CMSC828THW1 and CMSC426 (Fall 2018) GTSAM Toy Problem Code

GTSAM Toy Problem <br> Authors: Nitin J. Sanket and Chahat Deep Singh.

What?

This is a Toy Example to perform SLAM (Simultaneous Localization and Mapping) on a 2D Map. The robot moves around the scene, gets a noisy estimate (distance and direction) of the observed landmarks (you cannot see all landmarks at once) from a simulated LIDAR/Camera sensor.

This code simulates the environment, robot movement, robot observations and then formulates the SLAM problem as a Factor Graph using the famous GTSAM graph optimization framework.

Related Resources

• The accompanying Video lecture for this Toy Problem is can be found here.
• The accompanying slides can be found here.

Usage Instructions

• Install GTSAM 3.2.1 from here. Try to avoid the latest version of GTSAM (GTSAM 4.0) as it has some bugs.
• Be sure to be able to run any of the example from the toolbox.
• To run this code, run the Test1.m script.

Assumptions

The assumptions made are:

• The world extends from -WorldLim to WorldLim in both X and Y directions. Refer to Test1.m script.
• You have a noisy odometry estimate with noise covariance given by [0.3 0.3 0.1] ([MovementX MovementY MovementTheta]).
• The landmark noise is given by [0.1 0.1] in m.
• Landmarks are non blocking (you can see a landmark behind another landmark).