Persistent Navigation and Mapping using a Biologically Inspired SLAM System

Abstract

The challenge of persistent navigation and mapping is to develop an autonomous robot system that can simultaneously localize, map and navigate over the lifetime of the robot with little or no human intervention. Most solutions to the simultaneous localization and mapping (SLAM) problem aim to produce highly accurate maps of areas that are assumed to be static. In contrast, solutions for persistent navigation and mapping must produce reliable goal-directed navigation outcomes in an environment that is assumed to be in constant flux. We investigate the persistent navigation and mapping problem in the context of an autonomous robot that performs mock deliveries in a working office environment over a two-week period. The solution was based on the biologically inspired visual SLAM system, RatSLAM. RatSLAM performed SLAM continuously while interacting with global and local navigation systems, and a task selection module that selected between exploration, delivery, and recharging modes. The robot performed 1,143 delivery tasks to 11 different locations with only one delivery failure (from which it recovered), traveled a total distance of more than 40 km over 37 hours of active operation, and recharged autonomously a total of 23 times.

Environmental stability modulates the role of path integration in human navigation

“…Many artificial or robotic navigation systems are inspired by findings about biological navigation system (Hübner & Mallot, 2007;Kuipers, 2000;Milford & Wyeth, 2010;Trullier, et al, 1997). Like human and animal navigators, these systems face the same problems of acquiring spatial knowledge about the environment and localizing/orienting themselves, and attempt to solve them by combining path integration or "odometry", vision-based navigation, and other strategies.…”

“…Like human and animal navigators, these systems face the same problems of acquiring spatial knowledge about the environment and localizing/orienting themselves, and attempt to solve them by combining path integration or "odometry", vision-based navigation, and other strategies. Because most artificial navigation systems assume the environment is stable (but see Milford & Wyeth, 2010), they do not consider back-up system and reference system functions for path integration. However, they often use visual input to recalibrate the path integrator, which accumulates errors due to odometry noise.…”

Semi-task-dependent and uncertainty-driven world model maintenance

“…In the field of persistent simultaneous localization and mapping (SLAM) the aim is to allow an autonomous robot to simultaneously localize, map and navigate despite (visual) changes in the real world (Milford and Wyeth 2010). Rat-SLAM (Milford and Wyeth 2010) uses an 'experience map' and adds experiences if the robot visits new locations or if previously observed places have a new visual appearance.…”

“…Rat-SLAM (Milford and Wyeth 2010) uses an 'experience map' and adds experiences if the robot visits new locations or if previously observed places have a new visual appearance. In Churchill and Newman (2012) the map is extended if localization in a previously visited area fails given previous experiences, e.g., due to a different visual appearance caused by changing weather conditions, and (Konolige and Bowman 2009) use similar ideas to create lifelong visual maps which are maintained online at a frame rate of 30 Hz.…”

“…In Churchill and Newman (2012) the map is extended if localization in a previously visited area fails given previous experiences, e.g., due to a different visual appearance caused by changing weather conditions, and (Konolige and Bowman 2009) use similar ideas to create lifelong visual maps which are maintained online at a frame rate of 30 Hz. Because of their focus on maps used for localization and navigation instead of semantic object maps (Milford and Wyeth 2010;Churchill and Newman 2012;Konolige and Bowman 2009), and the references therein, are considered to be complementary to the proposed approach rather than being alternatives. In Ekvall et al (2007), knowledge about object poses is incorporated in a SLAM approach, however, due to the computational complexity only ten objects are tracked.…”