rda_ros

This is the ROS Wrapper of the RDA planner

Preqrequisite

Please install the following libraries and packages first.

• Python >= 3.8
• ROS Noetic
• RDA_planner

Note

We recommend using Python 3.8 and Conda to manage Python environments, as some examples are based on CARLA.

Test Environment

• Ubuntu 20.04

Installation

You can install the package by running the commands below:

mkdir -p ~/rda_ws/src
cd ~/rda_ws/src
git clone https://github.com/hanruihua/rda_ros
cd ~/rda_ws && catkin_make
cd ~/rda_ws/src/rda_ros 
sh source_setup.sh && source ~/rda_ws/devel/setup.sh && rosdep install rda_ros 

Demonstration

Dynamic collision avoidance

We provide the dynamic collision avoidance examples in Gazebo shown as follows. To run these examples, please see example/dynamic_collision_avoidance for detail.

• Dynamic collision avoidance with obstacles information input

demo_rda_gazebo_obs.mp4

• Dynamic collision avoidance with scan information input

demo_rda_gazebo_scan.mp4

Autonomous Driving

We provide the Autonomous Driving examples in Carla shown as follows. To run these examples, please see example/auto_driving for detail.

• Autonomous Driving in Town02 with Topology Map and regular traffic vehicles

demo_rda_carla_town02.mp4

• Autonomous Driving in Town04 with traffic chaos

demo_rda_carla_town04.mp4

Node API

Published Topics

Topic Name	Message Type	Description
/rda_cmd_vel	geometry_msgs/Twist	Publishes velocity commands for the robot's movement.
/rda_opt_path	nav_msgs/Path	Publishes the optimized path generated by the RDA planner.
/rda_ref_path	nav_msgs/Path	Publishes the reference path that the robot should follow.
/rda_ref_states	nav_msgs/Path	Publishes the reference states trajectory for visualization.
/rda_obs_markers	visualization_msgs/MarkerArray	Publishes obstacle markers for visualization in RViz.

Subscribed Topics

Topic Name	Message Type	Description
/rda_obstacles	costmap_converter/ObstacleArrayMsg	Subscribes to obstacle information when use_scan_obstacle is set to False.
/scan	sensor_msgs/LaserScan	Subscribes to laser scan data for obstacle detection when use_scan_obstacle is True.
/rda_sub_path	nav_msgs/Path	Subscribes to the reference path that the RDA planner should follow.
/rda_goal	geometry_msgs/PoseStamped	Subscribes to the goal point to generate the reference path for rda planner

Parameters

Robot Information

Parameter	Type	Default Value	Description
~robot_info	dict		Configuration dictionary for robot specifications.
~robot_info.vertices	list or None	None	Vertices defining the robot's shape. If None, uses length and width to create a rectangular shape.
~robot_info.radius	float	None	Radius of the robot (used if the robot shape is circular).
~robot_info.max_speed	list	[10, 1]	Maximum speed parameters [linear, angular] or [linear, steer].
~robot_info.max_acce	list	[10, 0.5]	Maximum acceleration parameters [linear, angular] or [linear, steer].
~robot_info.length	float	2	Length of the robot (used if vertices are not provided).
~robot_info.width	float	1	Width of the robot (used if vertices are not provided).
~robot_info.wheelbase	float	1.5	Wheelbase of the robot.
~robot_info.dynamics	string	"diff"	Type of robot dynamics (e.g., differential: diff, ackermann: acker).
~robot_info.cone_type	string	"Rpositive"	Type of collision cone used, polygon: Rpositive, circle: norm2

