The task is to develope a systematic method for a mobile robot to learn
how to navigate autonomously in indoor and ourdoor natural environment
based on video images. No active sensors, such as sonar or infrared proximity
sensors, can be used. Learning and performance stages are a single stage:
performing while learning, and are both conducted in real time. Currently,
a preliminary version of system runs on a SUN SPARC-1 on board of Rome,
a mobile robot built on a Labmate platform from TRC. The navigation control
signals are corrected heading direction, speed and step distance. In the
learning phase, the Rome was manually controlled to take pictures at typical
positions of a hallway section for training. The intended control signal
associated with each scene was also record as a desired output. The system
learns the association between the input image and the output control signal
in real time. In many test drives Rome has performed so far, Rome has successfully
navigated, even along the straight section and the turn that it has not
learned. Rome is not confused by passers-by during its navigation.
S. Chen and J. Weng, ``State-Based SHOSLIF
for Indoor Visual Navigation,'' IEEE Trans. Neural Networks, vol.
11, no. 6, pp. 1300-1314, Nov. 2000. Dowload PostScript. J. Weng and S. Chen, ``Visual Learning
with Navigation as an Example,'' IEEE Intelligent Systems,
vol. 15, no. 5, pp. 63-71, 2000. Dowload PDF
file. J. Weng and S. Chen, ``Vision-guided
navigation using SHOSLIF,'' Neural Networks , vol. 11, pp. 1511-1529,
1998. Dowload PostScript. J. Weng and S. Chen ``Incremental learning
for vision-based navigatioon,'' in Proc. International Conference on
Pattern Recognition, Vienna, vol. IV, pp. 45- 49, Austria, Aug. 1996.
Click here
to down load the paper (PostScript). J. Weng and S. Chen, ``SHOSLIF-N: SHOSLIF
for Autonomous Navigation (Phase II),'' Technical Report CPS-95-22, Department
of Computer Science, MSU, May, 1995. Click here
to down load the paper (PostScript). S. Chen and J. Weng, ``Vision-based
navigation using self-organizing learning,'' in Proc. 2nd Asian Conf.
on Computer Vision, Singapore, pp. 239-243, Dec. 5-8, 1995. S. Chen and J. Weng, ``Autonomous navigation
through case-based learning,'' in Proc. IEEE Int'l Symposium on Computer
Vision, Coral Gables, FL, pp. 359-364, Nov. 20-22, 1995. S. Chen and J. Weng, ``Autonomous navigation
using recursive partition tree,'' in Proc. Workshop on Vision for Robots,
Pittsburgh, PA, pp. 130-135, Aug. 7-9, 1995. S. Chen and J. Weng, ``SHOSLIF-N: SHOSLIF
for Autonomous Navigation (Phase I),'' Technical Reports, CPS-94-62, Department
of Computer Science, MSU, December 1994.
To Weng's Home Page: http://web.cps.msu.edu/~weng/