Robotics Research

One of the most challenging aspects of concurrent mapping and localization (CML) is the problem of data association. Because of uncertainty in the origins of sensor measurements, it is difficult to determine the correspondence between measured data and features of the scene or object being observed, while rejecting spurious measurements. This paper reviews several new approaches to data association and feature modeling for CML that share the common theme of combining information from multiple uncertain vantage points while rejecting spurious data. Our results include: (1) feature-based mapping from laser data using robust segmentation, (2) map-building with sonar data using a novel application of the Hough transform for perception grouping, and (3) a new stochastic framework for making delayed decisions for combination of data from multiple uncertain vantage points. Experimental results are shown for CML using laser and sonar data from a B21 mobile robot.

This paper summarizes recent research on developing autonomous robot systems than can acquire volumetric 3D maps with mobile robots in real-time. The core of our system is a real-time version of the popular expectation algorithm, developed for extracting scalar surfaces from sets of range scans (Martin, Thrun, 2002). Maps generated by this algorithm consists of a small number of planar rectangular surfaces, which are augmented by fine-grained polygons for non-flat environmental features. Experimental results obtained in a corridor-type environment illustrate that compact and accurate maps can be acquired in real-time from range and camera data.

This work addresses the real time implementation of Simultaneous Localization and Mapping (SLAM). It presents the integration if the Compressed Extended Kalman Filter (CEKF) and a new decorrelation algorithm to reduce the computational and memory requirements of SLAM to ∼ O(N *N _a ), being N and N _a proportional to the total number of landmark in the global map and local area respectively. It also presents the problematic of outdoors navigation using natural feature based localization methods. The aspect of feature detection and validation is investigated to reliable detect the predominant features in the environment. Experimental results obtained in outdoor environments are presented.

This paper describes a full probabilistic solution to the Simultaneous Localisation and Mapping (SLAM) problem. Previously, the SLAM problem could only be solved in real time through the use of the Kalman Filter. This generally restricts the application of SLAM methods to domains with straight-forward (analytic) environment and sensor models. In this paper the Sum-of-Gaussian (SOG) method is used to approximate more general (arbitrary) probability distributions. This representation permits the generalizations made possible by particle filter or Monte-Carlo methods, while inheriting the real-time computational advantages of the Kalman filter. The method is demonstrated by its application to sub-sea field data consisting of both sonar and visual observation of near-field landmarks.

The capability of action imitation constitutes a fundamental basis of higher human intelligence. In this paper, we first analyze the concept of imitation and mark the essential problems underlying imitation, e.g. redundant sensory and motor degrees of freedom, adaptive mapping, and strong embodiment. It supports a synthetic approach to understanding imitation capabilities, which requires a versatile humanoid robot platform. An overview of our full-body humanoid robot system is presented, which is signified by its versatile physical capabilities and complete open architecture. Then we present our early experiment on multi-modal architecture and behavior imitation with the humanoid. It can continuously interact with humans through visual, auditory and motion modalities in an unmodified everyday environment. And when a person attends to the robot, starting to show a dual arm motion, the robot spontaneously starts to copy it.

This paper describes our research efforts aimed at developing several low-level autonomous capabilities required for remote-operation tasks involving humanoid-type robots. The low-level autonomy considered falls into three categories: 1) walking functions, such as online dynamic balance compensation and online walking trajectory generation, 2) manipulation functions such as arm motion planning and 3D vision based interactive planning, and 3) human interaction functions such as human identification and tracking, face recognition, and voice recognition/speech synthesis. We describe experimental results implemented on the humanoid robot research platforms H6 and H7.

