Waypoint Planning Networks (WPN)

Alexandru-Iosif Toma*, Hussein Ali Jaafar**, Hao-Ya Hsueh**, Stephen James*, Daniel Lenton*, Ronald Clark*, Sajad Saeedi**

* Imperial College London ** Ryerson University

What is Waypoint Planning Networks (WPN)

Waypoint Planning Networks, or WPN, is a hybrid motion planning algorithm based on LSTMs with a local kernel, a classic algorithm such as A*, and a global kernel using a learned algorithm. WPN produces a more computationally efficient and robust solution than other learned approaches. WPN is open source, see the GitHub repo for more a more technical tutorial. [Github]


With the rapid development of machine learning (ML) algorithms, planning algorithms are also evolving; however, the learned path planning algorithms often have difficulty competing with success rates of classic algorithms. To solve this, we propose waypoint planning networks, which uses a hybrid solution; a classic local kernel, and a ML global kernel. Additionally, classic algorithms such as A* are unable to plan into unknown environments, or with partial knowledge of the maps without additional exploration packages. WPN is able to plan into unknown environments. See the supplementary video for demonstrations.

Furthermore, WPN is also able to compete with A*'s high success rates, but it is able to outperform in terms of search space. This is outlined in the submitted paper, and supplementary video.


If you have any questions, feel free to reach out to us at the following email: rcvl@ryerson.ca