LOLA Environment Perception Package v2: understanding the environment sensed by a Kinect-like sensor.
The goal of the project is to approximate arbitrary objects in the video feed of a Kinect-like sensor, using a number of basic geometric shapes, such as spheres and cylinders, in real time.
The approximations can then be used as a basis for a robot control system, providing it with precise information on the positions of obstacles.
As such, in order to meet the real-time requirement, precision in the generated approximations is sometimes sacrificed, but with the goal of never under-approximating an object, so that the robot does not accidentaly end up colliding with it.
As part of the research and development of humanoid robots that the Institute of Applied Mechanics of the Technical University Munich does, a subsystem for obstacle detection based on the sensory information provided by Kinect-like sensors was developed.
Initially, the system was based on analyzing pure depth maps (2D images, where each pixel is additionally annotated with depth measurements), as returned by the sensor. It subsequently moved to leveraging PCL, in the hopes of simplifying, as well as improving the subsystem by making use of the algorithms already implemented by PCL.
This project represents the latest evolution of this subsystem. It builds upon ideas that were previously implemented in the vision subsystem, improving the object approximation algorithms, implementing some new approaches for obtaining the best possible approximations, as well as optimizing and improving the efficiency and modularity of the previously existing codebase.
It also extracts the purely vision-related components of the subsystem --
the obstacle detection and approximation -- into an independent and
reusable C++ library. The library is found in the
src/lepp2
directory.
This not only facilitates better code organization and modularity of the humanoid robot system, but also contributes back to the open source community, by openly sharing the exact implementation of the object/obstacle detection and approximation used by the Institute in its humanoid robot systems.
The implementation of the vision subsystem is also found within this
repository, in the
src/lola tree.
Most of the code there represents communication logic between the robot
control and the vision subsystems, such as obtaining the robot's kinematic
parameters and sending the detected obstacles' descriptions back to the control,
according to previously specified communication protocols. This can be
considered as a good example how the lepp2 library can be used within the
context of a larger system.
Whether the vision subsystem executable gets built is controlled by the
LEPP_BUILD_LOLA option passed to cmake when generating the build
configuration (on by default).
The detector executable, on the other hand, built when
LEPP_BUILD_DETECTOR is set (on by default), is independent from any robot
specifics and displays the detected obstacles' approximations by overlaying
them over a point cloud displayed in PCL's PCLVisualizer.
The project depends on the PCL library. You should first install it following PCL's directions.
Then, you can compile the detector using cmake. In order to compile the
project with only the default options set, use the build.sh script.
The detector executable is built by default and has a number of flags
that can be passed to it from the CLI. Check its usage by running it with
no flags.
For some examples of how to use the library itself, you may check the
examples directory.
The project is published under the terms of the MIT License.