RRG Research Mobile Manipulator

Research

Mobile Manipulation

Motivation

A mobile manipulation system offers a dual advantage of mobility offered by the mobile platform (AKA Unmanned Graound Vehicles (UGVs)) and dexterity offered by the manipulator. The mobile platform offers unlimited workspace to the manipulator. The extra degrees of freedom of the mobile platform also provide user with more choices. However the operation of such a system is challenging because of the many degrees of freedom and the unstructured environment that it performs in. The current generation of the system needs to be teleoperated over a large workspace for doing manipulation tasks requiring skilled operators with significant amount of training.

Introduction

Robotics Research Group at University of Texas at Austin (RRG) and Idaho National Lab (INL) have been engaged in a collaborative research in the area of mobile manipulation for the past four years which has resulted in an improved interface for robotic operation and control technologies for arms and mobile robots individually as well as for mobile manipulator as a whole. RRG has developed a manipulation control software framework, namely OSCAR (Operational Software Components for Advanced Robotics), for highly-dexterous and field-configurable (in terms of geometry, dexterity, and end effector tools) robotic arms. OSCAR is a fully modular, open-architecture framework that can efficiently control robotic arms ranging from over constrained configurations to hyper redundant configurations and from a single arm to multi-arm systems. OSCAR has been demonstrated in challenging environments such as nuclear cleanup and surgical robotics. INL has developed modular control software RIK (Robot Intelligence Kernel) for mobile robots to navigate in highly unstructured environment with changing levels of autonomy. RIK is a portable, reconfigurable onboard architecture that integrates perception, world-modeling, adaptive communication, dynamic tasking, and behaviors for mapping, localization, obstacle avoidance, waypoint navigation, search and detection.

The overall goal of this research collaboration is to develop a unified interface and control architecture to support both the navigation and manipulation so that the system can function effectively as a human surrogate in critical and hazardous environments such as power plants, industrial settings, and defense domains. The system is designed to run at varying levels of autonomy (ranging from teleoperation to full autonomy) in order to help the operators reduce their burden so that they can focus on critical tasks (such as grasping, visual servoing, reaching etc.) while letting the system take care of such necessary, but ancillary tasks as navigation and target acquisition.

Testbed

A state of the art mobile manipulation test bed has been developed at the Idaho National Lab as shown in Figure 1.

The WMR is a Segway RMP 400, which is a 4 wheeled skid steered robot with a payload capacity of about 180 kg (400 lb). It has a top speed of about 29 km/h (18 mph). It is composed of two Segway RMP200 robots such that there are two controllers, one for front wheel pair and one for the rear pair. The control of the robot needs coordination of both the control units. The onboard Li battery set supplies 1600 AmpHrs at 72V. An extra battery set with Lead-Acid batteries supplying 58 AmpHrs at 24V is also mounted onboard that runs the arm, the onboard PC as well as the sensors. There is a Linux (Mandriva LE 2005) based computer (Kontron ETX-PM processor board with 1 GB memory, 1.8 GHz Pentium M 745 processor) onboard that controls all the hardware components. A 360 degree laser range finder is placed at the center and 180 degree laser range finder is installed in the frontof the WMR for mapping and localization. The top cover is raised to accommodate the payload battery pack, the computer and the 360 degree laser scanner. A special attachment is designed and fabricated to mount the arm as well as the 180 degree laser scanner.

The LW3 Arm is a 3rd generation light weight arm from Schunk GmbH. A 7 DOF configuration is used to enhance obstacle avoidance capabilities of the system. It has a payload capacity of 10 kg and has a reach of 1 m at tool plate. The arm is mounted at 45 degrees with horizontal plane as shown in Fig. 1 so that its workspace aligns better with the task space. One other reason is with that the arm is able to reach to the ground. A three-fingered Barrett hand from Barrett Technologies is mounted at the end effector for grasping . The arm is powered by the payload battery pack. All the arm modules run at 24 V unlike the first generation modules thereby simplifying the energy requirement. The arm is controlled by the onboard computer using CAN protocol based CAN USB mini by ESD Electronics similar to the system at RRG. The CAN USB mini communicates with the arm at 500 KBPS giving the arm a band width of 165 Hz. This effort has also produced a novel system for capturing 3D mosaics of the local environment using a Swiss ranger flash IR camera that is located on the end of the arm. This allows for 3D range data out to four meters. The Swiss ranger is mounted at the end of the arm. It is connected via USB to the computer and powered using the onboard batteries.

