Shushman Choudhury

Robotics Institute, CMU

Theses

Anytime Geometric Motion Planning on Large Dense Roadmaps [PDF]
MS Thesis, Robotics Institute, Carnegie Mellon University, July 2017

Publications

Incorporating Qualitative Information into Quantitative Estimation via Sequentially Constrained
Hamiltonian Monte Carlo Sampling
[PDF]
Daqing Yi, Shushman Choudhury and Siddhartha Srinivasa
IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) 2017 (To Appear)

Densification Strategies for Anytime Motion Planning over Large Dense Roadmaps [arXiv]
Shushman Choudhury, Oren Salzman, Sanjiban Choudhury and Siddhartha Srinivasa
IEEE International Conference on Robotics and Automation (ICRA) 2017

Pareto-Optimal Search over Configuration Space Beliefs for Anytime Motion Planning [PDF]
Shushman Choudhury, Christopher Dellin, Siddhartha Srinivasa
IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) 2016

A System for Multi-Step Mobile Manipulation: Architecture, Algorithms, and Experiments [PDF]
Siddhartha Srinivasa et al.
International Symposium on Experimental Robotics 2016

Currency Recognition on Mobile Phones [PDF]
Suriya Singh, Shushman Choudhury, Kumar Vishal, C. V. Jawahar
International Conference on Pattern Recognition (ICPR). IEEE, 2014

Research (Past)

My MS thesis proposed an algorithmic framework for efficient anytime geometric motion planning on large and dense roadmaps. These structures are capable of solving a wide range of motion planning problems and allow for important pre-processing benefits. However, the size of the roadmap graph creates difficulties for existing approaches to graph-based planning algorithms. We explore two key ideas to deal with this.

Firstly, we frame the problem of anytime planning on roadmaps as one of searching for the shortest path over a sequence of subgraphs of the entire roadmap graph. We study the space of subgraphs and formulate densification strategies to traverse the space of subgraphs, all the way to the complete roadmap graph. Secondly, we develop an anytime roadmap planning algorithm, that is efficient with respect to collision checks, for searching each subgraph generated by the densification strategy. This algorithm searches for paths that are Pareto-optimal in path length and collision probability (obtained from some belief model) and adjusts the tradeoff to find successively shorter feasible paths.

We analyze and implement our framework and show favourable performance with respect to current approaches to anytime motion planning. Please see my papers at IROS 2016 and ICRA 2017 for the individual ideas and we have an upcoming journal paper where we present the combined framework.