Most quadrupedal robots today are designed with mechanically simple, fore/hind symmetric legs. This design choice is taken to simplify modeling and control tasks, but it likely comes at the expense of versatility. The overarching goal of this project is to explore the benefits that different types of leg designs might bring to bio-inspired legged robots. For example, designs based on closed-loop (e.g. six-bar) linkages can be created to shape the aesthetics, range of motion, movement envelope and mechanical advantage of a leg (the fore and hind legs of a robot probably need to be designed with different goals in mind), while flexible materials can be used to passively improve stability in unstructured environments. Your task will be to develop whole-body controllers for legged robots that feature these types of leg designs.
On a technical level, two types of approaches can be pursued:
1) The Jacobian of a leg's kinematic structure, or similarly the force Jacobian of a flexible leg, can be used to establish a mathematical relationship between instantaneous motor torques, joint accelerations and the ground reaction forces that are applied by the robot's feet. This relationship can then be used to extend traditional whole-body locomotion control formulations.
2) Our lab is developing an analytically differentiable physics simulator for rigid bodies, elastic objects, frictional contact, etc. The derivatives output by this simulator can be used to optimize parameters of higher-level locomotion controllers, to directly compute motor torques that lead to a desired outcome over a short-horizon planning window, or to optimize kinematic and elastic parameters of a leg's design.
Most quadrupedal robots today are designed with mechanically simple, fore/hind symmetric legs. This design choice is taken to simplify modeling and control tasks, but it likely comes at the expense of versatility. The overarching goal of this project is to explore the benefits that different types of leg designs might bring to bio-inspired legged robots. For example, designs based on closed-loop (e.g. six-bar) linkages can be created to shape the aesthetics, range of motion, movement envelope and mechanical advantage of a leg (the fore and hind legs of a robot probably need to be designed with different goals in mind), while flexible materials can be used to passively improve stability in unstructured environments. Your task will be to develop whole-body controllers for legged robots that feature these types of leg designs.
On a technical level, two types of approaches can be pursued:
1) The Jacobian of a leg's kinematic structure, or similarly the force Jacobian of a flexible leg, can be used to establish a mathematical relationship between instantaneous motor torques, joint accelerations and the ground reaction forces that are applied by the robot's feet. This relationship can then be used to extend traditional whole-body locomotion control formulations.
2) Our lab is developing an analytically differentiable physics simulator for rigid bodies, elastic objects, frictional contact, etc. The derivatives output by this simulator can be used to optimize parameters of higher-level locomotion controllers, to directly compute motor torques that lead to a desired outcome over a short-horizon planning window, or to optimize kinematic and elastic parameters of a leg's design.
Please see the project description. Both BSc and MSc theses can be done within the context of this project. Knowledge of rigid body dynamics, and experience programming in C++ are required.
Please see the project description. Both BSc and MSc theses can be done within the context of this project. Knowledge of rigid body dynamics, and experience programming in C++ are required.