Unconstrained omni-directional walking in virtual worlds

CyberWalk project

 

Last update: December 20, 2011

 NEWS

CyberCarpet "ball-array" platform

In December 2011, the paper "Motion Control of the CyberCarpet Platform" has been accepted in the IEEE Transactions on Control Systems Technology and has then appeared in vol. 21, no. 2, pp. 410-427, 2013.

You can find here the ten videos accompanying the paper. These videos refer to: 

-----

CyberWalk "omni-directional/belt-array" platform

In February 2011, the paper "CyberWalk: Enabling unconstrained omnidirectional walking through virtual environments" has been accepted in the ACM Transactions on Applied Perception and has then appeared in vol. 8, no. 4, October-November 2011. The paper summarizes the results obtained with the CyberWalk omnidirectional platform, including also perceptual evaluation of the platform by several users.

In April 2010, the IROS'09 paper was featured also in the Automaton blog of IEEE Spectrum.

In February 2010, the paper "Making virtual walking real: Perceptual evaluation of a new treadmill control algorithm" (pdf) has been published in the ACM Transactions on Applied Perception, vol. 7. no. 2, pp. 11.1-11.14, 2010. The paper contains the control approach for keeping a walking user at the center of a 1D linear treadmill, without extra supports and limiting perceptual effects. See the video. The used control algorithm is the basis for the design in the 2D case.

In October 2009, we presented the paper "Control Design and Experimental Evaluation of the 2D CyberWalk Platform" (pdf) to IROS'09. Have a look at the video attached to this paper that shows how a clever design of the control gains can facilitate a more natural walking behavior for the user. 

-----

The CyberWalk project ended in April 2008 with a workshop where the system was presented to the press and scientific community. Many attendees have tested the behavior of the full system under the control laws developed by our group. This video shows the first steps of an untrained person moving on the platform, while this movie (made by the Max Planck Institute for Biological Cybernetics) gives a more complete description of the overall system. A nice history of the project can also be found at this link (thanks to Martin Schwaiger from TUM for this nice work).

 

Despite recent improvements in Virtual Reality technology, it is at present still impossible to physically walk through virtual environments in a natural way. The goal of the European research project CyberWalk is to significantly advance the scientific and technological state-of-the-art by enabling natural, unconstrained, and omni-directional walking in virtual worlds. This should be achieved through the use of an actuated platform (the CyberCarpet) that compensates for the walker's locomotion in such a way to keep her/him close to the platform center.

A walker will be allowed to execute slow or fast locomotion in any planar direction, or even step over and cross his/her legs, while remaining on the CyberCarpet platform. A markerless visual tracking system will locate the instantaneous position of the walker on the carpet, providing this information to the platform motion control system. The latter will command two actuation devices linearly moving the belt and rotating the turntable in such a way that the walker is pulled toward its center.

Documents

We derived a first-order kinematic model together with several control laws that ensure the recovery of the walker position when he/she is moving unpredictably on the platform. You can find the details in this paper presented at ICRA 2006 and in this paper presented at Syroco 2006, both written by A. De Luca, R. Mattone and P. Robuffo Giordano.

Simulations

We present some video clips in order to show the behavior of the control laws developed for the CyberCarpet. In the movies, the walker is always represented as an upright yellow "wheel" rolling on a turntable, while two other cyan wheels, placed at both sides of the platform, indicate the motion of the belt. The platform is represented as square only for the purpose of a clear visualization of the belt main direction. The simulations were generated using the Simulink and Visual Nastran environments. For the actual control laws, please refer to the formulas in the related paper.



ICRA 2006

In this first set of videos the walker is asked to move along a square path with 3 m sides, starting at rest from the Init absolute position, traveling along each edge with a trapezoidal velocity profile and stopping at each corner to change his/her orientation. Note that, without motion control of the platform, the walker would exit from the boundary of the circular platform (the dotted circle in figure on the right).

  • Square path traveled by the walker without actuation of the platform
  • Square path traveled by the walker under the platform control law (3-4) (static feedback)
  • Square path traveled by the walker under the platform control law (3) and (7) (static feedback + integral term)

The second controller has a better performance in recovering the walker's straight motion, but an unwanted overshoot is added to the overall behavior, yielding a more erratic global motion.

SYROCO 2006

In this second set of videos the walker is asked to move along the same square path as before, but starting at a different absolute position, i.e., one that would immediately lead to control singularity problems with the previous laws. We avoid the use of an integral term in the previous scheme (in order to recover a straight line motion), by implementing a non-linear observer of the walker's velocity. This information is then used as a feed-forward term to compensate for the walker's voluntary motion.

Finally, we dealt with the singularity at the origin by adopting two different strategies

  1. Hysteresis controller: two small circular dead-zones around the origin switch on/off the platform motion when the walker is close to the center. This solution provides a good overall performance, but some chattering is introduced between the two dead-zones boundaries
  2. Smooth controller: at the cost of a slower rate of convergence (quadratic instead of exponential), the singularity at the origin is avoided, yielding a more "natural" behavior
  • Square path traveled by the walker without actuation of the platform
  • Square path traveled by the walker under the hysteresis control law + velocity observer
  • Square path traveled by the walker under the smooth control law + velocity observer

Despite the slower convergence, the smooth feedback provides a better global behavior, gently pulling the walker towards the origin.

Other simulations

We tested our control laws in many other cases. Hereafter you can find several clips involving other possible scenarios

Experimental results

An experimental validation campaign has been carried out in October 2006 at the TUM (Technischen Universität München). Follow this link for more information.


Back to the LabRob Home Page