Sensor Fusion in Mobile Autonomous Robot | ROS | IMU+Wheel Odometry | Kalman Fliter | Jetson Nano

in this video we will see sensor fusion on mobile robots using robot localization package first we'll find out the need for sensor fusion then we'll see how to use robot localization node for sensor fusion and finally we'll see the comparison of photometric data with and without sensor fusion this is our mobile robot it currently has two sensor data laser scan and audiometry from wheel encoders amcl node uses odometry data and laser scan data to localize the robot in the map as the robot moves the localization of the robot becomes more accurate this is good until the surface provides good friction to the wheels and there is no drifting off wheels but we cannot guarantee that the surface is even everywhere with good friction let's see what happens when robot moves over less friction surfaces to demonstrate such bad surface i have placed a mat on the floor i'll try to go over the mat and see what happens the robot started losing the accuracy in localization why is this happening the reason for this is faulty odometry data from wheel encoders when there is less friction on the surface the wheels drift and the robot does not move in my case i am trying to rotate the robot but due to less friction right wheel is drifting and making the robot stationary but the odometry data from the wheel encoder says to amcl node that the robot is rotating so due to this mismatch in automatic data and the real state of the robot amcl node fails to localize the robot in the map to improve this condition we have to make the odometry data more reliable to match with the real state of the robot we can do this by sensor fusion we can add another sensor like imu which can track the rotations and accelerations of the robot and fuse the imu data with v loadometry to get more accurate robot audiometry for sensor fusion in mobile robots we can use robot localization package this package can be used to fuse wheel odometry imu sensor data and also gps data it internally uses kalman filter to predict the real state of the robot with the fused data i am adding mpu6050imu sensor to my report to use robot localization package all we have to do is just add this node in the launch file and setup its parameters let's have a look into the parameter file this file contains information of frame ids input sources that is reloadometry data and imu sensor data also we can add covariance matrices for kalman filter as i did not write any covariance matrices here the default values will be used these are the default covariance matrices now let's launch this file and test the robot's localization by driving over the mat even though we drive the robot on bad surface robot's location in the map is proper so this is how fusing imu data and wheel odometry improves slowboard localization now let's see a demo comparing the audiometry with just wheel data and filtered odometry with imu plus wheel data fused the red marker indicates wheel odometry and the blue marker indicates the fused odometry as we can see on good surface both odometries are relatively same we can see the difference once the robot get onto the mat see here the robot is not moving but the wheel odometry is still rotating the fused odometry is not rotating matching the real state of the robot according to the wheel loadometry the robot is here which is already out of map but the fused automatically shows the exact location of the robot so this is how sensor fusion helps in getting the accurate state of the system hope you like this video then don't forget to hit the like button and do subscribe to my channel for more interesting videos on robotics and raws for more stuff visit my website www.roboticslearning.com thanks for watching

Channel: Robotics and ROS Learning

Views: 30,118

Keywords: sensor fusion, mobile autonomous robot, differential drive, imu, mpu6050, lidar, slam, autonomous navigation, rplidar, fusion, robot localization, kalman filter, robotics

Id: 0yICGqriN3g

Length: 6min 15sec (375 seconds)

Published: Tue Apr 20 2021