Contents
This node is a robot controlled based on the Dynamic Window Approach described in "Fox, D.; Burgard, W.; Thrun, S., "The dynamic window approach to collision avoidance" Robotics and Automation Magazine, IEEE , vol.4, no.1, pp.23-33, Mar 1997".
It differs from the original approach in that the best speed are found by random sampling the search space instead of using a fixed step incremental search. This allows the controller to run at a specific frequency while still guaranteeing a good solution.
Input
odom: the robot nav_msgs/Odometry message.
base_scan: the laser sensor_msgs/laserscan message.
cmd_vel_dwc: the desired linear/rotational speed as a geometry_msgs/twists message.
base_link: the robot base frame of referece.
Output
cmd_vel_dwc: the node publishes the linear/rotational speeds as geometry_msgs/twists messages.
debug_point_cloud: if enabled, the node publishes sensor_msgs/Pointcloud messages corresponding to the laser scan.
Parameters
alpha, beta, gamma: the importance weights given to heading, clearance and speed. See the original paper for the mathematical meaning of this parameters.
vmax: the maximum linear velocity of the robot.
vpmax, vpbrake: maximum linear acceleration and brake (reverse acceleration) of the robot.
wmin, wmax: minimum and maximum rotational velocities of the robot.
wpmin, wpmax: minimum and maximum rotational accelerations of the robot
tcmax: for how long (in seconds) should the controller search for an optimal value.It is the inverse of the frequency at which the controller should run.
laser_max: any laser reading below this value is not considered.
robot_radius: this is the radius of the circle that approximates the robot base.
publish_pcl; if the pointcloud showing the laser will be published.
Code Stability
This code is stable. Although I cannot guarantee the absence of bugs, I've used it for several years with different robot platform and it always proved reliable. Please feel free to send any comment and/or bug report to lorenzo.riano (at) gmail.com.
