[ISL Logo]

Robotics Research

The ISL web site is maintained by Sean Andersson. Most recent update 03/08/01.

Robot localization

If a robot does not know its position it can be difficult to implement global planning and higher level tasks. Many robots are equipped with precise odometry. This odometry can drift from the actual position to due robot slip, errors in the odometry calculation, etc. Simple experiments shows that the robots in our lab lose track with reality quite quickly. A method is needed, therefore, for localizing from other information.

The importance of this problem is evidenced by the large body of work in the literature. Several probabilistic algorithms have shown promising performance. We would like to implement such an approach and investigate improvements. A few examples can be found in Sebastian Thrun's work. A few example choices are:

Ongoing and future work

  • Implement robot localization using the map server
  • Investigate particle filter approaches in localization
  • Investigate efficient localization plans under MDLe
  • Investigate landmark based localization