Estimation of neighbor velocity in a leader-follower network
(English)Manuscript (preprint) (Other (popular science, discussion, etc.))
In applications where mobile robots are set to follow other moving objects, it is often assumed that not only the relative position, but also the velocity of the object in question is available to the tracking robots. In this paper, we consider a leader-follower system where the followers are equipped only with vision and range sensors. Thus, the followers can detect the positions of their neighbors, but they are unable both to measure the velocity of other agents directly and to obtain estimates of the unknown velocities through communication with other agents. In order to implement the desired leader-follower controls, the followers must use available sensor data to compute individual estimates of their respective leaders' velocities. One option would be to use the Extended Kalman Filter to obtain the estimates, but in this paper we take another approach and propose two separate state observers that can be used by the followers to obtain estimates of the missing velocities and that are also shown to stabilize the tracking controls used by the robots. To evaluate the observers, the tracking performance is compared in simulations for a system that uses observer estimates of the leader's velocity in the control equations and a system that computes the velocity estimates using the standard Extended Kalman Filter algorithm.
IdentifiersURN: urn:nbn:se:kth:diva-10907OAI: oai:DiVA.org:kth-10907DiVA: diva2:231581
QC 201007152009-08-142009-08-142010-07-15Bibliographically approved