Collective Robot Construction of Discrete Lattice Structures
focuses on robotic construction of discrete lattice structures. Once we
have individual robotic construction, we will want to look at
collaborative construction with multiple robots.
The goal of this project will be to
develop a tool for evaluating various approaches for robot construction of
lattice structures of arbitrary scale and shape.
what is the goal?
This tool will (for now) not zoom in
to the level of individual voxel placement, rather, it will focus on
global geometry strategies as it relates to path planning for multi-robot
Initial experiments will be
constrained to 2D geometries, with the final goal of addressing 3D
how will you evaluate progress?
Hierarchy can be introduced, which
can lead to higher performance, but also higher part count and geometric
what is the prior art?
Collective robot behavior is a well-researched topic. Collective
robot construction, less so.
It looks like a gradient descent algorithm could work, but would likely
need iterative sensitivity updating to prevent local minima from creating
what techniques will you use?
what questions will you need to address?
-Interesting things to address include
-centralized v decentralized
-dynamics of robots (with mass)
walking around on the structure as its being built
progress update (3/23)
thinking about the swarm construction environment/simulator, here's what
the GUI could look like
have recently been interested in the dynamics of the structure and the
robot as it builds the structure. this is a fairly unexplored topic.
most climbing robots consider their structure to be A) rigid, and B)
For space construction, both (A)
and (B) aren't necessarily true. Most space structures are optimized for
mass, and therefore at large scale become flexible. Further, all space
structures are not necessarily fixed. They are technically floating in
microgravity, and their orientaiton, position, and direction of travel
is governed by spaceship controls (ie: propulsion).
Therefore, a second topic would
look at these (comparatively) smaller scale interactions. While (A) is
indeed interesting, it will be more tractable to look at (B) and only
consider rigid body motions. The design tool/simulator would then
investigate how a robot with much greater mass than the structure it is
building would alter the center of gravity (CG) of the structure, and
how that would result in rotation of the structure in 2D (for now)
Here's a sketch: