Uncertainty guided method improves robot object navigation in new places
Have You Seen It Clearly? Uncertainty-Guided Belief Refinement for Zero-Shot Object Navigation
Robotics
Summary
Finding objects in places a robot has never been before is hard because the robot doesn't know what to expect. The authors propose a method that helps the robot keep track of how sure it is about where the object might be, and to update its beliefs when new information conflicts with old information. This method helps the robot decide when to keep exploring or revisit places it has been before to find objects more efficiently. They tested this on simulated home environments and a real robot, showing better success without special training on those new objects or places.
What this means in practice
- •For robotics engineers: Improve autonomous robots’ ability to find objects in new, unfamiliar indoor environments without retraining for each scenario.
- •For warehouse operators: Enhance robots that navigate warehouses to locate misplaced inventory items more reliably by revisiting stored knowledge with updated certainty assessments.
Authors
Jingzhi Huang, Junkai Huang, Yuhang Peng, Keyu Mu, Jiaru Zhong, Yi Wang, Lujia Wang, Lap-Pui Chau, Siyang Gao
Abstract
Zero-Shot Object Navigation (ZSON) requires an agent to locate target objects in unfamiliar environments without additional training for the test categories or environments. Recent approaches leverage large language models and vision-language models to construct spatial representations and guide exploration. However, whether the available evidence sufficiently supports a navigation choice, and how new evidence changes the search value of previously visited locations, remain insufficiently addressed. We propose RefBeliefMap, an uncertainty-guided mapping framework that distinguishes target-location belief from spatial and semantic evidence uncertainty. By jointly modeling intra-observation uncertainty and inter-observation disagreement over a sliding memory window, RefBeliefMap assesses the reliability of both evidence channels. When competing frontiers remain insufficiently resolved, these uncertainty estimates guide active observation and semantic reasoning to refine the evidence supporting target-location beliefs before navigation commitment. Building on the refined map, we introduce RefBeliefNav, which jointly evaluates unexplored frontiers and historical waypoints according to their target-finding potential, expected information gain, and navigation cost. This unified objective allows historical locations to be reconsidered as new evidence changes their search value, coordinating exploration and revisitation. Experiments on HM3D and MP3D demonstrate state-of-the-art performance among the evaluated zero-shot baselines, with ablation studies examining the contributions of modules. Real-world experiments on a Unitree Go2 quadruped robot further demonstrate the sim-to-real transferability of our method.