This paper introduces an open architecture humanoid robotics platform (OpenHRP for short) on which various building blocks of humanoid robotics can be investigated. OpenHRP is a virtual humanoid robot platform with a compatible humanoid robot, and consists of a simulator of humanoid robots and motion control library for them which can also be applied to a compatible humanoid robot as it is. OpenHRP also has a view simulator of humanoid robots on which humanoid robot vision can be studied. The consistency between the simulator and the robot are enhanced by introducing a new algorithm to simulate repulsive force and torque between contacting objects. OpenHRP is expected to initiate the exploration of humanoid robotics on an open architecture software and hardware, thanks to the unification of the controllers and the examined consistency between the simulator and a real humanoid robot.

Human can perform variety of limber whole-body motions using numerous muscles and huge number of various sensors. The human brain has all the connections to the sensors and muscles, and learn how to manage them for whole-body motions. In this research, we have aimed to build a complex body with physically massive parallel sensor-motor systems to enter the next stage for studying humanoid brain systems. It is designed to have a flexible spined torso and a whole-body with fully muscle-tendon driven systems. In this paper the design and implementation of the first model of the humanoid is described with some experiments.

We demonstrate an adaptation strategy for adjusting the stride period in a hexapedal running robot. The robot is inspired by discoveries about the self-stabilizing properties of insects and uses a sprawled posture, a bouncing alternating-tripod gait, and passive compliance and damping in the limbs to achieve fast, stable locomotion. The robot is controlled by an open-loop clock cycle that activates the legs at fixed intervals. For maximum speed and efficiency, this imposed stride period should be adjusted to match changes in terrain or loading conditions. An ideal adaptation strategy will complement the design philosophy behind the robot and take advantage of the self-stabilizing role of the mechanical system. In this paper we describe an adaptation scheme based on measurements of ground contact timing obtained from binary sensors on the robot’s feet. We discuss the motivation for the approach, putting it in the context of previous research on the dynamic properties of running machines and bouncing multi-legged animals, and show results of experiments.

We have been trying to induce a quadruped robot to walk dynamically on irregular terrain by using a neural system model consisting of a CPG (central pattern generator) and reflexes. In this paper, we define adaptive walking using a neural system model as “coupled-dynamics-based motion generation”, in which a neural system and a mechanical system are coupled and generate motion by interacting with the environment emergently and adaptively. In order to clarify how a CPG is coupled to a mechanical system, we use simulations to investigate the relationship between the parameters of a CPG and the dynamics of the mechanical system. We propose the essential conditions for stable dynamic walking on irregular terrain in general, and we design a neural system by comparing biological knowledge with the essential conditions described in physical terms. We report our experimental results of dynamic walking on terrains of medium degrees of irregularity using a planar quadruped robot and a newly developed three-dimensional quadruped robot. MPEG footage of these experiments can be seen at: http://www.kimura.is.uec.ac.jp.

An optimal kinematic design method suited for parallel manipulators is described. The optimal configuration for a Delta-type three-degree-of-freedom spatial, translational manipulator, known as the New University of Western Australia Robot (NUWAR) is presented, and shown to be advantageous over the Delta configuration in terms of workspace volume. These results led to the construction of a prototype and an Australian Patent application.The kinematic optimisation process yielding a design, which delivers the best compromise between manipulability and a new performance index: space utilisation, is presented. The process leading to finding an optimal configuration of Linear Delta robot is described.An example medical application of a parallel robot is discussed. The robot is designed to work inside an open magnetic resonance scanner. A concept of a robot control system, based on biomechanical models of organs the robot operates on, is presented.

We have developed a humanoid robot called “Robovie”. The task is to communicate with humans and establish relationships by using various sensors and actuators. For designing the robot behavior, we have performed cognitive experiments, implemented the results on the software architecture, and verified the effectiveness in human-robot communication. This paper proposes an interdisciplinary approach between cognitive science and robotics for developing the communicative robot.

