About

The Dartmouth Robotics Lab is a center for robotics research and education at the Dartmouth Computer Science Department. The director is Devin Balkcom.

Our primary interest is the mechanics of locomotion and manipulation -- the interface between robots and the physical world. We have studied exact optimal trajectories for vehicles, robot origami folding, efficient algorithms for planning stacking and unstacking, and robust minimal-sensing control of rigid-body contact tasks.

Education and outreach are also a critical part of our work. Current and past undergraduate research projects include a toy rock-climbing robot, and a vehicle that can rock back and forth to free itself when its wheels are buried in sand. Our lab is also involved in the yearly Dartmouth Summer Robotics Camp for middle-school students.

Featured Research: Robotic Origami Folding

Balkcom, Mason, Demaine, and Demaine

Thin materials like sheet metal, paper, and cardboard are lightweight, inexpensive, and can be stored and shipped in bulk. Folding allows the construction of semi-rigid 3-D structures, including fast-food containers, paper bags, and file cabinets. Folding can also allow a single large surface or chain to be stored in a small volume; motivating examples include car airbags, space-telescope mirrors, and proteins. Finally, folding allows reconfiguration, without the need for disassembly and reassembly.

We have built the first origami-folding robot (left), capable of folding a paper hat, paper airplane, and paper cup. We have also analyzed more complicated folding techniques; work with Erik and Martin Demaine questions whether ordinary paper shopping bags can be mathematically folded.

Featured Research: Exact Optimal Trajectories for Robots

Balkcom, Kavathekar, Mason, Chitsaz, and LaValle

Robot control algorithms, like computer algorithms, allocate resources to accomplish tasks; resources include time, energy, degrees of freedom, steps of computation, and precision. A fundamental problem is to determine the minimum resources required -- the physical complexity of of the task. The optimal trajectories provide insight into the capabilities of a mechanism, independent of compromises introduced by particular planning or control techniques.

The exact optimal trajectories were previously known for two vehicles: the Dubins car, and the Reeds-Shepp car. We have discovered the optimal trajectories for a third vehicle; the common differential-drive (wheelchair-like) robot, and work on two more vehicles is nearing completion.

Featured Research: Rigid-body Mechanics

Loomis, Balkcom, Trinkle, and Gottlieb

Rigid-body models are fundamental to almost all algorithms for physical simulation, planning, and control. Even flexible objects are often modelled as collections of rigid bodies loosely joined by springs. As with all models, however, rigid bodies do not capture all desirable characteristics of the physical system. The solutions to rigid-body dynamics equations with Coulomb friction may be inconsistent, or indeterminate. Efficiency is also a concern -- computing the solutions to the dynamics equations in the inner loop of a design or control algortihm is expensive.

Balkcom, Trinkle, and Gottlieb designed an algorithm for computing all possible forces and torques consistent with constraints on the motion of rigid bodies, and used the algorithm to compute parts-seating plans that are robust to incomplete or incorrect sensor information and to frictional indeterminacy. More recently, Loomis and Balkcom have explored methods for efficiently computing manipulation plans for unstacking piles of objects by exploiting the similarity between the dynamics computations at different stages of the disassembly.

Undergraduate Research Project: Freeing Stuck Vehicles

Rosa, Pyke, and Balkcom

Wheeled robots get stuck. We have begun an exploration of increasing vehicle mobility in soft terrain by using the wheels to dig. We have explored two behaviors: oscillation of wheel-spin direction to dig out stuck wheels, and the use of controlled landslides to remove a hill that is too steep for the robot to cross. The work is preliminary and empirical, but suggests promising directions in which robot capabilities may be expanded.

We have implemented two open-loop escape techniques. The robot has escaped from being buried in sand using the oscillation technique. The robot has also climbed a sand dune by creating a pass shallow enough for it to cross using the digging technique.

Undergraduate Research Project: Toy Climbing Robot

Bell and Balkcom

We built a simple toy climbing robot in order to explore problems related to grasping, path planning, and robot control. The robot is capable of climbing a wall of pegs either under remote control, or on the basis of a set of pre-recorded keyframes. In addition, the robot can climb certain peg configurations using a cyclic gait. All communications are sent through an infrared connection, and the tether to the robot consists only of two power wires. Due to the minimalist, non-prehensile grasping method the robot is capable of actively removing error while climbing, which is necessary to enable the robot to climb open-loop in the absence of environmental sensing.