There are gazillion different aspects forming a human life as we know it, but when push comes to shove, none is more defining for us than our tendency to improve at a consistent pace. This firm commitment towards getting better, no matter the situation, has enabled the world to clock some huge milestones, with technology emerging as quite a major member of the group. The reason why we hold technology in such a high regard is, by and large, predicated upon its skill-set, which guided us towards a reality that nobody could have ever imagined otherwise. Nevertheless, if we look beyond the surface for one hot second, it will become abundantly clear how the whole runner was also very much inspired from the way we applied those skills across a real world environment. The latter component, in fact, did a lot to give the creation a spectrum-wide presence, and as a result, initiated a full-blown tech revolution. Of course, the next thing this revolution did was to scale up the human experience through some outright unique avenues, but even after achieving a feat so notable, technology will somehow continue to bring forth the right goods. The same has turned more and more evident in recent times, and assuming one new discovery ends up with the desired impact, it will only put that trend on a higher pedestal moving forward.

The researching team at MIT’s Computer Science and Artificial Intelligence Lab has successfully developed a whole new framework called Graphs of Convex Sets (GCS) Trajectory Optimization, which is designed to help the world access a scalable, collision-free motion planning system for a better robotics infrastructure. In order to understand the stakes here, we must acknowledge how, despite all their technological brilliance, our current robotics systems have largely failed to navigate through a complex surrounding on their own. This, in turn, has unsurprisingly limited their application across the board. However, the development in question promises to change that big time. But how does it work? Well, it effectively merges graph search (a method for finding discrete paths in a network) with convex optimization (an efficient method for optimizing continuous variables so that a given cost is minimized) to manufacture a result which can guide robots to find paths through maze-like environments. Not just that, the framework also makes a point to optimize the robot’s overall trajectory. Talking on a slightly deeper level, the first main ingredient bears the responsibility to search for graphs, calculate different properties at each one to find hidden patterns, and identify the shortest path to reach the target. Moving on to convex optimization, GCS banks upon that technology to conceive different trajectories that account for different angles to ensure the robot avoids colliding with the edges of its obstacles. Such attention to detail means the robot can seamlessly squeeze by potential hurdles, maneuvering through each turn the same way a driver avoids accidents on a narrow street.

“Robots excel at repetitive, pre-planned motions in applications such as automotive manufacturing or electronics assembly but struggle with real-time motion generation in novel environments or tasks. “This paper presents a novel approach that has the potential to dramatically enhance the speed and efficiency of robot motions and their ability to adapt to novel environments,” said David M.S. Johnson, Co-founder and CEO of Dexai Robotics.

The researchers have already conducted some initial tests on their brainchild, tests where they observed that the system was able to skillfully guide two robotic arms holding a mug around a shelf, while simultaneously optimizing for the shortest time and path. In subsequent demos, the researchers removed the shelves, and the robots swapped the positions of spray paints and handed each other a sugar box, again displaying sizeable success. Given what they saw during the tests, there is a belief that the framework can be of huge help in sectors like manufacturing, given two robotic arms working in tandem could bring down an item from a shelf. As for how it is any different from all the previous efforts we have seen around this discipline, in simpler terms, those previous models would employ a ‘hub and spoke’ approach, using pre-computed graphs of a finite number of fixed configurations. Now, even though these configurations are safe, they demanded the robot to strictly follow the provided roadmap, something that eventually led to inefficient robot movements. On the other hand, the concept in question takes a completely opposite approach, as it empowers robots to easily adapt to different configurations within pre-computed convex regions.

We referred to some initial tests already conducted on the technology, but we still haven’t acknowledged the ones that involved using GCS in a quadrotor to let it fly through a building without crashing into trees or failing to enter doors and windows at the correct angle. Here, too, the system was able to deliver great efficiency in optimizing the path around the obstacles, all while staying mindful about the quadrotor’s unique dynamics.

For the immediate future, the MIT CSAIL researchers hope to reach a level where GSC can facilitate the robot’s contact with the environment, such as pushing or sliding objects out of the way. Beyond that, the team is also focused on exploring applications of GCS trajectory optimization in regards to robot-led tasks.

“I’m very excited about this application of GCS to motion planning. But this is just the beginning. This framework is deeply connected to many core results in optimization, control, and machine learning, giving us new leverage on problems that are simultaneously continuous and combinatorial,” said Russ Tedrake, MIT Professor, CSAIL Principal Investigator, and co-author on a new paper about the work. “There is a lot more work to do.”