Workshop: Sandy Enoch and Vladimir Ivan

Sandy Enoch

BLUE: A bipedal robot with variable stiffness and damping. Construction and initial testing

Human walking is a dynamic process, which makes use of the compliance of the human body in order to be efficient and stable. We have a particular arrangement of muscles and tendons which allows us to change the stiffness and damping of our joints as we move. Robots, however, typically have very rigid joints, which make them non-ideal for handling the "controlled fall" of walking, and not inherently robust to disturbances that occur. BLUE is a new compliant bipedal robot designed around variable impedance joints and will, uniquely, be capable of independently varying both joint stiffness and damping. This ability should make it more efficient at walking, more robust, and open up a host of possible applications for the technology.

In this talk I will recap the motivations for and design of BLUE, detail the construction, electronics design and low level software, and show some initial testing. I will also detail first moves towards the controller design, and present the design of miniBLUE, a lightweight 3D printed variable impedance biped.

Vladimir Ivan

Hierarchical motion planning in topological representations

Motion can be described in alternative representations, including joint configuration or endeffector spaces, but also more complex topological representations that imply a change of metric or topology of the motion space. Certain types of robot interaction problems, e.g. wrapping around an object, can suitably be described by so-called writhe and interaction mesh representations. However, considering motion synthesis in only a topological space is insufficient since it does not account for additional tasks and constraints in other representations. I will present methods to combine and exploit different representations for motion synthesis and generalization of motion to novel situations. Our motion synthesis approach is formulated in the framework of optimal control as an approximate inference problem. Motion generalization is performed by projecting motion from topological to joint configuration space. Direct path finding in joint configuration space is extremely hard whereas local optimal control exploiting a representation with different topology can efficiently find optimal trajectories.