R2022b

New Features, Bug Fixes, Compatibility Considerations

Spatial Math: Use SE(2), SE(3), SO(2), and SO(3) transformations

Use the se3 and so3 objects to represent 3-D transformation matrices and 3-D rotation matrices, respectively. These objects make it easier to perform common operations, such as transforming 3-D points, normalizing the rotation, or interpolating between positions and orientations. To represent multiple transformations and rotations, you can arrange these objects into arbitrarily shaped arrays.

The se2 and so2 objects represent 2-D transformation matrices and 2-D rotation matrices, respectively.

Use plotTransforms to visualize the transformations and rotations in a figure. This can be helpful to see the underlying positions and orientations and to see interpolation results.

Custom Sensor Support in Robot Scenarios: Add custom sensor models to robots in simulated scenarios

Specify custom sensor models and define their behavior in simulation using the robotics.SensorAdaptor class. To generate a template for implementing the class, use the createCustomRobotSensorTemplate function.

Time Optimal Trajectory: Create time optimal trajectory subject to kinematic constraints

Use the contopptraj function to generate a time optimal trajectory subject to velocity and acceleration limits. This is also useful for retiming existing trajectories such that they are also within acceleration and velocity limits. For more information, see Generate Time-Optimal Trajectories with Constraints Using TOPP-RA Solver.

TOPPRA generated trajectory graph of position, velocity, and acceleration

Capsule Collision Geometry: Use additional collision capsule geometry

Create collision capsules as collisionCapsule objects and use the fitCollisionCapsule function to fit a collision capsule over a collisionBox, collisionCylinder, collisionSphere, or collisionMesh object. This is useful for reducing motion planning times by approximating collision meshes with collision capsules. See Reduce Motion Planning Times Using Capsule Approximation for more information.

Use the genspheres object function to decompose a collision capsule into spheres for faster collision-checking performance.

Collision capsule and box with capsule fitted over it and spherically decomposed box

Collision Checking: Check collisions using collision capsules

The checkCollision function now supports collision checking between collisionCapsule objects.

Capsule Approximation: Approximate rigid body trees using capsules

Use the capsuleApproximation object to create an approximation of a rigid body tree robot by using collisionCapsule objects, for more efficient collision checking. By creating approximations of collision meshes of the rigid body tree, you can reduce motion planning times. See Reduce Motion Planning Times Using Capsule Approximation for more information.

Valkyrie collision geometry and capsule approximation

Robot Models: Use additional manipulator and gripper rigid body tree robot models introduced to robot model library

You can now retrieve these additional robots from the robot model library using the loadrobot function:

"meca500r3" — Mecademic Meca500 R3 six-axis robot arm
"robotiq2F85" — Robotiq 2F-85 two-finger gripper

Self-Collision Checking: Alter rigid body tree self-collision checking behavior change, new default self-collision checking behavior

You can now specify self-collision checking behavior for a rigid body tree robot model by using the SkippedSelfCollisions name-value argument of the checkCollision and by setting the SkippedSelfCollisions property of the manipulatorCollisionBodyValidator and manipulatorRRT objects. Specify the SkippedSelfCollisions argument as "parent" or "adjacent".

"parent" — Collision checking ignores self-collisions between parent and child rigid bodies.
"adjacent" — Collision checking ignores self -collisions between rigid bodies of adjacent indices.

As of R2022b, the default behavior of collision checking is to ignore self-collisions between parent and child rigid bodies. In previous releases, the default behavior of self-collision checking was to ignore self-collisions between rigid bodies at adjacent indices. To instead ignore self-collisions between rigid bodies of adjacent indices, specify the SkippedSelfCollisions argument as "adjacent".

Gazebo Co-Simulation: Gazebo co-simulation enhancements

The Gazebo co-simulation features have been updated with these enhancements:

You can now connect to multiple Gazebo simulations from one or more machines. You can now specify a cell array of IP addresses and a cell array of port numbers in the MATLAB^®, workspace and then specify their variable names to the Hostname/IP Address and Port boxes, respectively, of the Configure Gazebo Simulation dialog box.
To connect to a single Gazebo session from MATLAB, specify the port number and IP address of the computer running the Gazebo simulator.
```
portnum = 14580; 
ipaddress = '172.18.250.125'; 
```
To connect to multiple Gazebo sessions from MATLAB, specify the port numbers and IP addresses of the computers running the Gazebo simulator.
```
portnum = {14580,14581}; 
ipaddress = {'172.18.250.125','172.18.250.125'}; 
```
The Gazebo Read block now automatically suggests to change the dimension of the bus to match the signals from Gazebo. You can now toggle the joint-axis-related bus signal to variable or fixed dimension.

Gazebo Co-Simulation: Gazebo Read block sensor read performance improvements

The Gazebo Read block now has improved sensor read performance. The performance is measured on two different machine set ups while keeping the Real Time Factor in the Gazebo simulator at 1 for all the conditions. All the sensors are running at 30Hz.

The tables show the time taken to complete a 10 second simulation to read messages from each of the sensors. The sensors with read time that matches the simulation time of 10 seconds are considered to be at real time.

Gazebo and MATLAB running on the same machine

Release Version	IMU	Lidar	Camera
R2022a	769 sec	12.82 sec	66.66 sec
R2022b	10 sec	10 sec	30.3 sec

Gazebo and MATLAB running on different machines

Release Version	IMU	Lidar	Camera
R2022a	909 sec	33.33 sec	100 sec
R2022b	25 sec	25 sec	50 sec

These simulations were timed in MATLAB running on a Ubuntu^® Bionic 20.04, Intel^® Xeon^® W-2133 CPU @ 3.60 GHz, with 12 threads and 64 GB of RAM.

The Gazebo simulator is running on a Ubuntu Bionic 20.04 virtual machine, Intel Xeon W-2133 CPU @ 3.60 GHz, with 8 threads and 8 GB of RAM.

waypointTrajectory Reverse Motion: Specify wait and reverse motion for waypoint trajectory

You can now specify wait and reverse motion using the waypointTrajectory System object™.

To let the trajectory wait at a specific waypoint, simply repeat the waypoint coordinate in two consecutive rows when specifying the Waypoints property.
To render reverse motion, separate positive (forward) and negative (backward) groundspeed values by a zero value in the GroundSpeed property.

Support Package Updates

The Robotics System Toolbox™ Support Package for Manipulators has been split into these support packages:

Robotics System Toolbox Support Package for Kinova^® Gen3 Manipulators
Robotics System Toolbox Support Package for Universal Robots UR Series Manipulators

New Examples

This release contains several new examples: