Robust Feature Tracking by Coordinating Heterogeneous Visual Processing Agents

Japanese version

We have proposed a robust visual feature tracking method that combines static feature extracting agents and feature tracking agents [3]. This method can robustly track rotating features, which has been difficult with only a correlation-based tracking method (Figure 1).
Multi-agent based feature tracking
Figure 1: Multi-agent based feature tracking

Multiagent-based Peg-in-hole

By combining heterogeneous visual agents and joint agents, the robot can accomplish a task of inserting pegs with various cross-sectional shapes [3] (Figure 2).

Multi-agent based Peg-in-hole
Figure 2: Multi-agent based peg insertion task

  1. Mori, A., Hiramatsu, K., Naya, F. and Osato, N.: A Robot-Controlling Agent Description with Finite State Machines, IEEE 4th International Symposium on Distributed Autonomous Robotic Systems (DARS'98), May 1998.
  2. Mori, A., Hiramatsu, K., Naya, F. and Osato, N.: A Robot-Controlling Agent Description with Finite State Machines (in Japanese), Proc. 55th Annual Convention IPS Japan, Vol.2, pp. 434--435 (1997).
  3. Naya, F., Hiramatsu, K., Mori, A. and Osato, N.: Active Tracking based on Cooperation of Heterogeneous Visual Processing Agents (in Japanese), Proc. 55th Annual Convention IPS Japan, Vol.2, pp. 136--137 (1997).

Contact: Futoshi Naya; Email: naya@cslab.kecl.ntt.co.jp

Back to Multi-agent based Robotics Page

Last modified on Tue Oct 6 17:22:35 1998