Proprioceptive Localization for Robots
Loading...
Date
2021-04-07
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Localization is a critical navigation function for mobile robots. Most localization methods employ a global position system (GPS), a lidar, and a camera which are exteroceptive sensors relying on the perception and recognition of landmarks in the environment. However, GPS signals may be unavailable because high-rise buildings may block GPS signals in urban areas. Poor weather and lighting conditions may challenge all exteroceptive sensors. In this dissertation, we focus on proprioceptive localization (PL) methods which refer to a new class of robot egocentric localization methods that do not rely on the perception and recognition of external landmarks. These methods depend on a prior map and proprioceptive sensors such as inertial measurement units (IMUs) and/or wheel encoders which are naturally immune to aforementioned adversary environmental conditions that may hinder exteroceptive sensors. PL is intended to be a low-cost and fallback solution when everything else fails.
We first propose a method named proprioceptive localization assisted by magnetoreception (PLAM). PLAM employs a gyroscope and a compass to sense heading changes and matches the heading sequence with a pre-processed heading graph to localize the robot. Not all cases can be successful because degenerated maps may consist of rectangular grid-like streets and the robot may travel in a loop. To analyze these, we use information entropy to model map characteristics and perform both simulation and experiments to find out typical heading and information entropy requirements for localization.
We further propose a method which allows continuous localization and is less limited by map degeneracy. Assisted by magnetoreception, we use IMUs and wheel encoders to estimate vehicle trajectory which is used to query a prior known map to obtain location. We named the proposed method as graph-based proprioceptive localization (GBPL). As a robot travels, we extract a sequence of heading-length values for straight segments from the trajectory and match the sequence with a pre-processed heading-length graph (HLG) abstracted from the prior known map to localize the robot under a graph-matching approach. Using HLG information, our location alignment and verification module compensates for trajectory drift, wheel slip, or tire inflation level.
%The algorithm runs successfully in finding robot location continuously and achieves localization accuracy at the level that the prior map allows (less than 10m).
With the development of communication technology, it becomes possible to leverage vehicle-to-vehicle (V2V) communication to develop a multiple vehicle/robot collaborative localization scheme. Named as collaborative graph-based proprioceptive localization (C-GBPL), we extract heading-length sequence from the trajectory as features. When rendezvousing with other vehicles, the ego vehicle aggregates the features from others and forms a merged query graph. We match the query graph with the HLG to localize the vehicle under a graph-to-graph matching approach. The C-GBPL algorithm significantly outperforms its single-vehicle counterpart in localization speed and robustness to trajectory and map degeneracy.
Besides, we propose a PL method with WiFi in the indoor environment targeted at handling inconsistent access points (APs). We develop a windowed majority voting and statistical hypothesis testing-based approach to remove APs with large displacements between reference and query data sets. We refine the localization by applying maximum likelihood estimation method to the closed-form posterior location distribution over the filtered signal strength and AP sets in the time window. Our method achieves a mean localization error of less than 3.7 meters even when 70% of APs are inconsistent.
Description
Keywords
Autonomous vehicle navigation, localization, sensor fusion, multi-robot systems