Home Page - Approximate POMDP Planning Software

Approximate POMDP Planning Software

APPL is a C++ implementation of the SARSOP algorithm [1] for solving discrete POMDPs. It uses the factored MOMDP representation [2] for computational efficiency. APPL takes as input a POMDP model in the POMDP or POMDPX file format and produces a policy file. It also contains a simple simulator for evaluating the quality of the computed policy. More information can be found at here. For bug reports and suggestions, please email motion@comp.nus.edu.sg

H. Kurniawati, D. Hsu, and W.S. Lee. SARSOP: Efficient point-based POMDP planning by approximating optimally reachable belief spaces. In Proc. Robotics: Science and Systems, 2008.
S.C.W. Ong, S.W. Png, D. Hsu, and W.S. Lee. POMDPs for robotic tasks with mixed observability. In Proc. Robotics: Science and Systems, 2009.

Monte Carlo Value Iteration (MCVI) for POMDP

MCVI is a C++ implementation of the MCVI algorithm [1], combined with macro-actions [2]. It takes as input a POMDP model (coded in C++) and produce a policy file. It also contains a simple simulator for evaluating the quality of the computed policy. For bug reports and suggestions, please email motion@comp.nus.edu.sg.

H. Bai, D. Hsu, M. Kochenderfer, and W.S. Lee. Unmanned Aircraft Collision Avoidance using Continuous-State POMDPs. Proc. Robotics: Science and Systems, 2011.
Z.W. Lim, D. Hsu, and W.S. Lee. Monte Carlo Value Iteration with Macro-Actions. Proc. Neural Information Processing Systems (NIPS), 2011.

News

Our team won the POMDP track at the ICAPS 2011 International Probabilistic Planning Competition (IPPC). The source code used can be found here. We used SARSOP together with POMCP, a UCT based Monte Carlo algorithm in the competition. We will be implementing POMCP in APPL soon - look out for that.