RDA MPC Configuration

Parameter	Type	Default Value	Description
~receding	int	10	Receding horizon parameter for MPC.
~iter_num	int	2	Number of iterations for the MPC solver.
~enable_reverse	bool	False	Enables reverse movement if set to True.
~sample_time	float	0.1	Sampling time interval for the MPC.
~process_num	int	4	Number of parallel processes for MPC computation.
~accelerated	bool	True	Enables accelerated computation in MPC.
~time_print	bool	False	Enables time logging for MPC operations.
~obstacle_order	bool	True	Determines if obstacle ordering by distance is applied.
~max_edge_num	int	5	Maximum number of edges to consider for obstacles.
~max_obs_num	int	5	Maximum number of obstacles to consider.
~goal_index_threshold	int	1	Threshold for goal index determination.
~iter_threshold	float	0.2	Iteration threshold for convergence in MPC.
~slack_gain	float	8	Gain for slack variables in MPC constraints.
~max_sd	float	1.0	Maximum safety distance.
~min_sd	float	0.1	Minimum safety distance.
~ws	float	1.0	Weight for the state in the cost function.
~wu	float	0.5	Weight for the control inputs in the cost function.
~ro1	float	200	Weight parameter for the first term in the cost function.
~ro2	float	1.0	Weight parameter for the second term in the cost function.

Reference Speed

Parameter	Type	Default Value	Description
~ref_speed	float	4.0	Reference speed for the robot.

Scan Configuration

Parameter	Type	Default Value	Description
~use_scan_obstacle	bool	False	Determines whether to use laser scan for obstacle detection.
~scan_eps	float	0.2	Epsilon parameter for DBSCAN clustering on scan data.
~scan_min_samples	int	6	Minimum number of samples for a cluster in DBSCAN.

Reference Paths

Parameter	Type	Default Value	Description
~waypoints	list	[]	List of waypoints for generating the reference path.
~loop	bool	False	Determines if the path should loop upon reaching the end.
~curve_type	string	"dubins"	Type of curve generator used (e.g., dubins, reeds, or line).
~step_size	float	0.1	Step size for generating the reference path.
~min_radius	float	1.0	Minimum turning radius for the reference path.

Frame Configuration

Parameter	Type	Default Value	Description
~target_frame	string	"map"	Target frame for transformations.
~lidar_frame	string	"lidar_link"	Frame ID for the LiDAR sensor.
~base_frame	string	"base_link"	Base frame of the robot.

Visualization

Parameter	Type	Default Value	Description
~marker_x	float	0.05	Scale factor for marker size in visualization.
~marker_lifetime	float	0.1	Lifetime of markers in RViz (in seconds).

Citation

If you find this code or paper helpful, please consider starring the repository and citing our work using the following BibTeX entries:

  @ARTICLE{10036019,
  author={Han, Ruihua and Wang, Shuai and Wang, Shuaijun and Zhang, Zeqing and Zhang, Qianru and Eldar, Yonina C. and Hao, Qi and Pan, Jia},
  journal={IEEE Robotics and Automation Letters}, 
  title={RDA: An Accelerated Collision Free Motion Planner for Autonomous Navigation in Cluttered Environments}, 
  year={2023},
  volume={8},
  number={3},
  pages={1715-1722},
  doi={10.1109/LRA.2023.3242138}}

@ARTICLE{10601554,
  author={Li, Guoliang and Han, Ruihua and Wang, Shuai and Gao, Fei and Eldar, Yonina C. and Xu, Chengzhong},
  journal={IEEE/ASME Transactions on Mechatronics}, 
  title={Edge Accelerated Robot Navigation With Collaborative Motion Planning}, 
  year={2024},
  volume={},
  number={},
  pages={1-13},
  doi={10.1109/TMECH.2024.3419436}}

Known Issues

• RDA Planner is an optimization-based planner. Better performance can be achieved with a more powerful computer, especially when running with the CARLA and Gazebo simulators.
• If your computer doesn't match the demo's performance, try adjusting the parameters in the RDA MPC configuration to reduce computation costs.
• Since obstacles in some scenarios are moving and set randomly, there may be cases where the robot cannot avoid them. Adjust the parameters in the RDA MPC configuration can improve the performance.
• To manage the many parameters for this node, we recommend using YAML files to store them and loading them in the launch file, as shown in the demonstrations.

Acknowledgement

• limo_ros
• Carla
• Carla-ROS-bridge
• Gazebo

Contact

If you have any issues about the paper and code:

Ruihua Han

Guoliang Li

Shuai Wang

Chengyang Li