The Bottleneck

Our previous stack running on ROS 1 Noetic was serving us well, until we started adding high-bandwidth sensors. The centralized roscore became a single point of failure, and the TCP overhead was introducing latency in our control loops.

Enter ROS 2 Humble

We decided to bite the bullet and migrate to ROS 2 Humble. The key advantages we were looking for:

• DDS (Data Distribution Service): No more master node!
• Lifecycle Management: Better state control of our nodes.
• Python 3 Modernization: leveraging the latest libraries.

The Migration Process

It wasn't smooth sailing. We had to rewrite our custom messages and adapt our heavy SLAM algorithms.

# Old ROS 1 Publisher
pub = rospy.Publisher('chatter', String, queue_size=10)

# New ROS 2 Publisher
self.publisher_ = self.create_publisher(String, 'chatter', 10)

Results

After 2 months of grueling refactoring, our latency dropped by 40%. The rover felt snappier, and the peace of mind of having a decentralized network was worth every sleepless night.