Robotic perception seems so cool, but it's not all deep learning and YOLO.

Every robot system is going to employ some middleware. The most common is the Robotic Operating System (ROS).

But why should a perception engineer care about ROS?

Sensors Are Useless Without A Framework

Robots are complex machines with lots of software components, that need to do at least the following things:

• Low level motor control.

• Fuse all of its sensors including encoders, cameras and lasers into a common coordinate frame.

• Perform path planning and obstacle avoidance.

Rather than write custom algorithms from scratch, you can use ROS to make this much easier.

But in order to use ROS you will need to integrate your sensor into the ROS system.

Nearly Every Sensor Will Need Some Specific Pre-Processing

If your laser intersects with your robot body you may need to apply pre-processing to remove points.

Alternatively the best sensor may not have a vendor provided ROS node, or perhaps the ROS node is not suitable. Other times you may need to upgrade the ROS node to your required ROS version.

Fundamentally software integration is an unavoidable key skill a robotics perception engineer must have.