS. This paper describes the goals of the project, the guidelines we follow to face each aspect of it, and the motivations that led us to work with a walking robot. The current state of the project in the areas of legged locomotion, landmark-based navigation and vision is succintly described.

Reinforcement learning constitutes a promising approach to achieve the adaptation of an animat's behavior to its environment. However, existing algorithms tend to be too time and space consuming, and become useless when the number of sensors and actions reach the typical sizes of a realistic animat situation. Existing generalization techniques mitigate somehow the problem, but they are still insufficient. In this paper, we suggest that the efficiency of learning observed in animals is possible, in part, because they capture important regularities, existing in the dynamics of its interaction with the environment, that are not exploited enough by the generalization algorithms. We characterize a kind of regularity that is plausible to hold for many animat environments, that we call categorizability, and propose an algorithm that takes advantage of such regularity. The main features of the algorithm axe: a competitive evaluation of actions through different partial views of the same situation, and the on-line generation of new partial views to improve the accuracy of action evaluation. The described algorithm is applied to the task of learning to walk with a simulated six-legged robot.

In this paper, we address the problem of moving a legged robot in an unknown environment according to externally provided driving commands assuming an arbitrary initial position of legs. This general problem is decomposed in three sub-problems: first, decide when to lift a given leg, second, choose where to move lifted legs, and third, coordinate body and leg movements to make the robot advance in the desired direction. To solve this third problem, we extend the posture control mechanism introduced in (1) so that body movements are limited to the desired trajectory. The resulting controller performs smooth transitions between different trajectories and can cope with irregular terrain where valid footholds are scarce.

The categorization process defines sensor and action categories from elementary sensor readings and basic actions so that the necessary elements for solving a task are correctly perceived and manipulated. In reinforcement learning, a previous categorization process is needed to define sensor and action categories with special requirements that we analyze and that, in general, are difficult to achieve, especially in complex tasks such as those that arise when working with autonomous robots. We show how these special requirements should be relaxed and we sketch a reinforcement learning algorithm that uses a less restrictive form of sensory categorization than existing algorithms. Additionally, we show how a given sensory categorization can be improved so that it better fits the demands of the previous algorithm.

We present a general method for the analysis of the gaits used by a six-legged robot independently of the mechanism used to generate the gait. The gait state of the robot is de ned as a function of the last executed steps and several classes of gait states as well as the transitions between them are identi ed. As an example, we apply our method to the well-know wave gaits (the most e cient and stable gaits for straight line locomotion on at terrain) showing interesting properties about how they can be generated, the relation between all the possible wave gaits and about the emergence of the tripod gait under certain assumptions likely to hold when walking on at terrain.

The control of a legged robot walking on difficult terrain demands the development of efficient and reliable algorithms to coordinate the movement of multiple legs according to a diversity of requirements. We present a control structure, implemented on a six-legged robot, in which the aspects of stability, mobility, ground accommodation, gait generation, and robot heading are integrated in a coherent and simple way.

Legged robots are well suited to walk on difficult terrains at the expense of requiring complex control systems to walk even on flat surfaces. But simply walking on a flat surface is not worth using a legged robot. It should be assumed that walking on abrupt terrain is the typical situation for a legged robot. With this premise in mind, we have developed a robust controller for a six-legged robot that allows it to walk over difficult terrains in an autonomous way, with a limited use of sensory information (no vision is involved). This walk controller can be driven by an upper level which need not be concerned about the details of foot placement or leg movements, taking care only of high level aspects such as global speed and direction.

An algorithm to obtain the solution of underconstrained rotation equations in the form of sets of values that the variables can take is given. Furthermore, an interval propagation algorithm is presented which permits obtaining the subsets of the solution that are compatible with a given interval for one of the variables. The propagation technique provides a way to take into account non-intersection constraints in the solution. In addition, propagation can be used to solve systems of rotation equations. Finally, it is shown how the algorithms described can be used in the solution of certain spatial problems, including 6-bar mechanisms.