In this paper a brief overview on the German Collaborative Research Center on Humanoid Robots is given. The research topics focused by the collaborative research center are multimodal man-machine interaction, learning of motion, skills and tasks, man-robot cooperation, perception and action and mechatronics and controls for a human like torus consisting of 2 redundant arms, 2 dexterous 5-finger hands, a head with eyes and ears, a neck and a spine. The project started in July 2001 and is planned in long terms for a period of up to 12 years. Over this period it is envisaged to elaborate and establish basic methodologies and technology for humanoid robots to act and cooperate close with humans in daily living environment. An integrated perspective towards continuous humanoid inter-action gaining multimodality, adaptivity, redundancy and flexibility is proposed.Programming, cooperation and interaction with a humanoid robot is assumed to be multimodal which means to let the user program the robot simply by speech, gesture or demonstrating a task. The robot observes, interprets and then tries to imitate and to learn the performed user action. On the basis of a redundant humanoid type 2 arm robot system equipped with an active stereo head and a microphone array observation of the demonstrated task is realized as a first prototype using an active vision system. Grasping of objects is detected with the help of data gloves and active vision. The system interprets and stores the observed actions, segments them into meaningful sequences in a given context. Due to sensor errors and the complexity of the intended interaction, the system generates queries concerning intention, manipulation and objects. Motion catch techniques in combination with learning capabilities are combined towards imitation learning and active observational learning strategies.

This paper describes the overall control structure of a new generation of compliance controlled manipulators implemented by the design and control of the DLR lightweight robot. To achieve the compliance property, the manipulator does not have to revert to a human-like, bionic design. A serial link manipulator, control technology, and sensorized joint actuators can generate a kinesthesis similar to the compliance properties of the human arm. The design and control of the new robot generation implement these properties in a very different way than its biological counterpart to enable high fidelity interaction of the robot with its environment.

This work develops a methodology for building haptic reality-based modeling systems by exploiting the complementary goals of haptic exploration and display. While the generally unstructured nature of haptic exploration makes it difficult to develop and control autonomous robotic fingers, simultaneous consideration of exploration and display provides a way to characterize the required sensing, control, finger geometry, and haptic modeling components. Both the exploring robot and haptic display must complement an appropriate virtual model, which can accurately recreate remote or hazardous environments. We present an example of haptic exploration, modeling and display for surface features explored using a three-degree-of-freedom spherical robotic fingertip with a tactile sensor.

A new field of robotics is emerging. Robots are today moving towards applications beyond the structured environment of a manufacturing plant. They are making their way into the everyday world that people inhabit. The paper focuses on models, strategies, and algorithms associated with the autonomous behaviors needed for robots to work, assist, and cooperate with humans. In addition to the new capabilities they bring to the physical robot, these models and algorithms and more generally the body of developments in robotics is having a significant impact on the virtual world. Haptic interaction with an accurate dynamic simulation provides unique insights into the real-world behaviors of physical systems. The potential applications of this emerging technology include virtual prototyping, animation, surgery, robotics, cooperative design, and education among many others. Haptics is one area where the computational requirement associated with the resolution in real-time of the dynamics and contact forces of the virtual environment is particularly challenging. The paper describes various methodologies and algorithms that address the computational challenges associated with interactive simulations involving multiple contacts and impacts between human-like structures.

Teleoperation can be improved if humans and robots work as partners, exchanging information and assisting one another to achieve common goals. In this paper, we discuss the importance of collaboration and dialogue in human-robot systems. We then present collaborative control, a system model in which human and robot collaborate, and describe its use in vehicle teleoperation.

In the Expectation-based, Multi-focal, Saccadic (EMS-) Vision system of UBM, the vision sensor consists of four cameras with different focal lengths mounted on a highly dynamic pan-tilt camera head. Image processing, gaze control and behavior decision interact with each other in a closed loop form. The behavior decision module specifies the relevance of obstacles like road segments, crossings or landmarks in the situation context. The gaze control unit takes all this information in order to plan, optimize and perform a sequence of smooth pursuits, interrupted by saccades. The sequence with the best information gain is performed. The information gain depends on the relevance of objects or object parts, the duration of smooth pursuit maneuvers, the quality of perception and the number of saccades. The functioning of the EMS-Vision system is demonstrated in a complex and scalable autonomous mission with the UBM test vehicle VAMORS.

