Murphey:Murphey Projects

This work focuses on control and estimation for vehicles that must slip in order to turn. They have the unusual property of being controllable in a hybrid-systems sense, but they are not locally controllable with bounded inputs. Hence, the hybrid structure of the dynamics must be taken into account. We are implementing our results on the autonomous vehicle seen to the right.

• In its prime, Flexy-Flyer with the help of Mike Konshak and team Robot Dojo, took top honors in season two of Robotica.
• Now because of a generous donation from Mike Konshak, Flexy-Flyer has found a new home at the University of Colorado, Boulder.

Mechanical grasping is a traditional area of study in robotics. However, when the mechanism is highly constrained, the contact will undergo stick/slip transitions and contact/no-contact transitions that must be taken into account for control purposes. The experimental system to the right allows us to investigate these effects in a dynamic setting, unlike the PI's prior work in distributed manipulation that was effectively kinematic.

• The multiple-point manipulator to the right is only capable of 2DOF motion in each finger. It uses real-time operating system and vision feedback for determining the contact state of the system. Right now we are using standard hybrid observers for determining the contact state, but we are currently developing observers that take advantage of the geometric structure of the dynamics and kinematics.

Cooperative control has been widely studied in the past decade, but cooperative manipulation has received comparatively little attention. My interests in both areas are to show that "simple" controllers can achieve the desired objectives in most cases, and in particular to show that many of the current cooperative control techniques that involve nonsmooth analysis may be replaced by standard adaptive control techniques without any theoretical loss.

• Cooperative Control is being studied through the use of a suite of Roombas controlled wirelessly and independently through ARM embedded systems.
• Source code and samples of the experiments are available.
• Visit the Cooperative Control web page

Marionettes are high degree-of-freedom (40-50 DOF) systems that are extremely nonlinear with degenerate Lagrangians. Moreover, they have a strongly geometric structure. We are using them as a test-bed for studying the theoretical foundations of combining classical Riemannian geometry with graph theoretical techniques for efficient simulation and optimization of complex systems.

As an embedded control system, marionettes provide challenges in simulation, optimal control, hybrid control, and networking. We are solving these issues in a unified setting using the marionettes as an example, but the developed techniques are generically applicable to networked nonlinear systems. Plus, they are fun to play with!

• Aim: To develop a robotic puppet stand that can interpret and perform the choreography of a play with several marionettes.
• Visit the Puppeteer Website

• Analysing limit sets directly can allow one to design parts and their environment so that part feeding or assembling naturally emerges from the dynamics. In this work (with Kevin Lynch from Northwestern University) we did experiments validating this forced "self-assembly".