The previous newsletter described SubT and indicated that development has started to focus on its needs. SubT is a long running program with a final competition to be held in the fall of 2021. This schedule places certain restrictions on the set of software that we can use, assuming we want to avoid EOL distributions at the time of the final competition. A larger concern is the scope of simulation requirements, which can be summarized as tens of heterogeneous robots operating in underground spaces with a traversable length of tens of kilometers.
Based on time schedule and feature requirements, we are pursing a two-phased development strategy. Phase 1 is aimed at satisfying near term needs through the use of existing software which includes Gazebo 9, ROS Melodic and Ubuntu Bionic. In this phase, the largest limitation will be scale. The size of robot teams and environment complexity will be restricted. Phase 2 will address scale issues and incorporate new features through the use of the Ignition libraries in Gazebo 11.
You may have noticed that Gazebo version 10 was skipped. This is because we are too close to the Gazebo 10 release date to make the changes needed for SubT. So, Gazebo 10 will be an incremental improvement to Gazebo 9 and Gazebo 11 will contain the major new features required by SubT.
As a user and/or developer of Gazebo, you can expect the following.
We'd like to thank the following person for their contributions to Gazebo.
The following list includes releases made over the past month.
Ignition Transport 5, currently in prerelease, ships with a message logging API and a helpful command line tool. Logging is acheived by writing serialized Protobuf messages along with metadata, such as topic name and message type, into an SQLite3 database.
We chose to use SQLite3 due to its stability, performance, cross-platform support, and large community with numerous tools, applications, and language bindings. Performance is an important consideration since we may need to log large quantities of messages, both in size and count. We evaluated the performance of SQLite using statistics published by the authors of SQLite.
We also ran a few tests ourselves to augment the official SQLite statistics. For 10KB messages we achieved a write speed of 210MB/s or roughly 21k msgs/sec. For comparison a 4.5GB file copied from one part of a disk to another achieved a rate of 189MB/sec. This result indicates that SQLite is utilizing the disk well, and is likely bound by disk bandwidth.
Take a look at this tutorial to get started with logging through Ignition Transport.
This month's featured model is a Toyota Prius equipped with a variety of sensors. The sensors include an IMU and GPS along with set of cameras, sonars, and LIDARs. This Prius model is perfect for testing and developing autonomous vehicle software.
Most recent active questions