This paper presents our experience towards the conception of a virtual reality medical simulator coupled with haptic interaction aimed at training surgeons. This area of research has a long history and a wide variety of approaches have been used. Generally, human tissue can be considered as a deformable body of viscoelastic material. To enable dynamic simulation of these bodies, we have patched three well known physical models onto their geometrical model: mass-spring networks which is more of a discrete object model, finite element method (FEM) based on continuum mechanics and recently long element method (LEM) which we believe to be more promising. We make some comparisons between these models. We also present some numerical resolution method for simulation of deformable bodies. As far as real-time interactions are concerned, we present our work on collision detection, haptic interaction and topology modifications. In the haptic system, we separate the physical simulation and the haptic interaction to ensure stability; the link between the two process is acheived by means of a local model which will be eloborated. We present some experimental results to highlight these works.

In order for a mobile robot to provide information services in real offices, the robot has to maintain the map of the office. Rather than a completely autonomous approach, we chose to interact with office people to learn and update the topological map using spoken dialogue. To successfully apply a speech recognition technology to conversation understandings in real offices, we implemented a multiple microphone array system and a context and attentional manager in the robot. The robot could demonstrate simple map learning, route guidance, and information service about people’s location.

In this paper the use of robots as intelligent appliances is discussed. A number of advertised systems are reviewed and their basic characteristics analyzed. Open issues in terms of navigation, user interfaces, robustness, price are discussed as a basis for issues for future research to enable commercial delivery of such systems.

Field Robots are machines that work in unstructured environments, including un- der water, in mines, in forests and on farms, and in the air. These applications involve both advanced ideas in robotics and careful attention to engineering details. This paper discusses the overall challenges of field robotics and the current state of the art. The paper is not a thor- ough survey, nor a tutorial; but is instead intended to serve as a discussion starter for further design and development.

We discuss the role of spatial representations and visual geometries in vision-based navigation. To a large extent, these choices determine the complexity and robustness of a given navigation strategy. For instance, navigation systems relying on a geometric representation of the environment, use most of the available computational resources for localization rather than for “progressing” towards the final destination. In most cases, however, the localization requirements can be alleviated and different (e.g. topological) representations used. In addition, these representations should be adapted to the robot’s perceptual capabilities.Another aspect that strongly influences the success/complexity of a navigation system is the geometry of the visual system itself. Biological vision systems display alternative ocular geometries that proved successful in different (and yet demanding and challenging) navigation tasks. The compound eyes of insects or the human foveated retina are clear examples. Similarly, the choice of the particular geometry of the vision system and image sampling scheme, are important design options when building a navigation system.We provide a number of examples in vision based navigation, where special spatial representations and visual geometries have been taken in consideration, resulting in added simplicity and robustness of the resulting system.

This paper demonstrates some interesting connections between the hitherto disparate fields of mobile robot navigation and image-based visual servoing. A planar formulation of the well-known image-based visual servoing method leads to a bearing-only navigation system that requires no explicit localization and directly yields desired velocity. The well known benefits of image-based visual servoing such as robustness apply also to the planar case. Simulation results are presented.

Insects, being perhaps more reliant on image motion cues than mammals or higher vertebrates, are proving to be an excellent organism in which to investigate how information on optic flow is exploited to guide locomotion and navigation. This paper describes one example, illustrating how bees perform grazing landings on a flat surface. A smooth landing is achieved by a surprisingly simple and elegant strategy: image velocity is held constant as the surface is approached, thus automatically ensuring that flight speed is close to zero at touchdown. No explicit knowledge of flight speed or height above the ground is necessary. The feasibility of this landing strategy is tested by implementation in a robotic gantry. We also outline our current efforts at exploring the applicability of this and related techniques to the guidance of UAVs.

