Searching for object in an unknown environment using ROS

Question

Using ROS, I'm trying to build a robot able to search for specific objects in an unknown environment. Right now, I'm experimenting with a mobile robot in a simple gazebo maze, using a LIDAR for navigation/mapping and its camera for object recognition on "STOP" signs. Here's a picture from the Husarion tutorial (now deprecated):

As a ROS beginner, I find myself writing more code than I expected, and I'm wondering if I'm actually reinventing the wheel. After some googling, these are my findings:

1. The find_object package, that seems without activity since 2010.
2. Husarion's object search tutorial, that's been deprecated and not supported anymore.

Is this task well-known and I'm reinventing the wheel? If that's the case, is there a standard/best-practice way to approach it? I would very much appreciate if you can point me to any books/videos/tutorials/examples.

Answer 1

Is this task well-known and I'm reinventing the wheel?

To truly answer this, you will have to conduct a full literature review of what is out there. 'find objects' is a bit too vague. Could you be more specific as to what objects are you trying to detect and what sensors are available to you?

I'd say that general object detection is well know in robotics and there are several methods to detect objects but classifying them is another story and one that depends on your application and available resources.

If you only want to detect objects in lidar point clouds you can use something like euclidean cluster extraction. If your aim is to classify objects in the image space you could use some NN based approach. The darknet_ros package would be a good starting point.