Revisiting Sparse Rewards for Goal-Reaching Reinforcement Learning

Fuente: arXiv
Guardado en:
Detalles Bibliográficos
Autores principales: Vasan, Gautham, Wang, Yan, Shahriar, Fahim, Bergstra, James, Jagersand, Martin, Mahmood, A. Rupam
Formato: Preprint
Publicado: 2024
Materias:
Acceso en línea:
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866910518466314240
author Vasan, Gautham
Wang, Yan
Shahriar, Fahim
Bergstra, James
Jagersand, Martin
Mahmood, A. Rupam
author_facet Vasan, Gautham
Wang, Yan
Shahriar, Fahim
Bergstra, James
Jagersand, Martin
Mahmood, A. Rupam
contents Many real-world robot learning problems, such as pick-and-place or arriving at a destination, can be seen as a problem of reaching a goal state as soon as possible. These problems, when formulated as episodic reinforcement learning tasks, can easily be specified to align well with our intended goal: -1 reward every time step with termination upon reaching the goal state, called minimum-time tasks. Despite this simplicity, such formulations are often overlooked in favor of dense rewards due to their perceived difficulty and lack of informativeness. Our studies contrast the two reward paradigms, revealing that the minimum-time task specification not only facilitates learning higher-quality policies but can also surpass dense-reward-based policies on their own performance metrics. Crucially, we also identify the goal-hit rate of the initial policy as a robust early indicator for learning success in such sparse feedback settings. Finally, using four distinct real-robotic platforms, we show that it is possible to learn pixel-based policies from scratch within two to three hours using constant negative rewards.
format Preprint
id arxiv_https___arxiv_org_abs_2407_00324
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Revisiting Sparse Rewards for Goal-Reaching Reinforcement Learning
Vasan, Gautham
Wang, Yan
Shahriar, Fahim
Bergstra, James
Jagersand, Martin
Mahmood, A. Rupam
Robotics
Machine Learning
Many real-world robot learning problems, such as pick-and-place or arriving at a destination, can be seen as a problem of reaching a goal state as soon as possible. These problems, when formulated as episodic reinforcement learning tasks, can easily be specified to align well with our intended goal: -1 reward every time step with termination upon reaching the goal state, called minimum-time tasks. Despite this simplicity, such formulations are often overlooked in favor of dense rewards due to their perceived difficulty and lack of informativeness. Our studies contrast the two reward paradigms, revealing that the minimum-time task specification not only facilitates learning higher-quality policies but can also surpass dense-reward-based policies on their own performance metrics. Crucially, we also identify the goal-hit rate of the initial policy as a robust early indicator for learning success in such sparse feedback settings. Finally, using four distinct real-robotic platforms, we show that it is possible to learn pixel-based policies from scratch within two to three hours using constant negative rewards.
title Revisiting Sparse Rewards for Goal-Reaching Reinforcement Learning
topic Robotics
Machine Learning
url https://arxiv.org/abs/2407.00324