RSS 2023 Recap

Last week, I attended Robotics: Science and Systems in Daegu, South Korea. While there, I presented our paper, Non-Euclidean Motion Planning with Graphs of Geodesically-Convex Sets. This was my first experience presenting at a single-track conference, and giving a talk in front of hundreds of people was pretty anxiety-inducing! But I think the talk and poster session went very well. You can view the poster here, and I've clipped a recording of the talk from the RSS livestream here.

There were several papers at the conference that I found particularly interesting. Path Planning for Multiple Tethered Robots Using Topological Braids presents a group-theoretic framework for tethered quadrotors, which may be in a knotted configuration. It would have been very nice to know about their work when I was working on my final project with Seiji Shaw for Underactuated this past spring. In that project, we ended up making the simplifying (and somewhat unrealistic) assumption that the cables would not tangle.

Another paper that stood out to me was Motion Planning (In)feasibility Detection using a Prior Roadmap via Path and Cut Search. This paper discussed roadmap-based planning, where it's unknown at planning-time whether or not an edge is valid. They either present an optimal path from the start to the goal (with respect to all paths in the graphs), or a certificate of infeasibility, in the form of a graph cut that separates the start and the goal.

As an overall trend, there were many papers utilizing diffusion policies, and I got the chance to discuss one of these papers with its authors. Among other things, they emphasized that their use of a continuous-time formulation for diffusion was significantly better than discrete-time approximations. I'm far from an expert on diffusion policies, but I feel it's worthwhile to pass that advice on. (The code from that paper can be found here.)

Finally, I got the chance to discuss G*: A New Approach to Bounding Curvature Constrained Shortest Paths through Dubins Gates with one of its authors. This type of graph problem with neighborhoods feels spiritually similar to GCS, and we had a lot to talk about. Perhaps a convex decomposition could yield "gates" that yield a tighter bound on the optimal path length? Maybe there's a way to plan Dubins paths through such gates? It's definitely something both of us will be thinking about.

Now, I'm back in Cambridge (although my circadian rhythm still thinks it's in Daegu). The research direction we presented on at RSS is continuing, so maybe I'll be able to post a preview sometime soon.

Comments

Post a Comment

Popular posts from this blog

Reimplementing GCS (Shortest Paths in Graphs of Convex Sets)

Image
In a previous blog post , I discussed my experiences reimplementing Iterative Regional Inflation by Semidefinite programming (IRIS) , which is an algorithm for finding large convex regions of collision free space (C-Free). In this blog post, I'll be discussing my experiences reimplementing shortest paths in Graphs of Convex Sets (GCS) , an algorithm for motion planning when the configuration space is described as the union of convex sets. Disclaimer: the intent is not to build a codebase for future use, since the Drake toolbox has mature implementations of both IRIS and GCS . Disclaimer: if you have not read the previous blog post , you should do so before reading this one. In this blogpost, I'll give a brief overview of the GCS paper, and then discuss my experiences reimplementing it. Before reading this blog post, you should read my previous blog post about reimplementing IRIS, as it provides some background and context which I will not repeat in this post. I have also ma
Read more

Reimplementing IRIS (Computing Large Convex Regions of Obstacle-Free Space Through Semidefinite Programming)

Image
I've just begun my studying towards my PhD in computer science at MIT, and am working in the Robot Locomotion Group . This group has developed a series of projects that use formal mathematical optimization for tasks in motion planning. Because I'm interested in this area of research, I've been reading several of their papers; to better understand these papers, I've endeavored to reimplement some of their work. The two specific algorithms I'm interested in are close companions: Iterative Regional Inflation by Semidefinite programming (IRIS) can be used to find large convex regions of collision free space, and shortest paths in Graphs of Convex Sets (GCS) can be used for motion planning when the configuration space is described as the union of convex sets. I've attempted to reimplement these algorithms for simple, toy problems. Disclaimer: the intent is not to build a codebase for future use, since the Drake toolbox has mature implementations of both IRIS and
Read more

Defining Geodesic Convexity

In my previous blog post , I discussed the notion of geodesic convexity , which is a generalization of ordinary convexity in Euclidean space to Riemannian manifolds. This is essential for disciplined optimization over manifolds; hence, it's very relevant to my research in optimization-based motion planning over manifolds. Unfortunately, in the current literature, there are various definitions of geodesic convexity, with key differing properties. This is chiefly because if a geodesic connecting two points exists (it may not exist), it may not be unique, and it may not be minimizing. In this brief blog post, I'll discuss the various definitions that are used, and then explain why we use the one we did in our GGCS paper . The Definitions Nicholas Boumal's textbook An introduction to optimization on smooth manifolds is an excellent reference on the subject, and he discusses the various definitions of geodesic convexity in Section 11.3. The first definition of geodesic convexi
Read more