Logo: Universität Hannover

Bipedal Autonomous Robot LISA

LISA-UH LISA is an experimental robot for doing research in the field of autonomous bipedal walking. Goal of the project is to develop an alternative mobile plattform for service robots.

Up to now, several robots have demonstrated how two-legged locomotion can be realized technically. However, in order to fully take advantage of two-legged locomotion, bipedal robots need quite more intelligent methods for gait stabilization and walking pattern generation. For the ongoing research in this field, LISA is an experimental plattform that allows us to implement and test new methods practically and explore issues beyond simulation.

A specialty of the shown walking machine is the construction of the hip joint, which is built like a ball-and-socket joint but with all hip motors attached to the upper body. The two legs of the robot are movable in a way, so that each foot can be moved spatially and rotated around all its three axes. LISA is a lower limb model without any degree of freedom in its upper body. Our research in the field of gait stabilization concentrates on methods that do not depend on a movable upper body such with rotating and/or compensating masses. Bipedal robots should be able to walk without them as far as possible. With this approach in mind, the new stabilization methods found will provide the upper body with as much flexibility as possible. The larger the stability margins the less restrictive the biped has to be with application specific upper-limbs.

The walking machine LISA is an autonomous system and does not require any cable or link to some external base station during operation. All microcontrollers, processors, power modules and the rechargeable battery have been integrated into the torso.

Technical Data

LISA-UH
  • Total height 124 cm
  • Total weight approx. 40 Kg
  • Length of thigh and shank 35 cm
  • Distance between legs 18 cm
  • Spherical parallel manipulator hip joints
  • 12 actuated degrees of freedom, each leg:
    • 3 in each hip joint,
    • 2 in an ankle and
    • 1 in a knee joint
  • DC motors with Harmonic Drive gears, encoders and motor current sensors
  • 6D-Force-/Torquesensors at the feet
  • 3 inertial measurement units (IMUs) at the trunk and both thighs
  • Flexible toe link (passive only, with potentiometer)
  • 1 GHz CPU, 3x 32 Bit uC, CAN-network
  • Autonomous system: battery, CPU and power electronics on board

Technology

LISA has twelve active degrees of freedom. All axes are actuated by DC motors with harmonic drive gears and are equipped with incremental encoders for the servo control feedback. Each leg has two rotatory degrees of freedom (DOF) in the ankle, a revolute joint in the knee and a ball-and-socket-like hip joint with three rotatory DOF. The hip has been built as a spherical parallel manipulator (see below). Thus, the legs are lightweight, quick and have nearly the workspace (region of movability) of a human leg.

The robot has one six dimensional force-torque-sensor in each foot, which can measure the floor reaction forces and torques simultaneously. The force and torque signals are intensely used for feedback control to keep the robot's balance. Additionally, an inertial measurement unit (IMU) has been integrated into the torso for the purpose of gait stabilization. The IMU is a cluster of several gyroscopes and accelerometers and is able to measure the orientation of the robot in space.

Besides the rechargeable battery, the torso carries the twelve power electronic modules generating the output voltages for the motors. For each leg, a single, PowerPC-based 32-bit microcontroller (MPC555, 40MHz) samples the encoders and computes the feedback control. Another PowerPC-CPU with 1 GHz is used for central computational tasks. Both, CPU and microcontrollers run the realtime operating system RTOS-UH.

Based on the bipedal robot BARt-UH (as of 2003), which was able to walk in one direction only, a new robot was developed with twice as much degrees of freedom. Thus, the robot is able to walk straight ahead aswell as sidewards and can turn around on the spot. The pictures show the trunk furthermore the knee and ankle joints of LISA. The rightmost picture shows the first and very simplified feet without any sensors. Further below, the new "intelligent" feet with integrated force-torque sensors are presented in another picture.

Banner Aufbau

The development of LISA is a typical mechatronical task. Both the mechanical and the electrical design as well as the processor modules were developed at the institute. Typical components available on the market are less suited for the construction of a bipedal robot (too big in size, too heavy). Due to the optimized design and the mutual development of the components the best possible utilization was achieved.

Parallel manipulator hip joint

Hüftgelenke The hip joint consists of three active rotational degrees of freedom whose rotational axes intersect in one point. In contrast to most hip joints of other bipedal robots LISA's hip joint are built as spherical parallel manipulators. A comparable cardanian joint would lead to a heavier weight and due to the functionality the masses of some engines would have to be accelerated by other engines during motion.

