Stanford Seminar - When Design = Planning

()

Formalizing Design Thinking in Robotics

Introduced the topic of formalizing design thinking in robotics.
Presented a modular design system called BuildaBot, which allows users to design and simulate robots.
Conducted a study with 37 participants using BuildaBot to design robots for a specific task.
Suggested that the goal of design systems may not be to find the optimal design but to help users find a design that works.

Discussed the challenges of mechanical design compared to VLSI and computer programming due to the interconnectedness of mechanical components.
Proposed an approach to mechanical design that focuses on modularity and emphasizes the importance of considering what designs are possible and practical rather than solely optimizing for optimality.

Presented a project aimed at designing physically realizable manipulator arms with specified kinematics, using origami-inspired modules that implement specific paths.
Drew inspiration from path planning under curvature constraints, particularly Dubin's path planning, and showed how these paths can be used to create rigid transformations and design manipulator arms.
Demonstrated a method for designing fabricable structures with specified kinematics by converting kinematic specifications into origami patterns, which can then be folded or 3D printed to create physical structures.
Designed a quadrupedal robot with eight joints using the origami-based method, which is lighter and has fewer degrees of freedom than existing quadrupeds.

Focused on modeling soft robots and their interactions with compliant surfaces.
Used a network of springs to model the deformation of soft robots and control their deformation by manipulating the stiffness of these springs.
Extended these ideas beyond body robot design to consider how robots move on compliant surfaces, using signals from the motors in robots to detect ground forces and track joint positions to understand the robot's interaction with the ground.
Designed robot configurations that allow them to navigate over compliant surfaces more reliably.

Spiral zipper and variable topology truss allow robots to connect and apply forces to each other, creating lateral motion and resisting instabilities.
New controller combines reaction forces from the ground and trusses to move robots in sandy environments.
Robot stiffness can be manipulated through P control to modify system dynamics based on ground properties.