Hello ROS Fans,
For purposes of testing, I am trying to use move_base together with the ArbotiX differential drive controller in "fake simulation" mode to move a simulated TurtleBot along a 1 meter square. In simulation mode, the ArbotiX differential controller moves the robot exactly where you tell it so that odometry is essentially perfect. So there is no physics involved as when using Gazebo.
I want to move a simulated TurtleBot under these conditions along a 1 meter square as accurately as possible. Using a script that directly monitors the /odom topic and uses Twist commands to move the robot, I can get a pretty good result as shown in this image:
However, if I use move_base and a blank map to try to accomplish the same thing, I'm having a lot of trouble finding parameters that give a good result. Here for example is the best I've been able to do so far:
I am using tolerances of 0.05 meters for xy and 0.087 radians for yaw. I would have thought that with perfect odometry, it would be relatively easy for move_base to hit these tolerances but as the image above shows, there is a lot of slop in the trajectory. No doubt I am missing something basic but I can't figure it out.
Below are the config files I am using. After those is the script I am using to move the robot in a square:
base_local_planner_params.yaml:
controller_frequency: 20.0 #planner_patience: 5.0 #controller_patience: 5.0 TrajectoryPlannerROS: max_vel_x: 0.2 min_vel_x: 0.05 max_rotational_vel: 1.0 min_in_place_rotational_vel: 0.2 escape_vel: -0.10 acc_lim_th: 0.8 acc_lim_x: 0.2 acc_lim_y: 0.2 holonomic_robot: false yaw_goal_tolerance: 0.087 # about 5 degrees xy_goal_tolerance: 0.05 # 5 cm goal_distance_bias: 0.6 path_distance_bias: 0.8 heading_lookahead: 0.325 meter_scoring: true heading_scoring: false oscillation_reset_dist: 0.05 occdist_scale: 0.05 publish_cost_grid_pc: false sim_time: 1.0 sim_granularity: 0.01 angular_sim_granularity: 0.01 vx_samples: 6 vtheta_samples: 20 dwa: false costmap_common_params.yaml:
obstacle_range: 2.5 raytrace_range: 3.0 robot_radius: 0.15 inflation_radius: 0.5 max_obstacle_height: 0.6 min_obstacle_height: 0.08 observation_sources: scan scan: {data_type: LaserScan, topic: /scan, marking: true, clearing: true} global_costmap_params.yaml:
global_costmap: global_frame: /odom robot_base_frame: /base_link update_frequency: 20.0 publish_frequency: 1.0 static_map: false rolling_window: true width: 10.0 height: 10.0 resolution: 0.05 transform_tolerance: 1.0 local_costmap_params.yaml:
local_costmap: global_frame: /odom robot_base_frame: /base_link update_frequency: 20.0 publish_frequency: 1.0 static_map: false rolling_window: true width: 1.0 height: 1.0 resolution: 0.05 transform_tolerance: 1.0 blank_map.yaml:
image: blank_map.pgm resolution: 0.05 origin: [-10, -10, 0] occupied_thresh: 1.0 free_thresh: 1.0 negate: 0 move_base_square.py:
#!/usr/bin/python import roslib roslib.load_manifest('rbx_nav') import rospy import actionlib from actionlib_msgs.msg import * from geometry_msgs.msg import Pose, PoseWithCovarianceStamped, Point, Quaternion, Twist from move_base_msgs.msg import MoveBaseAction, MoveBaseGoal from tf.transformations import quaternion_from_euler from math import radians class MoveBaseSquare(): def __init__(self): rospy.init_node('nav_test', anonymous=True) rospy.on_shutdown(self.shutdown) square_size = 1.0 # meters turn_angle = radians(90) # degrees q_turn_angle = quaternion_from_euler(0, 0, turn_angle, axes='sxyz') quaternion_turn_angle = Quaternion() quaternion_turn_angle.x = q_turn_angle[0] quaternion_turn_angle.y = q_turn_angle[1] quaternion_turn_angle.z = q_turn_angle[2] quaternion_turn_angle.w = q_turn_angle[3] # Goal state return values goal_states = ['PENDING', 'ACTIVE', 'PREEMPTED', 'SUCCEEDED', 'ABORTED', 'REJECTED', 'PREEMPTING', 'RECALLING', 'RECALLED', 'LOST'] # Publisher to manually control the robot (e.g. to stop it) self.cmd_vel_pub = rospy.Publisher('cmd_vel', Twist) # Subscribe to the move_base action server self.move_base = actionlib.SimpleActionClient("move_base", MoveBaseAction) rospy.loginfo("Waiting for move_base action server...") self.move_base.wait_for_server(rospy.Duration(60)) rospy.loginfo("Connected to move base server") rospy.loginfo("Starting navigation test") for i in range(4): # First move along a side self.goal = MoveBaseGoal() self.goal.target_pose.pose.position.x = 1.0 self.goal.target_pose.pose.orientation = Quaternion(0, 0, 0, 1) self.goal.target_pose.header.frame_id = 'base_link' self.goal.target_pose.header.stamp = rospy.Time.now() self.move() # Now make a turn self.goal = MoveBaseGoal() self.goal.target_pose.pose.orientation = quaternion_turn_angle self.goal.target_pose.header.frame_id = 'base_link' self.goal.target_pose.header.stamp = rospy.Time.now() self.move() def move(self): # Start the robot toward the next location self.move_base.send_goal(self.goal) # Allow 1 minute to get there finished_within_time = self.move_base.wait_for_result(rospy.Duration(60)) if not finished_within_time: self.move_base.cancel_goal() rospy.loginfo("Timed out achieving goal") else: state = self.move_base.get_state() if state == GoalStatus.SUCCEEDED: rospy.loginfo("Goal succeeded!") def shutdown(self): rospy.loginfo("Stopping the robot...") self.move_base.cancel_goal() rospy.sleep(2) self.cmd_vel_pub.publish(Twist()) rospy.sleep(1) if __name__ == '__main__': try: MoveBaseSquare() except rospy.ROSInterruptException: rospy.loginfo("Navigation test finished.") Originally posted by Pi Robot on ROS Answers with karma: 4046 on 2012-05-27
Post score: 5
Original comments
Comment by Eric Perko on 2012-05-27:
Could you try a path that doesn't require a "spin-in-place" operation and see if you get the results you expect? base_local_planner's trajectory scoring function doesn't work well for choosing pivot turns.
Comment by SL Remy on 2012-12-24:
Patrick, could you officially post the answer to the question?