Due to the parallel manipulator all engines rest to the trunk. Only a coordinated interaction of all engines leads to a controlled motion of the thigh. This enables a design with a thigh of minimal and a trunk of maximal weight which is an advantageous weight distribution for bipedal walking. Because of the parallel manipulator structure forces applied on the thigh are distributed among all three engines and therefore the power of the engines adds up.

The design of the spherical parallel manipulator was based on the "Agile Eye" of Prof. Gosselin, Canada, a fast rotational orienting camera-positioning device. In opposite to the "Agile Eye" an additional supporting stand was implemented to carry the robots weight and therefore to disburden the parts of the parallel manipulator.

Hüftgelenke

Disadvantageous for most parallel manipulators is a bigger complexity of the mathemical modell which mostly can only be handled numerically. For the hip joint we could derive analytical expressions for the inverse as well as the direct kinematic and kinetic equations. This enables us manifold analysis of the joint motions and efficient calculations at run-time.

Real-time system

Software for simulation and analysis

Bild des Simulators

  • Simulation and measurement data acquisition in one system
  • Visualisation of simulation results or of received measurement data with one visualization toolkit (VTK)
  • Direct dynamics with Open Dynamics Engine (ODE) or based on Newton-Euler-Jourdain equations
  • Environment for development and verification of algorithms implemented on the robot
  • CAN- and TCP/UDP/IP-interface to the robot
  • Interface to the optical tracking system DTrack von ART
  • Interfaces to other software like Gnuplot, Matlab/Simulink, ...
  • Recording video data parallel to measurements, processing in VTK with feature-overlay and generating videos films

Downloads

Videos

  • The following video demonstrates the movability of the legs. As one can see, the workspace is large enough to turn on the spot with 90°-steps.
  • A so called statically stable step sequence is shown in the next file. The torso is pushed above the footprint of the standing leg before the swinging leg can follow. Statically stable motions are neither quick nor efficient. However, they are a good point to start from...
  • LISA turns on the spot with two statically stable steps.
  • The following file tries to illustrate the measurement of the floor reaction forces (green arrows), reaction torques (red arrows) and the resulting center of pressure (yellow point). In this video, the robot has no real idea of its position in space. When it turns around the edge of the foot, the absolute position of the robot has errors. However, please bear in mind that this fact does not influence the illustration of the COP measurement.
  • The next video shows an experiment on the stabilization of the robot. While the robot is hit the first two times, a feedback control based on the force-torque-signals is balancing the robot in its upright pose without overshooting into the direction of the swinging leg. The foot of the supporting leg stays on the floor and does not rotate. Before the third impact, the controller is disabled. The robot does not absorb the energy of the impact and overturns.

Bilder

Literature

A complete and searchable list of publications with abstracts and citations in bibtex format can be found here.

The development and construction of LISA has essentially been done in the context of the PhD thesis

  • Hofschulte, J.: Zweibeiniger Roboter mit parallelkinematischen Hüftgelenken. Books On Demand GmbH, Norderstedt, 2006.

The following publications describe the robot and/or its components.

  • Seebode, M.; Gerth, W.: Echtzeitsystem für einen zweibeinigen Roboter mit adaptiver Bahnplanung. In: Holleczek, P.; Vogel-Heuser, B. (Hrsg.): Mobilität und Echtzeit. PEARL 2007, Informatik Aktuell, Springer Verlag Berlin Heidelberg, 2007, S. 88-97.
  • Hofschulte, J.; Seebode, M.; Gerth, W.: Parallel Manipulator Hip Joint for a Bipedal Robot. Proceedings of the 7th International Conference on Climbing and Walking Robots and the Support Technologies for Mobile Machines - CLAWAR 2004, 22.-24.09.2004, Madrid, Spanien.
  • Gerecke, M.; Hofschulte, J.; Gerth, W.: Realization of a Lightweight Sensory Foot for a Bipedal Robot. Proceedings of the Sixth International Conference on Climbing and Walking Robots and their Supporting Technologies - CLAWAR 2003, 17.-19.09.2003, S. 895-902.
  • Strasser, R.; Seebode, M.; Albert, A.; Gerth, W.: Extrem kompaktes SoC-Konzept eines Gleichgewichtsorganes für einen Laufroboter. In: P. Holleczek, B. Vogel-Heuser (Hrsg.): Verteilte Echtzeitsysteme. PEARL 2003, Springer Verlag Berlin Heidelberg, 2003, S. 49-58.
 


Institute of Automatic Control - URL http://www.irt.uni-hannover.de/forschung/asr/lisa_en.html
 
Responsible: Institute of Automatic Control, Biped Team, last modification 01.09.2009
© Leibniz Universität Hannover 1998 - 2009