For most hobbyists, building a robot mostly involves buying a robot and assembling it and then (if that first part doesn’t make you hate robots) programming it to do stuff. Designing your own robot from scratch is much more difficult, especially if it’s a robot that has legs that are supposed to do something practical. ETH Zurich, in collaboration with Disney Research and CMU, has developed “an interactive design system that allows casual users to quickly create 3D-printable robotic creatures.” From walking bipeds to salamanders with actuated spines to quintapeds (that’s a thing, right?), the software does all of the hard work for you, and with just a 3D printer and some servos, you can design robots that are exactly as bizarre as you want them to be.
This approach is based on how easy it is to design digital characters by pointing, clicking, and dragging. Making physical robots this way isn’t nearly as easy, of course, because it involves dealing with minor annoyances like whether motions are physically possible. There also needs to be some consideration of cost and complexity, especially if the fabrication process is based on a consumer 3D printer. Disney’s method involves a combination of fabrication-oriented design, physical character design, motion planning, and of course robotics.
From the look of things, using the software is alarmingly easy. Each robot starts with an initial skeletal structure with bones connected by virtual motors placed at each joint position. This is freely editable, and you can click to add or subtract motors or alter their orientation to change the structure of the robot.
To get the robot walking, you can adjust which legs are on the ground when, while keeping your robot from falling over simply by making sure that the green ball representing the center of mass of the robot stays within the red box representing the stability polygon created by whichever legs are in contact with the ground. More detailed gait customization can be applied as well, including directionality, speed, turning rates, and individual feet trajectories. Behind the scenes, the software deals with all of the complicated stuff, optimizing the motor values to yield dynamically stable motions that can then be previewed in a physics-based simulation.
Once you’re happy with how things look, the final step is to generate 3D geometry for all of your robot’s body parts, including motor connectors. The software takes into account what kind of printer you’re using and what materials you’ve chosen: for example, if you’re using a MakerBot or something else with a filament material, the software will use infill for strength, but if you have a laser sintering setup, the software will change to a much more efficient truss structure.
To test out their software, the researchers build themselves a few different robots from scratch (including a five-legged robot called Predator) using 3D printed parts, Dynamixel MX-28s, a servo controller board, and a battery. Predictably, the real world performance of the robots varied somewhat from the (by definition simplified) physical simulation: things like friction, slightly bendy 3D printed body parts, and actuators that don’t respond like an ideal actuator should all lead to small but cumulative amounts of uncertainty. However, the researchers “observed good agreement between the overall motions of our physical prototypes and the behavior predicted in simulation.”
There are a few things that the software can’t handle right now. One is any kind of flight phase, so no running robots. Another is sensor feedback, which could certainly offer significant improvements in adaptive motion planning. From the sound of things, though, the researchers are more interested in enabling more options for creativity on the part of the end user:
We are excited about the challenge of making the process of authoring complex behaviors easily accessible to casual users. We are encouraged by the ease with which gaits and motion styles can be specified using our easy-to-use editing tools. We plan to extend these tools such that users can specify a motion repertoire that includes switching between gaits, portraying rich personalities and interacting appropriately with objects and humans. Finding appropriate abstractions for intuitively authoring such high-level behaviors is an interesting subject for future work.
“Interactive Design of 3D-Printable Robotic Creatures,” by Vittorio Megaro, Bernhard Thomaszewski, Maurizio Nitti, Otmar Hilliges, Markus Gross, and Stelian Coros from ETH Zurich, Disney Research Zurich, and Carnegie Mellon University was presented at SIGGRAPH Asia 2015.
[ Disney Research ]