Ph.D. Dissertation Defense: Dipankar Maity

Thursday, September 6, 2018
11:00 a.m.
AVW 2168
Maria Hoo
301 405 3681
mch@umd.edu

ANNOUNCEMENT:  Ph.D. Dissertation Defense


Name: Dipankar Maity

Committee:
Professor John S. Baras,  Chair/Advisor
Professor Nuno C. Martins
Professor Andre L. Tits
Professor P. S. Krishnaprasad
Professor Nikhil Chopra, Dean's Representative

Location:  AVW 2168

Title: CONTROLLER SYNTHESIS UNDER INFORMATION AND FINITE TIME LOGIC CONSTRAINTS

Abstract:
In robotics, networks, and many related fields, a typical controller design problem needs to address both logical and information constraints. The logical constraints may arise due to the complex task description or decision making process, while the information constraints emerge naturally as a consequence of the limitations on communication and computation capabilities.

In the first part of the thesis, we consider the problem of event-based controller synthesis that addresses the information constraint problems in controller design.  Classically, a controller requires continuous availability of the state/output measurements in order to compute its action; and hence increasing the cost of operation due to high network bandwidth utilization, and energy consumption in continuous sensing. In many situations, it so happens that the measurement does not change fast enough and there is no need to transmit these measurements continuously. Taking motivation from this fact, we consider the case where controller only obtains intermittent state feedback. Consequently, the new system will deviate from its behavior under continuous feedback. However, we are interesting in quantifying how much would it deviate? Given a bound on this deviation, can we design some intelligent controller that requires intermittent measurements and still manages to keep the deviation from continuous controller within this given bound? Two important questions remain to be answered here: 1) What will be the controller structure under intermittent feedback? and 2) How the system will know when to transmit the measurement to the controller? When the system sends out measurement to controller, it is termed as an "event”. Thus, we are looking for an event-generator and a controller to perform event-triggered control under the constraints on availability of the state information.

In the next segment we move on to the study of logical constraints on the system's behavior. This part focuses on controller synthesis problem with spatio-temporal constraints on the trajectory of the system; and a robot motion planning problem fits as one of the exemplary of these kind of problems.  The motion planning tasks that we consider here contain finite time logic constrains expressed using temporal logic. We propose a two stage framework, for this class of problems: In the first stage we find the optimal waypoints for the robot to achieve the task and in the second stage we implement a controller that moves the robot between the waypoints. We consider time bounded task and hence the waypoints are time-stamped For a general class of problems the framework would produce set of initial and final points and the system trajectory will pass from one initial point to final point and then to the next initial point. 

Finally, in the last section of this thesis, we consider the problem of addressing the logical and information constraints together. The problem is approached with the notion of robustness of logical constraints. We propose two different frameworks for this problem with two different notions of robustness, and two different approaches for the controller synthesis. One framework relies on the abstraction of the dynamical systems into a finite transition system, whereas the other relies on tools and results from prescribed performance control to design continuous feedback control to satisfy the robust logical constraints. However, at the end, both the continuous feedback controls are replaced with event-triggered controllers using the results obtained in the first part of the thesis.  

 
 

Audience: Graduate  Faculty 

remind we with google calendar

 

March 2024

SU MO TU WE TH FR SA
25 26 27 28 29 1 2
3 4 5 6 7 8 9
10 11 12 13 14 15 16
17 18 19 20 21 22 23
24 25 26 27 28 29 30
31 1 2 3 4 5 6
Submit an Event