# Why use GraphSLAM?

I am doing some project on Slam. I have a data set of a moving Robot which can give me the value of forward velocity(m/s) and angular velocity(rad/s) and time(s). Now if this data are provided I can find out the x,y,theta value of a Robot which I can plot and find the Robot path.

Now I am working on Graph Slam and little bit confused in this aspect. I know Graph slam also gives us path and map, but I can already determine the path using forward velocity and angular velocity then what is the need for Graph Slam?

• Do you mean path or pose? Jul 25 '18 at 20:11

## 2 Answers

SLAM (Graph or EKF) not only provides you a map of the environment (assuming you have some sort of an exteroceptive sensor like a camera or a rangefinder), but it also capable of performing much more accurate localization by improving your pose estimates.

If you simply fuse your onboard sensor data to obtain an $(X, Y, \theta)$ estimate of your poses, it is extremely likely that this estimate will drift from the actual ground truth over time. Observing additional information (landmarks) and performing sensor fusion will help in reducing this drift. At the same time, while using SLAM, you can keep track of if/when you're observing the same landmarks you've seen before and use that information to adjust your pose estimates, commonly known as loop closure.

I can already determine the path using forward velocity and angular velocity

If you already know the exact path of robot, you dont need SLAM algorithm, you just have to integrate your sensor readings(laser /camera) to get the map. Its known as Mapping with known poses.

I have a data set of a moving Robot which can give me the value of forward velocity(m/s) and angular velocity(rad/s) and time(s)

In general with consumer grade IMUs, calculating robot pose from forward velocity(m/s) and angular velocity(rad/s) and time(s) is not accurate and errors will accumulate with time.