Task-Oriented Active Perception and Planning in Environments with Partially Known Semantics

Mahsa Ghasemi, Erdem Bulgur, Ufuk Topcu
Proceedings of the 37th International Conference on Machine Learning, PMLR 119:3484-3493, 2020.

Abstract

We consider an agent that is assigned with a temporal logic task in an environment whose semantic representation is only partially known. We represent the semantics of the environment with a set of state properties, called \emph{atomic propositions} over which, the agent holds a probabilistic belief and updates it as new sensory measurements arrive. The goal is to design a joint perception and planning strategy for the agent that realizes the task with high probability. We develop a planning strategy that takes the semantic uncertainties into account and by doing so provides probabilistic guarantees on the task success. Furthermore, as new data arrive, the belief over the atomic propositions evolves and, subsequently, the planning strategy adapts accordingly. We evaluate the proposed method on various finite-horizon tasks in planar navigation settings where the empirical results show that the proposed method provides reliable task performance that also improves as the knowledge about the environment enhances.

Cite this Paper


BibTeX
@InProceedings{pmlr-v119-ghasemi20a, title = {Task-Oriented Active Perception and Planning in Environments with Partially Known Semantics}, author = {Ghasemi, Mahsa and Bulgur, Erdem and Topcu, Ufuk}, booktitle = {Proceedings of the 37th International Conference on Machine Learning}, pages = {3484--3493}, year = {2020}, editor = {III, Hal Daumé and Singh, Aarti}, volume = {119}, series = {Proceedings of Machine Learning Research}, month = {13--18 Jul}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v119/ghasemi20a/ghasemi20a.pdf}, url = {https://proceedings.mlr.press/v119/ghasemi20a.html}, abstract = {We consider an agent that is assigned with a temporal logic task in an environment whose semantic representation is only partially known. We represent the semantics of the environment with a set of state properties, called \emph{atomic propositions} over which, the agent holds a probabilistic belief and updates it as new sensory measurements arrive. The goal is to design a joint perception and planning strategy for the agent that realizes the task with high probability. We develop a planning strategy that takes the semantic uncertainties into account and by doing so provides probabilistic guarantees on the task success. Furthermore, as new data arrive, the belief over the atomic propositions evolves and, subsequently, the planning strategy adapts accordingly. We evaluate the proposed method on various finite-horizon tasks in planar navigation settings where the empirical results show that the proposed method provides reliable task performance that also improves as the knowledge about the environment enhances.} }
Endnote
%0 Conference Paper %T Task-Oriented Active Perception and Planning in Environments with Partially Known Semantics %A Mahsa Ghasemi %A Erdem Bulgur %A Ufuk Topcu %B Proceedings of the 37th International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2020 %E Hal Daumé III %E Aarti Singh %F pmlr-v119-ghasemi20a %I PMLR %P 3484--3493 %U https://proceedings.mlr.press/v119/ghasemi20a.html %V 119 %X We consider an agent that is assigned with a temporal logic task in an environment whose semantic representation is only partially known. We represent the semantics of the environment with a set of state properties, called \emph{atomic propositions} over which, the agent holds a probabilistic belief and updates it as new sensory measurements arrive. The goal is to design a joint perception and planning strategy for the agent that realizes the task with high probability. We develop a planning strategy that takes the semantic uncertainties into account and by doing so provides probabilistic guarantees on the task success. Furthermore, as new data arrive, the belief over the atomic propositions evolves and, subsequently, the planning strategy adapts accordingly. We evaluate the proposed method on various finite-horizon tasks in planar navigation settings where the empirical results show that the proposed method provides reliable task performance that also improves as the knowledge about the environment enhances.
APA
Ghasemi, M., Bulgur, E. & Topcu, U.. (2020). Task-Oriented Active Perception and Planning in Environments with Partially Known Semantics. Proceedings of the 37th International Conference on Machine Learning, in Proceedings of Machine Learning Research 119:3484-3493 Available from https://proceedings.mlr.press/v119/ghasemi20a.html.

Related Material