Figure 1. 2nd Generation Mobile Manipulation testbed The mobile manipulation system shown in Fig. 2 was developed at RRG as a test bed for the research collaboration. Following section details various components of the system.

The WMR is an iROBOT’s ATRV2 which is skid steered robot with four independently driven wheels. The robot is powered by four onboard batteries supplying 24 Volts at 66 AmpHrs. These batteries power the robot, the sensory suite and a part of the arm. There is a Linux based onboard computer that controls all the hardware components. For mapping and localization, the system is equipped with a suite of sensors such as sonar sensors and a 180 degree laser range finder. For more accurate localization of the mobile manipulation system, an indoor GPS system from ArcSecond is also integrated into the system. For mounting the arm and external battery pack, a frame was designed and fabricated as shown in the picture.

The arm is a PowerCube arm from Schunk GMBH. It consists of modular actuators and links as building blocks. The actuators are self contained with motor, gear train and even high level motion control. While it is possible to configure the robot as per task requirements, the current configuration is a 6 DOF Puma configuration with approximately 1 m reach and 5 kg payload. There is a two fingered parallel gripper, also by Schunk GmbH, at the end effector for grasping. A force Torque sensor from ATI is mounted just before the gripper for force control applications. The arm needs to be powered through onboard batteries. There are two different voltage requirements for the arm modules. The two shoulder modules need power at +48 V while the remaining five modules (four of the six modules for arm and one parallel gripper) need +24 V power. The five modules (including gripper) are powered by the main battery pack that also runs the WMR. However, the two shoulder joints are powered by external battery pack as shown in Fig. 2. With the onboard power available, the system runs for about 3 hours before needing to charge the batteries. The arm is controlled by the onboard Linux computer using CAN protocol based CAN USB mini by ESD Electronics. The CAN USB mini communicates with the arm at 250 KBPS giving the arm a band width of 65 Hz.

Figure 2. First Generation Mobile Manipulation Testbed Back to Top

Research Results

Demonstration Vidoes

	Simulation showing a Mobile Manipulator Entering a Doorway This short clip shows a simulation of a Mobile Manipulator (the First Generation System at RRG) entering a doorway while holding the door knob with its hand. Notice that the whole system acts in a coordinated manner to keep the hand position while the platform moves through space.(Download ~2.2 MB)
	Object Retrieval Using Swiss Ranger The video shows a graphical window in the top left corner. This window shows the video feed from Swiss Ranger (SR3000) with the 3D point cloud overlay. User can select a particular object and command the manipulator to retrieve it in real time. (Download ~7.5 MB)
	User Interface Demo The demonstration shows integration of the mobile manipulator with the 3D Interface that is developed by INL. The video shows that the arm can be tasked by dropping the hand target in the 3D world shwon in the interface. Based on whether the target is reachable to manipulator or not, the mobile robot is commanded to approach the target location and the manipulator is then commanded to reach out. (Download ~26 MB)
	A Door Opening Demo A door opening demonstration. The iGPS system is used for the locating the door knob and also for real-time positioning of the mobile manipulator. (Download~13 MB)
	Object Retrieval Using iGPS The demonstration of an object retrieval. The object is identified using iGPS. (Download~10 MB)
	Real-Time Manipulator Motion Planning The demonstration of iGPS-PowerCube Integration. The iGPS sensor is actively tracked by the PowerCube manipulator. (Download ~1.2 MB)

Publications

Kulkarni, A., Bruemmer, D., Kapoor, C., Kinoshita, R., Atherton, J., Whetten, J., Nielsen, C., and Pryor, M., "Software Framework for Mobile Manipulation", Proceedings of ANS 2nd International Joint Topical Meeting on Emergency Preparedness and Response and Robotic and Remote Systems, Albuquerque, New Mexico, March 2008.

Bruemmer, D., Few, D., Kapoor, C., and Goza, M., "Dynamic Autonomy for Mobile Manipulation," In Proc. of the ANS / IEEE 11th Annual Conference on Robotics and Remote Systems for Hazardous Environments, Salt Lake City, UT, Feb. 12-15, 2006