#probabilistic robotics

[towardsrobotics-blog]

Probabilistic Robotics: Introduction - Towards Robotics

Probabilistic robotics is a new and expanding place in robotics, involved with understanding and control in the face of uncertainty.

Probabilistic robotics is a new and expanding place in robotics, involved with understanding and control in the face of uncertainty.

[makokal-blog]

OpenPRM::Updates

OpenPRM has been updated significantly, some of the changes include but not limited to:

New Collision Checkers based on the Binary approach for faster collision checking

Addition of Halton Sampler known to generate significantly smaller number of samples compared to Uniform sampling for a similar scenario.

A new Spatial structure using Boost Graph Library to enable extensibility and ease of use of various Graph Algorithms

Improvements to the Visibility PRM algorithms implementation, showing a significant reduction in time taken and number of nodes in the roadmap for the same C-Space coverage

Introduction of Task Space Regions(TSR) for representing constraints in the Goal Configurations

[makokal-blog]

Official::OpenPRM Released

It is my pleasure to announce the official release of OpenPRM, open source implementation of Probabilistic Roadmap planners. Currently release has the following planners;

SBL - Single Query Bi-Directional PRM Planner with Lazy collision checking

VisibilityPRM - Visibility based Probability Roadmap planning

ClassicPRM - Classical PRM planner for high dimensional spaces

The initial release does not include path optimizers which are scheduled for the next release among other features. Find out more about OpenPRM here.

[makokal-blog]

openPP

openPP is an open source probabilistic motion planning plugin that I am currently developing for later integration into openrave. It will be based primarily on probabilistic roadmap methods starting with basic PRM and SBL variants. Since it is opensource, I will soon be calling one more people interested in the same to join into the project once I have something major running.

[makokal-blog]

PRMs for Motion Planning

Motion Planning has been dominated by algorithms that lie towards RRTs and related derivatives such as BiRRTs. Openrave for instance comes with RRT planners which do the job just a nuce as one needs it. PRMs(Probabilistic RoadMaps) however remain under-exploited even though they pose a great potential in motion planning owing to the probabilistic nature and the fact that classical graph algorithms can be optimized to be used in the stages of PRMs. I have now started looking at PRMs in deeper perspective with the aim of getting a better understanding of how they work and how they are implemented. For more on PRMs, check this out.