Local Control Policies for Global Control and Navigation

Dave Conner

Abstract

In this talk, I present a method for decomposing integrated navigation and control tasks into discrete sub-tasks, and solving those sub-tasks with feedback control policies. The planning system is no longer responsible for specifying a continuous path for the system; it instead chooses from a discrete set of sub-tasks. The feedback control policies associated with each sub-task are then responsible for execution of each sub-task. The technique presented is based on a cellular decomposition of the system's free space, and includes a generic control policy, applicable over each cell in the decomposition. The resulting composition of local control policies induces a global control strategy that brings any savable initial condition to the goal, provided that there is a single connected component of free space containing both the start and goal configurations. The underlying control policies are designed to respect environmental constraints such as obstacles, velocity bounds, and acceleration bounds. Control policies for fully actuated kinematic and dynamical systems will be discussed.
I conclude by outlining the future work extending the approach to systems with non-holonomic constraints, as well as underactuated systems.
(This talk covers research in support of my upcoming proposal, so feedback is appreciated.)

Presentation Slides

Ppt (8.2Mb)

 

Speaker Appointments

For appointments, please contact Dave Conner (dcconner@cs.cmu.edu).