Document Type

Working Paper

Date of this Version



Pratik Chaudhari


An agent performing Simultaneous Localization and Mapping (SLAM) constructs a map of the environment while estimating its location at the same time. SLAM algorithms primarily focus on finding the best estimate of the location of the agent, and by extension, that of the landmarks in the environment, given observations from sensors. These algorithms do not typically address how an agent should explore an unknown environment to build a map efficiently. This ability for active exploration is important for autonomous robots to work in unknown, unstructured environments such as forests or caves.

This paper proposes an active SLAM system that allows an agent to explore its surroundings, using visual-inertial data from an RGBD camera. We formalize this problem as taking actions that maximize the amount of information obtained from the scene. At each time step, a utility function that computes the incremental information gain is used to take actions. We conduct experiments using an Intel RealSense camera mounted on a custom-built quadrotor and show that we can explore indoor environments (while restricting the actions to choosing new viewpoints in SO(3) for practical reasons).


Active Exploration, Simultaneous Localization and Mapping, Autonomous Robot, Quadrotor



Date Posted: 26 September 2022


To view the content in your browser, please download Adobe Reader or, alternately,
you may Download the file to your hard drive.

NOTE: The latest versions of Adobe Reader do not support viewing PDF files within Firefox on Mac OS and if you are using a modern (Intel) Mac, there is no official plugin for viewing PDF files within the browser window.