RoboGrammar: Computer-Aided creativity in robot design


There are various ways to find out which robot design is most efficient for crossing various terrains. It is time-consuming to build every variation required for a robot and test them in the real world. RoboGrammer is a computer-aided system developed by MIT.

This system can test various robot designs virtually to determine which is best for crossing particular types of terrain. You can tell the RoboGrammer what parts are lying around your shop – joints or wheels and then it generates an optimized structure and control program for your robot.

Even though Robotic is a mechanical thing it still needs manual processing. RoboGrammar is becoming a way to make robot designs that could potentially be more effective. Robots are built to perform endless tasks yet they look similar in their overall design and shape. RoboGrammar helps create a more innovative design that could improve the functionality.

Initially, Robots are just a structure, selecting the best robot design is about controlling each robot’s movements and assessing their function. The controller and shape of a robot are interconnected, that is why the optimization of the robot is required. Once each Robot is dissembled to move freely the researchers look at high-performing robots with a “graph heuristic search”. This whole process improves as time passes.

This whole work is the result of great achievements in the 25-year quest to automatically design the morphology and control of robots.  According to Allan Zhao, RoboGrammar is a way towards, more inventive robot designs that could be more effective.

Zhao and his team develop something called Graph Grammar; it is a set of constraints for the arrangements of the robot’s component. This was developed to overcome the challenges faced when connecting the parts in arbitrary ways. The idea of rules in graph grammar was inspired by animals-arthropods in particular. RoboGrammar uses this graph grammar and set off in three steps firstly defines the problem, secondly, draw up possible robotic solutions, and then finally select the optimal ones.

With these steps and also using the rules of graph grammar RoboGrammar design hundreds of thousands of potential robot structures. Some look like a race car, and some like spiders. 


Please enter your comment!
Please enter your name here