Work

Visual robot navigation incorporating causal models in deep reinforcement learning

Year: 2024
Type: conference-paper
Abstract: Deep reinforcement learning can be used for visual navigation in mobile robots but needs substantial computational resources and long training time. To reduce exploration time and improve adaptation to novel situations or environments, we created an algorithm to learn a Causal Bayesian Network of the task and use it in a Deep Reinforcement Learning algorithm (DQN). Experimentally, we show its effectiveness in simulated environments for a visual robot navigation task.
Source: HAL
Language: English
Open Access status: green
OpenAlex ID: W4403241271
FWCI: 0.081
Cites: 26
Cited by: 1
Related to: 10