Giulio Milighetti

On a primitive skill-based supervisory robot control architecture

Smart interaction of humanoid robots in a complex public, private or industrial environment requires the introduction of primitive skill-based discrete-continuous supervisory control concepts. The functionality of the proposed hierarchical robot supervisory control architecture captures both the hierarchy that is required for representing complex skills as well as the mechanisms for detecting failures during their execution. At first by means of several complementary (e.g. internal, optical, tactile or acoustic) sensors and by neuro-fuzzy based fusion of relevant sensor features, the actual motion phase or fault event is continuously diagnosed. Depending on the identified motion phase or random fault event, the most appropriate discrete-continuous control strategy coping optimally with the corresponding situation will be selected and executed. First experimental and simulation results are reported in this paper

Adaptive predictive gaze control of a redundant humanoid robot head

A general concept for the gaze control of a redundant humanoid robot head is presented. It is based on an adaptive Kalman filter that predicts the next state of the moving target, processing the position information provided by a head-mounted stereo camera. The trajectory tracking control at the task level combines a proportional feedback and a feedforward term. The gains of both control actions are adapted in order to provide optimal dynamic response for unknown arbitrary target trajectories. Inverse differential kinematics is evaluated so that human-like joint motions are achieved. To exploit kinematic redundancy, a weighted pseudoinverse is realized that takes into account different optimization criteria. Additional self-motions of the head are also considered. Experimental results on the head of the humanoid robot ARMAR-III are presented.

On the Discrete-Continuous Control of Basic Skills for Humanoid Robots

Within the next years a new generation of humanoid robots is going to be developed being able to manage autonomously sophisticated tasks in a complex, time variant domestic environment. To cope with these advanced requirements a new multi-sensor based discrete-continuous supervisory control concept is proposed, which is able to accomplish even complex human skills. Each skill is divided into a sequence of primitive skills (PS). Depending on the multisensor perception of the robot state, the discrete control has to provide an optimal selection and activation of different PS. On the other hand the continuous control has to assure that each PS is performed by means of the most appropriate sub-controllers. Both theoretical approach and experimental results of the ongoing research are presented in this paper

Human-like reflexes for robotic manipulation using leaky integrate-and-fire neurons

In this paper we present an approach to transfer human-like reflex behavior to robots by utilizing leaky integrate-and-fire neurons. For the acceptance of robots in general and humanoid robots, which are even closer to peoples daily life, in particular a main aspect is their appearance and how they act and move in human centered environments. Especially safety strategies are crucial for a widespread acceptance of these machines. In our work we target this safety aspect by approaching this issue from the direction how humans respond to external stimuli. To achieve such human-like reflexes a general reflex unit, based on special variants of the leaky integrate-and-fire neuron model has been built. Instances of this reflex unit are adapted to special reflex types and connected to form dependent reflex behaviors. The concept of these neural structures and its evaluation by means of several experiments are presented in this paper. The results are depicted in detail and future aspects of our ongoing work are addressed.

Combined Visual-Acoustic Grasping for Humanoid Robots

This paper presents an efficient approach for the combined audio-visual localization of a fallen object. The localization algorithm is incorporated into a multi-sensor robotic platform to supervise and to adapt dynamically the discrete plan of a robot performing a pick-and-place task. Initial audio position estimates of the fallen object are delivered by means of time-delay estimates in two microphone pairs. For a successful operation even in acoustically adverse environments, reliability criteria for the time-delay estimates are introduced. These audio position estimates are improved by visual data using the disparity information in stereo images of a stereo-camera system. The performance of the audio-visual localization algorithm is evaluated for real data in a common office environment. The proposed real-time system shows robustness and accuracy localizing with ease a fallen wooden brick

On the robot based surface finishing of moving unknown parts by means of a new slip and force control concept

Fast and accurate robot based machining, finishing or cleaning of surfaces of geometrically complex manufacturing parts requires the introduction of flexible sensor based automation concepts. The industrial robots currently available on the market can not comply with such advanced requirements. In this paper a new concept for managing such challenging robot tasks is presented. It relies both on a flexible continuous slip and force control algorithm as well as on the use of a new slip sensor which is able to measure relative motion between the robot end-effector and the machined object surface. The smart slip and force control concept is applied to a robotic arm finishing the surface of an unfixed randomly moved object. Both theoretical approach and experimental results of ongoing research are presented in this paper.

Fuzzy based decision making for the discrete-continuous control of humanoid robots

Within the next years a new generation of humanoid robots able to manage autonomously sophisticated tasks in a complex, time varying domestic and public environment is going to be developed. To cope with these advanced requirements a new multi-sensor based discrete- continuous supervisory control concept is proposed, which is able to accomplish even complex human skills. Each skill is divided into a sequence of elementary actions (so called Primitive Skills). Depending on the multi-sensor perception of the current state of the system, the discrete control has to provide an optimal selection and activation of the appropriate sequence of action and control strategy. An on-line decision making algorithm based on the structure of Primitive Skills (PS) has been implemented. On a lower level the continuous control has to assure that each PS is performed by means of the most appropriate sub-controllers. The theoretical approach and first experimental results of the ongoing research are presented in this paper.

Vision Controlled Grasping by Means of an Intelligent Robot Hand

In this paper a new visual servoing concept for dynamic grasping for humanoid robots will be presented. It relies on a stereo camera in the robot head for wide range observing in combination with a miniaturized close range camera integrated in a five finger hand. By optimal fusion of both camera information using a fuzzy decision making algorithm a robust visually controlled grasping of objects is achieved even in the case of disturbed signals or dynamic obstacles.

Skill-based telemanipulation by means of intelligent robots

In order to enable robots to execute highly dynamic tasks in dangerous or remote environments, a semiautomatic teleoperation concept has been developed and will be presented in this paper. It relies on a modular software architecture, which allows intuitive control over the robot and compensates latency-based risks by using Augmented Reality techniques together with path prediction and collision avoidance to provide the remote user with visual feedback about the tasks and skills that will be executed. Based on this architecture different skills with high dynamics are integrated in the robot control, so that they can be executed autonomously without the delayed feedback of the user. The skill-based grasping by adherence of smooth or fragile objects during a remote controlled picking and placing task will be exemplary presented.

Flexible diskret-kontinuierliche Überwachung und Regelung humanoider Roboter

Zusammenfassung Die Regelung humanoider Roboter stellt nicht nur hohe Anforderungen an den Entwurf der kognitiven Architektur und der Hardwarekomponenten, sondern auch an die Planungs-, Regelungs- und Überwachungskomponenten, die die Lücke zwischen der Kognition und der Hardware schließen. Dieser Beitrag baut auf einer Übersicht über aktuelle Trends bei Industrie- und Servicerobotern auf und präsentiert ein generisches Lösungskonzept in Form eines modularen, multisensoriellen, diskret-kontinuierlichen Regelungs- und Überwachungskonzepts. Hierbei wird insbesondere die Zerlegung komplexer Aktionen in einzelne Grundgeschicklichkeiten mit integrierten Regelungs- und Überwachungslösungen beschrieben. Abstract Control of humanoid robots makes great demands not only on the development of cognitive functions and hardware components, but also on the design of planning, supervision and control methods, which establish the connecting element between cognition and hardware. Starting with an overview about current trends in the field of industrial and service robotics, we present in this contribution a general concept for a modular multisensory discrete-continuous supervisory control. We will analyze in particular the decomposition of complex robot actions in a sequence of primitive actions, each with integrated control and supervision solutions.