Konrad Banachowicz

Opening a door with a redundant impedance controlled robot

Opening a door with a robotic manipulator equipped with a gripper, is a vital problem in service robotics research. This paper presents an approach to opening a cabinet door with an impedance controlled 7-DOF redundant manipulator. A control system is developed that considers redundancy in order to maximise the manipulation capabilities of the manipulator whilst avoiding reaching the joints limits. The control strategy is based on automatically estimating the kinematic parameters of the door. All conducted experiments were based on the assumption that the manipulator pulls a handle and opens a door using a single finger.

Research Oriented Motor Controllers for Robotic Applications

Motor controllers are vital parts of robotic manipulators as well as their grippers. Typical, commercial motor controllers available on the market are developed to work with high level robot industrial controllers, hence their adaptation to work as a part of a scientific, experimental robotic system is problematic. The general concept of research oriented motor controllers for robotic systems is presented in this article as well as an exemplary gripper and manipulator application based on this concept.

Multi-sensory Feedback Control in Door Approaching and Opening

In the article the robotic system behavior is investigated for the complex door opening task. The system consists of the 7-DOF KUKA LWR4+ manipulator, which is controlled in an impedance way and the BarrettHand gripper, which is controlled in a position way. The system utilizes multi-sensory feedback. The visual feedback is used to roughly localize door and to plan a door approach trajectory. The tactile feedback detects the contact with the door, and handle and determines an exact contact position with the handle. The system does not form a grip in a door opening stage, but the contact between the robot and the door is maintained by the gripper’s fingers (with intrinsic backlash), which are pushing the handle from its one side. This concept allows to open the door when there are obstacles in the neighborhood of the handle (e.g. door jamb or frame), which make the grip impossible.

Grasp planning taking into account the external wrenches acting on the grasped object

The paper presents an outline of the specification of a robot controller used in manipulation tasks requiring object grasping by a multi-fingered gripper and interaction with the environment. The specification assumes that the robot is represented as an embodied agent composed of real and virtual effectors and receptors and the control subsystem. The actions of those subsystems are defined by finite state machines invoking in each of their states appropriate behaviours. The presentation focuses on grasp planning and execution.

Two Mode Impedance Control of Velma Service Robot Redundant Arm

The previous research on reactive torque control of redundant arms led to conclusion, that initial arm kinematic configuration is vital for task executed with the use of Cartesian impedance control. To provide that, in the article the control system is proposed with the two following modes of impedance control of redundant manipulators: Joint space and Cartesian space. For this purpose the system was treated as embodied agent with two behaviors of its Virtual Effector (hardware abstraction layer). Each behavior has been decomposed to several components described by automatons and communicate asynchronously with upper layers of the control system. The whole system has been finally verified on real manipulator.

Automated generation of component system for the calibration of the service robot kinematic parameters

In the article we propose a fast and cheap strategy of service robot kinematic parameters calibration. Our method is based on sensors (cameras in particular) that are already mounted on the robot and inexpensive markers, that are easy to fix on the robot arms. We developed the method to compute the impact of manipulator velocity on markers localization error in camera image. Thanks to our method, the calibration data acquisition can be significantly shortened, because a certain, acceptable marker detection error threshold can be introduced that allows to acquire the data with non-zero velocity of the manipulator. Additionally, we propose a method of automatic component based data acquisition system generation, based on the embodied agent theory and tree like kinematic model representation of the robot. Our model is suitable for the most of service robots. The same model forms the base for an automatic generation of measure functions for cost functions used in the optimization process to find the optimal set of model parameters. The whole approach has been verified experimentally using Velma service robot.

Control system design procedure of a mobile robot with various modes of locomotion

In this article the design procedure of a control system of a mobile robot with various modes of locomotion has been considered. The procedure is based on an embodied agent theory and focuses on the agent behaviours and transitions between them. The procedure was introduced to speed up and organize the whole controller development process. The effectiveness of the proposed approach is proved through control system development and experimental verification of the custom-built Lynx mobile robot with vertical and horizontal modes of locomotion.

Graph-based potential field for the end-effector control within the torque-based task hierarchy

In the article, a reactive torque-based hierarchical task-space control for a robot with multiple redundant DOFs is presented that is suitable to plan, validate and execute a humanoid upper-body motion in human oriented environment. The strategy consists of: (i) joint limit avoidance, (ii) self-collision avoidance using repulsive force fields with additional environment collision avoidance, (iii) end-effector task with a specified discrete potential field with tricubic interpolation and (iv) a posture task on the lowest priority. The proposed approach is verified for Velma - a robot with a humanoid upper-body.

Multibehavioral position-force manipulator controller

In the article the whole process of a manipulator controller development is presented. The controller is designed for the execution of various benchmark tasks in the field of service robotics. Initially, a set of system behaviors is presented as the base of the universal controller structure. For that purpose the embodied agent theory and the System Modeling Language are utilized. The execution and switching between the behaviors relies on a particular algorithm. The control system is verified on a two arm robotic system that consists of position-controlled industrial manipulators and a number of extra force/torque sensors and cameras.

Distributed NAO robot navigation system in the hazard detection application

Advanced robot control algorithms, that are based on sensor data fusion and a lot of processing, require high computational power. Hence, most of the modern robots have their own highly efficient computers. It makes the service and assistant robots costly and still unavailable for the most of the home users. In this paper the navigation system of the humanoid NAO robot is presented that is distributed through the robot and the cloud. It reduces the cost of the particular robot and makes it affordable for customers. Furthermore, the multi-agent task specification and implementation method is proposed and illustrated on an example of hazard detection task - the assistant robot application to help the elderly people to dwell at home.