Date: April 2022
The simulator features a map of our classroom with a virtual robot in the map. The linear and angular velocities of the virtual robot can be changed, and the distance from a sensor on the front of the virtual robot can be collected. These functionalities are controlled with the Commander class.
Open Loop Control
To execute a square, the robot moves with a linear velocity of 0.1 m/s and an angular velocity of 0 rad/s for 2 seconds. cmdr.set_vel(0.1, 0)
Then the robot turns with a linear velocity of 0 m/s and an angular velocity of 90 deg/s for 1 seconds. cmdr.set_vel(0, math.radians(87.5))
Through trial and error, I determined that due to delays, turning 87.5 degrees (converted to radians) would results in a 90 degree turn in the simulator. I repeated this sequence for 4 times to create a square.
The controls are not perfectly repeatable and the exact same shape is not executed everytime. As seen in the plotting in the videos, the odometry does not match the ground truth well.
Closed Loop Control
To avoid obstacles, I have the robot first backup when it senses it’s a certain distance from a wall then turn 60 degrees. When the robot moves at 0.5 m/s and tries to detect wall from 0.1 m out, the robot crashes. I found that increasing the detected distance to 0.25 m allowed the robot to act accordingly when it was moving at 0.5 m/s.
I reworked by code to get the pose data and plot it more frequently to attempt to improve the odometry. I saw some improvements, but the odometry still wasn’t perfect.
I wrote a python function called readData() that gets the pose and distance sensor data, plots both of the pose data, and returns the odometry pose, ground truth pose, and distance arrays.
def readData():
pose, gt_pose = cmdr.get_pose()
cmdr.plot_odom(pose[0], pose[1])
cmdr.plot_gt(gt_pose[0], gt_pose[1])
dist = cmdr.get_sensor()
return pose, gt_pose, dist
The distance sensor only detects directly in front of the center of the robot. If an obstacle is only partially in the path of the robot, there is a chance the distance sensor will not detect it and instead pick up on something else behind the obstalce. In this case, the robot will run into the obstacle.