A simple construction is given for measuring angles in a plane from a perspective image. The method requires the identification of the horizon line of the plane in the image, as well as a point called the conformal point. The angle between two lines is measured by extending the two image lines until they meet the horizon, then connecting back to the conformal point. The angle between the resulting lines is the angle between the original lines in the plane.The position of the conformal point is computed by calibrating the camera. Alternatively, knowledge of angles in the plane allows the conformal point to be located without explicit calibration. In pure planar motion, where the horizon and the conformal point of the ground plane are preserved, applications such as motion estimation can be applied for robot navigation and measurement of road angles.

This paper describes a new probabilistic roadmap (PRM) path planner that is: (1) single-query — instead of pre-computing a roadmap covering the entire free space, it uses the two input query configurations as seeds to explore as little space as possible; (2) bidirectional — it explores the robot’s free space by concurrently building a roadmap made of two trees rooted at the query configurations; (3) adaptive — it makes longer steps in opened areas of the free space and shorter steps in cluttered areas; and (4) lazy in checking collision — it delays collision tests along the edges of the roadmap until they are absolutely needed. Experimental results show that this combination of techniques drastically reduces planning times, making it possible to handle difficult problems, including multi-robot problems in geometrically complex environments.

Recent research results suggested a conservative transformation to correct the well-known congruence transformation between Cartesian and joint stiffness matrices of a serial manipulator. This paper utilizes screw geometry to interpret the conservative congruence transformation (CCT). The analysis using screw theory provides better geometric insights into the CCT. The effective geometrical stiffness matrix, due to the change of manipulator geometry under stiffness control in the presence of external force, is confirmed. This paper also points out several erroneous assumptions that may have led to the incorrect formulation of the conventional congruence transformation.

This paper presents an analysis of frictionless contact between a rigid body belonging to a robot mechanism and one belonging to its environment. According to this analysis, it is possible to design a hybrid motion/force controller such that the motion and force subsystems are instantaneously independent of each other, and both are instantaneously independent of the environmental dynamics. A control system with this property should be able to operate in contact with an environment having unknown dynamics.

Much work by many researchers and developers (Schempf, 1999) (Menzel, D’Alusio, 2001) (Schraft, Schmierer, 1998) over the last century has developed a vast body of knowledge in the areas of locomotion utilizing wheels, legs, tracks, etc. for applications from research to real-world applications. This paper addresses a novel development of a steerable monotread, dubbed AURORAAdvanced Urban RObot for Reconnaissance and Assessment), proving that a single continuous belt, designed with key flexure and guide elements, is capable of steerable locomotion. This is believed to be a significant departure from the theory that tracked vehicles need to have at least two treads to steer. The system was built with a flexible elastomeric monobelt with a central drive and guide spine, which, when flexed, forces the tread into a shape allowing it to steer. The system is also capable of inverted operations, stair climbing (with the help of a deployable ramp/paddle), and is easily portable due to its small size and low weight. The system is battery-driven and controlled/monitored over a wireless link, allowing it to be deployed safely into hazardous and remote areas in urban terrain. On-board cameras provide multiple side- and bird’s-eye views, with on-board computing processing and interpreting imagery. A portable control-box is used for remote control. Preliminary tests have shown the capability of the system to handle rough terrain and steer in all of the environments tested so far. Future work will extend the autonomy capabilities of the system and ruggedize the tread and drive/steer elements even further.

Various types of rigid and flexible endoscopes are used to inspect and to perform therapeutic procedures on different parts of the gastrointestinal (GI) tract. Due to the working characteristics of conventional endoscopes, most GI endoscopy procedures are unpleasant for the patient, and are technically demanding for the endoscopist. The authors are developing minirobots for semi-autonomous or autonomous locomotion in the GI tract. In this paper, the authors illustrate the systematic approach to the problem of “effective” locomotion in the GI tract and the critical analysis of “inchworm” locomotion devices, based on extensor and clamper mechanisms. The fundamentals of locomotion and the practical problems encountered during the development and the testing (in vitro and in vivo) of these devices are discussed. Finally, two mini-robots capable of propelling themselves in the colon and potentially suitable to perform rectum-sigmoidoscopy and colonoscopy are presented.

