Narrow Results

This paper presents the design and implementation of a human–robot interface capable of evaluating robot localization performance and maintaining full control of robot behaviors in the RoboCup domain. The system consists of legged robots, behavior modules, an overhead visual tracking system, and a graphic user interface. A human–robot communication framework is designed for executing cooperative and competitive processing tasks between users and robots by using object oriented and modularized software architecture, operability, and functionality. Some experimental results are presented to show the performance of the proposed system based on simulated and real-time information.

In robotic demining, the robot relies on a path-planner capable of generating trajectories to search for all the mines while avoiding obstacles whose locations are unknown. Several families of coverage algorithms exist but there is only one that guarantees complete coverage, the exact cellular decomposition family. This paper details the modifications performed to a cellular decomposition method for unstructured environments for its application to walking robots. Experiments show preliminary results and improvements to the method are proposed.

This paper introduces a special actuator that is characterized by presenting both a dual nonlinear transmission ratio and resonance properties. The peculiar features of this actuator make it suitable for numerous applications, but its application in walking robots is one of the most interesting cases. Results of design and simulations are presented. It is demonstrated that the use of this drive allows a considerable increasing of robot's effectiveness.

This paper discusses the problem of multisensor perception for autonomous stair climbing. The perception system is mounted on the Messor six-legged walking robot. The robot, due to its static stability while walking, is able to traverse obstacles in urban space, especially stairs. Messor while climbing stairs uses an adaptive algorithm, which exploits on-line perception of the stair geometry and robot pose with regard to the stair. The ascent procedure consist of three main parts. The first – preparation – measurements are performed in order to obtain information about the geometry of the stairs. The second – climbing – ascending each stair with correction of the robot orientation and horizontal position on the stairs. The third – landing – detection of the last stair and the end of the stair climbing procedure. The paper is focused on the multisensor system and the perception algorithms.

Modern walking robots are able to negotiate rough terrain. However there are still open topics, especially when there is a need to climb or descend an obstacle. This article presents the perception system for descending stairs with the sixlegged walking robot. The perception system for stair descent differs much from that used for stair climbing, while during the descent most of the surfaces of the step suffer from occlusions. Author describes the solution to this problem. The perception system consists of two sensors. Namely a video camera and a force sensor at the tip of the leg for active haptic sensing. The monocular vision system gives the scale of the stairs and the active haptic sensing system provides the additional information to obtain the real geometry of the obstacle. In the presented paper experiments were performed on the walking robot Messor. The article describes the monocular vision system and the method for obtaining the geometry of stairs. Next the ways of improving the accuracy of the system are presented.

For a legged robot that walks on rough terrain it is very important to avoid falling down while executing its mission. For this purpose, maintaining the six legged robot statically stable is considered an important issue and challenging task. This paper will shed light on the applied approach for detecting the direction of ground inclination, obstacle avoidance as well as the effect of amputation on statically stable up- and downhill walking based on evaluating the local current consumption and angular position of each leg’s joint. The body posture will be adjusted in relation to the direction of inclination. Our approach is based on an organic computing architecture and was tested on a low-cost version of the OSCAR robot.

This paper presents the design of a quadrupedal robot that can automatically adapt its gait to, and climb, staircases of different configurations. This is accomplished by endowing the robot with a parameterized gait for stair climbing: First, a gait plan is synthesized that allows the robot to climb a stair of known dimensions. Second, the robot approaches a previously unseen stair and perceives its height and width by using an onboard vision system. Third, the synthesized gait plan is parameterized by the perceived estimates of height and width of the stair. Fourth, the robot executes the parameterized gait to climb the staircase; this thereby eliminates the need for a complex control system to achieve the same purpose. Whereas quadruped robots have previously demonstrated stair climbing, to the best of our knowledge, none have so far been capable of climbing stairs of variable height while simultaneously carrying all the needed perception, processing, and power modules on-board. Our work is one of the first successful attempts toward the above goal. Results with the robot climbing a variety of stair configurations demonstrate the effectiveness of our approach.

Improving robot walking over compliant surfaces is considered an important issue. However, walking on an uneven surface requires controlling the robot‘s leg position to span a varying distance between the robot’s body and the ground. This paper will shed light on the applied decentralized controller approach that enables the robot‘s leg to adapt its position to be commensurate with the type of walking surface in order to improve walking on compliant surfaces. When the substrate compliance is high (sandy ground), the leg has to correct its position close to the point of contact with the ground. When the substrate compliance is low (gravel ground), the correction of the leg position is low. The novelty of our approach is the evaluation of the local current consumption and the angular position of each leg’s joint as somatosensory feedback. Our approach is based on an organic computing architecture and was tested on a low-cost version of the OSCAR walking robot.

This article focuses on the influence of walking speed and direction of the robot movement on the tactile ground classification. The perception system comprise force/torque sensor mounted on the foot of a six-legged robot. The force/torque signals are registered during the negotiation of several terrains. Next, based on the statistical or spectral analysis of the signal the robot is able to classify the terrain. In this paper we are concentrated on the influence of the walking speed and the direction of the robot movement on the classification performance. The results obtained proved that it is possible to learn the characteristics of the terrain using generalized classifier, which is trained on the dataset containing measurements acquired for the whole range of the selected gait parameter.