Does move_base require Twist from Odometry?

edit

Hi, yes robot_localisation should work too. I'm currently getting good results with laser_scan_matcher with a custom odometry node, so I'll continue playing around with that for a while. Thanks a lot! You've been great help!

@parvizal, could you provide a link to your custom odometry node? I take it you're taking odom from tf and converting it to nav_msgs/Odometry, using good-enough values for covariance? I'm doing exactly what you're doing (with an RPLidar as well :) ) and came to the same conclusion.

Edit: odom is only published in tf when you run the laser_scan_matching with gmapping demo :) So I suppose you're just using the PoseStamped msg and v=(x-x0)/dt to get twist, and publishing that ..

Wrote this up as a base but can't test yet due to broken robot :/

    pose = data.pose
    ts = data.header.stamp
    #https://answers.ros.org/question/69754/quaternion-transformations-in-python/
    euler = tf.transformations.euler_from_quaternion(pose.orientation)
    yaw = euler[2]
    # different method
    #odom->pose.pose.orientation.z = sin(yaw_/2.0)
    #odom->pose.pose.orientation.w = cos(yaw_/2.0)

    dx   = pose.position.x   - self.prev_x
    dy   = pose.position.y   - self.prev_y
    dyaw = pose.position.yaw - self.prev_yaw
    dt   = ts                - self.prev_ts

    odom = Odometry()
    #odom.header = 
    #odom.child_frame_id = 
    odom.pose.pose = pose
    odom.pose.covariance[0]  = 0.2 # x, covariances here are likely widely off
    odom.pose.covariance[7]  = 0.2 # y
    odom.pose.covariance[35] = 0.4 # yaw
    odom.twist.twist.linear.x    = dx/dt    
    odom.twist.twist.linear.y    = dy/dt    
    odom.twist.twist.angular.yaw = dyaw/dt