Manipulation & Control
Robotic manipulation was part of the Robotics Institute from its beginning in 1979, primarily in the form of manufacturing applications, and the notable development of the first Direct Drive Arm. Manipulation emerged as a separate research focus with the founding of the Manipulation Laboratory in 1982, and thrived and grew over the subsequent years through the addition of the Microdynamic Systems Lab and the Biorobotics Lab.
Over the last five years interest in manipulation has surged. New applications have emerged, from the manipulation of everyday objects in human environments, to the disposal of roadside bombs. Mobile platforms and manipulators and sensory capabilities are improving rapidly. Together all of these factors are pushing manipulation research in interesting directions, and leading us to new perspectives and approaches. This surge is evident throughout the Robotics Institute, including Field Robotics, Graphics, Humanoids, Human Robot Interaction, Manufacturing, Medical Robotics, and Quality of Life Technology. Perhaps the most visible signs of change over the last five years are the creation of the Personal Robotics Lab at our Intel "lablet" and the Search-Based Planning Lab.
Manipulation, Control & Path-Planning
Fundamental manipulation research is a hallmark of the Robotics Institute, and Erdmann's recent work continues that tradition. Erdmann found a graph-theoretic technique for modeling planning problems, so that the global capabilities are revealed by the homotopy type of a "strategy complex". A key result is a controllability theorem: A robot can move anywhere in its state graph despite control uncertainty precisely when the graph's strategy complex is homotopic to a sphere of dimension two less than the number of states.
Several other less abstract (but still fundamental) manipulation results have been recently developed. Research on manipulation in human environments has expanded and changed our perspective on everyday manipulation. Led by Srinivasa, a collaboration of CMU and Intel researchers have identified interesting new problems and raised our consciousness on issues such as how to deal with clutter, how to hand objects to a human being, and how to plan motions that won't surprise nearby humans. The group has been remarkably successful in proceeding from discovery of new problems to development of new technology, which in many cases has been adopted by other groups both inside of CMU and out.
Unique Mobility, Motion-Planning and Control
A high-impact high-visibility "unique mobility" project is the Modular Snake Robots project led by Choset. The group has developed novel mechanical design elements, and planning and control techniques, to demonstrate a remarkable set of capabilities. Perhaps the single most popular demo in the Robotics Institute is the robot snake climbing up one's leg. Among the most remarkable accomplishments is the wide range of applications addressed: from exploration (archeological forensics), to manufacturing (airplane assembly), and robotic surgery (minimally invasive heart surgery).
Ballbot, led by Hollis, is another novel approach to mobility. Ballbot balances and moves on a single spherical wheel, a radical departure from traditional quasi-static approaches. This unique approach also brings unique problems straddling the boundary between planning and control. To achieve fast, graceful navigation for dynamically stable mobile robots like the ballbot, Hollis and Kantor and collaborators developed motion planning that is cognizant of the natural dynamics of the system and closed-loop controls that stabilize the system around those trajectories. A hybrid control framework pieces together locally-defined control policies to produce smooth, graceful, energy-efficient motions. The project's high impact is best indicated by the numerous imitations that have appeared since Ballbot's debut.
Motion-Planning for Mobile Manipulation
RI's most recent addition to the faculty is Likhachev who founded a new lab, the Search-Based Planning Lab. Likhachev's group will continue his focus on developing fundamental new graph-search techniques addressing the many unique challenges of mobile manipulation. During his time at the University of Pennsylvania, Likhachev and his collaborators developed and refined a variety of techniques for the efficient search of nominally high dimensional and complex search spaces, and demonstrated the capabilities on a Willow Garage mobile manipulator.
The highlights above focus on a few high-profile projects, but there is much more. Our work in the "unique mobility" category includes
- the RISE climbing robot,
- the DSAC and DTAR dynamic planar climbing robots, and
- locomotion techniques suitable for a high-centered mobile robot or an inverted turtle.
Other work in manipulation includes:
- folding of origami,
- manipulation preparatory to grasping,
- new approaches to autonomous manipulation with simple hands,
- assembly of consumer electronics,
- design of hands with scale-invariant or pose-invariant contact geometry,
- imitation learning of manipulation strategies,
- data-driven approaches to grasp recognition and localization, and
and many more projects in a diverse set of areas.
Table of Contents
- Robotics Institute Research Guide
Graph-theoretic technique for planning of modeling
Planning in the presence of clutter and humans
Modular Snake Robots
Rockin nd Rollin robot
Autonomous Door Opening with a Mobile Manipulator
Hands for simple grasping
Manipulation preparatory to grasping
Modular Snake Robots
CHOMP: Covariant Hamiltonian Optimization for Motion Planning
- Planning pre-grasp manipulation for transport tasks
- On the Topology of Discrete Strategies
- Grasp Synthesis in Cluttered Environments for Dexterous Hands
- Connection Vector Fields and Optimized Coordinates for Swimming Systems at Low and High Reynolds Numbers
- Two-link swimming using buoyant orientation
- A Dynamic Single Actuator Vertical Climbing Robot
- Legless Locomotion: A Novel Locomotion Technique for Legged Robots
- Trajectory Planning and Control of an Underactuated Dynamically Stable Single Spherical Wheeled Mobile Robot
- Hybrid Control for Navigation of Shape-Accelerated Underactuated Balancing Systems
- Planning for Autonomous Door Opening with a Mobile Manipulator
- Failure Detection in Assembly: Force Signature Analysis
- Grasp Invariance
- Robotic origami folding
- Preparatory object rotation as a human-inspired grasping strategy
- HERB: a home exploring robotic butler
- CHOMP: Gradient Optimization Techniques for Efﬁcient Motion Planning