Research Abstracts - 2007

Figure 1. JPL LEMUR II-B Climbing Robot (image courtesy of JPL)

We have been working on state estimation in hybrid discrete-continuous systems. We applied recent techniques to two robotic applications; a cooperative construction scenario on the MIT MERS robotics testbed[1], and a contact and force detection problem on the NASA JPL LEMUR testbed (above).

Our research in this area has developed methods for active hybrid estimation. The key idea behind active estimation is that much more information can be obtained by actively probing a system, rather than making observations passively. In the case of a UAV actuator fault, detection of the fault is impossible without requesting control effort from the actuator. We have developed novel methods for active diagnosis, whereby a controller can ensure that the nominal plan is successful, while optimally detecting faults[2][3][4].

Figure 2. Choosing control inputs to minimize the probability of
misdiagnosis (Bayes Risk).

References:

[1] Lars Blackmore and Steve Block. Control and Estimation for Cooperative Manipulation Tasks . CSAIL Technical Report MIT-CSAIL-TR-2006-011, 2006.

[2] L. Blackmore, S. Rajamanoharan and B. C. Williams. Active Estimation for Switching Linear Dynamic Systems . In the Proceedings of the Control and Decision Conference 2006.

[3] L. Blackmore and B. C. Williams. Finite Horizon Control Design for Optimal Discrimination between Several Models . In the Proceedings of the Control and Decision Conference 2006.

[4] L. Blackmore and B. C. Williams. Finite Horizon Control Design for Optimal Model Discrimination . In the proceedings of Control and Decision Conference 2005.