Robotics research is heavily dependent on fast, accurate, reliable and cheap sensors. Sonar sensing can fulfil these requirements in air and underwater environments. Moreover sonar physics provides robotics researchers with a natural selection capability for landmark detection in navigation problems. This paper presents new sonar results that allow highspeed accurate measurement and classification suitable for moving platforms that has been combined with interference rejection to allow multiple sonar systems to co-exist.

We have been developing a visual/haptic interface to virtual environments, called a WYSIWYF Display, which ensures spatially/temporally consistent visual/haptic sensation. Three key components are (i) vision-based tracking for correct visual/haptic registration (ii) blending live video and computer graphics (CG) images by a chroma-keying technique, and (iii) encountered-type haptic rendering that provides real free and real touch sensations.In this paper, the recent progress on the above key components are presented. First, accurate image overlay technique by using vision and accelerometers is shown. Then, a path planning for encountered-type haptic devices that render multiple virtual objects is presented. A virtual control panel of automobiles was developed to demonstrate the proposed path-planning algorithm.The developed visual/haptic interface system can be regarded as a kind of mechano-media, a new concept extending the framework of teleoperation, which transmit kinesthetic knowledge from a human to other humans.

This paper traces four years of evolution of the UNSW team in the RoboCup Sony legged robot league. The lessons learned in the creation of a competitive team are instructive for a wide range of applications of robotics. We describe the development of vision and localisation procedures for robot soccer, as well as innovations in locomotion and the design of game play strategies. Since all teams in the competition are required to use identical hardware, the key factor to success in this league is the creativity of the software designers in programming the robots to perform skills that the robots were not originally intended to do and to perform them in a highly dynamic and non-deterministic environment.

The paper presents application of a parallel robot to a new class of hybrid motion simulation in which most of simulation is done using a complex numerical model and only a small but indispensable part of simulation is done using a physical model. The advantage of this type of simulation is that it exploits rapidly growing computational means for testing a real object in the simulation. A key element in such a class of simulation is a motion table that realizes motion with high frequency and combines the results of the numerical simulation precisely with the motion of the physical model or vice versa. The paper presents application of a fast parallel robot to such a high-speed motion table. It also presents modeling of the impact dynamics of motion in the simulator and its verification through experiments.

Although odometry is nonlinear, it yields sufficiently to linearized analysis to produce a closed-form transition matrix and a symbolic general solution for both deterministic and stochastic error propagation. Accordingly, error propagation in vehicle odometry can be understood at a level of theoretical rigor equivalent to the well-known Schuler dynamics of inertial navigation. While response to initial conditions is path-independent, response to input errors can be related to path functionals. These trajectory moments are integral transforms which functions like the moment of inertia or the Laplace transform — enabling many error propagation calculations to be performed by hand in closed-form.

Most of current machine vision systems suffer from a lack of flexibility to account for the high variability of unstructured environments. Here, as the state of the world evolves the information provided by different visual attributes changes, breaking the initial assumptions of the vision system. This paper describes a new approach for the creation of an adaptive visual system able to selectively combine information from different visual dimensions. Using a probabilistic approach and uncertainty metrics, the system is able to take appropriate decisions about the more relevant visual attributes to consider. The system is based on an intelligent agent paradigm. Each visual algorithm is implemented as an agent, which adapts its behavior according to uncertainty considerations. The proposed system aims to achieve robustness and efficiency. By combining the outputs of multiple vision modules the assumptions and constraints of each module are factored out resulting in a more robust system. Efficiency is achieved through the online selection and specialization of the agents. An implementation of the system for the case of human tracking showed encouraging results.