Paper: A Software Framework for Planning under Partial Observability

September 25th, 2018

IROS 2018 paper: Paper: A Software Framework for Planning under Partial Observability

Planning under partial observability is both challenging and critical for reliable robot operation.

The past decade has seen substantial advances in this domain: The mathematically principled approach for addressing such problems, namely the Partially Observable Markov Decision Process (POMDP), has started to become practical for various robotics tasks.

Good approximate solutions for problems framed as POMDPs can now be computed on-line, with a few classes of problems being solved in near real-time.

However, applications of these more recent advances are often hindered by the lack of easy-to-use software tools. Implementation of state of the art algorithms exist, but most (if not all) require the POMDP model to be hard-coded inside the program, increasing the difficulty of applying them.

To alleviate this problem, we propose a software toolkit, called On-line POMDP Planning Toolkit (OPPT) (downloadable from˜oppt ).

By providing a well-defined and general abstract solver API, OPPT enables the user to quickly implement new POMDP solvers.

Furthermore, OPPT provides an easy-to-use plug-in architecture with interfaces to the highfidelity simulator Gazebo that, in conjunction with user-friendly configuration files, allows users to specify POMDP models of a standard class of robot motion planning under partial observability problems with no additional coding effort.

Click here to download the paper.

For more information, contact PhD Student Marcus Hoerger on marcus.hoerger [at]

Marcus Hoerger1; Hanna Kurniawati, Alberto Elfes “A Software Framework for Planning under Partial Observability” in IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), Madrid, Spain, October, 2018.