The tendon_cr_environment is a ROS2-based environment designed for the simulation and control of a tendon-driven continuum robot. This project enables users to run simulations and implement control schemes for the robot. Built on ROS2 Humble, the interactive environment supports the integration of multiple functions operating in parallel, facilitating a smooth transition to physical robot implementation.
The simulation of the continuum robot leverages PyElastica for its backend processing. The open-source nature of this project encourages code inspection and further development by users, whether that involves adding new control schemes, creating interactions between multiple robots, or modifying the actuation mechanisms.
The logic and mathematics behind tendon actuation in the open-loop case are similar to those described in TendonForces, which supports custom tendon-driven configurations for a continuum robot and requires the specification of the tension applied to each tendon, making it suitable only for open-loop control.
The closed-loop scenario necessitates modifications to TendonForces to accommodate feedback processing and the application of control signals. ROS2 simplifies this task by allowing feedback reception and processing to be handled by a controller node, while only the control signals must be applied to the simulated continuum robot.
For the closed-loop case, the QuadTendonForces module is used, which allows for active tendon updates and tension adjustments. The logic and mathematics of QuadTendonForces closely resemble those of TendonForces, differing mainly in that the tendon configuration is fixed with four long tendons and four short tendons. User-specified parameters adjust the parameters of these tendons (without changing the setup), and tension is managed by the controller node in the ROS2 environment. NOTE: "long" and "short" tendons refer to a setup in which there are a total of 8 rows of tendons and vertebrae located in the global +Z, -Z, +Y, -Y directions (one "short" and one "long" for each direction), the "long" ones being of longer length than the "short" ones. The purpose of this setup is to allow for the contorsion of the robot to allow for ONE concavity change by way of antagonistic tendons of different lengths activated at the same time.
If users desire a different tendon setup, a new forcing module must be created to cater to that specific setup. The QuadTendonForces module is designed to allow for a maximum of one concavity change caused by pairs of antagonist tendons.
The functionality of the nodes in this project is described in the following sections. For a more in-depth overview, please refer to: INSERT THESIS HERE.
All parameters specified in the GUI are in SI units.
This node provides a user-friendly GUI for simulating an open-loop tendon-driven continuum robot system. Users can specify the parameters for the rod, simulation, and a custom tendon actuation configuration, with the ability to add multiple tendons, each having its own custom parameters.
This node facilitates the simulation and control of a closed-loop tendon-driven continuum robot system. Users can specify the rod's parameters, simulation parameters, and QuadTendonForces configuration parameters for the long and short tendons. Additionally, users can define the desired XYZ tip position for the robot and set the proportional gains for the controller. The GUI includes a gain search function, allowing users to specify a desired convergence time, which is then used to fine-tune the proportional gains for achieving the desired position within the specified time.
The simulator node processes input from either the manual GUI or the control GUI, selecting the appropriate tendon actuation external forcing module (TendonForces or QuadTendonForces) accordingly, and then running the simulation of the system using PyElastica as the backend.
In this straightforward setup, the simulator node receives parameters from the manual GUI, runs an open-loop simulation, and publishes the results to the visualizer node for user review.
The closed-loop simulation requires additional nodes for publishing and subscribing to data. Two extra nodes are initiated for this purpose, which are used within PyElastica simulator's callback function. The published information is primarily received by the controller node, while the subscribed information is used update the activated tendons in the system and their respective tensions. After the simulation, results are published to the visualizer node for processing.
The visualizer node receives parameters from the chosen GUI node, along with position and directors historical data from the simulator node. It then creates a video of the system's evolution over time (saved to the home directory) and generates a 3D plot of the robot's final pose.
This node receives parameters from the control GUI as well as current simulation information. It determines which tendons should be activated to achieve the desired XYZ tip position and updates the tensions assigned to the activated tendons. If the gain searcher option is activated, it iteratively adjusts the proportional gains based on logic defined by the gain searcher node.
The gain searcher node's goal is to identify optimal proportional gains for the controller node, to reach the desired XYZ tip position within the specified time frame. This is accomplished through simulations that vary the proportional gains until the required time tolerance is met.
Ensure you have the following prerequisites installed:
-
Clone the Repository
Clone this repository to your local machine:
git clone https://github.com/gabotuzl/tendon_cr_environment.git cd tendon_cr_environment -
Build the ROS2 Workspace
Navigate to the
environment_wsfolder and build the workspace:cd tendon_cr_environment_ws colcon build -
Source the Workspace
After the build is successful, source the setup script to add the workspace to your ROS2 environment:
source install/setup.bashYou may want to add this line to your shell startup file (e.g.,
~/.bashrc) for convenience:echo "source /path/to/repo_name/tendon_cr_environment_ws/install/setup.bash" >> ~/.bashrc
To run the full simulation environment, you can use the provided launch files.
To use the manual (open-loop) simulation environment which contemplates custom tendon configurations, open a terminal and execute:
ros2 launch tendon_cr_environment manual_tcre.launch.pyTo use the control (closed-loop) simulation environment which allows for the control of the robot tip's position by using a predetermined tendon setup (with customizable parameters), open a terminal and execute:
ros2 launch tendon_cr_environment control_tcre.launch.pyIf you prefer to run each node individually, you can do so by following these steps:
-
Open multiple terminal windows (or tabs).
-
Source the workspace in each terminal:
source /path/to/repo_name/tendon_cr_environment_ws/install/setup.bash -
Start each node by running the following commands in separate terminals:
-
GUI Node(s): For the open-loop simulation GUI:
ros2 run tendon_cr_enviroment gui_manual_node
For the closed-loop simulation and control GUI:
ros2 run tendon_cr_environment gui_control_node
-
Simulator Node:
ros2 run tendon_cr_environment simulator_node
-
Visualizer Node:
ros2 run tendon_cr_environment visualizer_node
-
Controller Node:
ros2 run tendon_cr_environment controller_node
-
Gain Searcher Node:
ros2 run tendon_cr_environment gain_searcher_node
-