[robocup-nao] SPQR Team announce B-Human code usage for RoboCup 2018

[Vincenzo Suriani]

Dear RoboCup 2018 Standard Platform League Technical and Organizing
Committees,
Dear RoboCup SPL teams,

in accordance to the terms and conditions of the BHuman code release
license (https://github.com/bhuman/BHumanCodeRelease/blob/master/License.md)
the SPQR Team would like to announce the use of the B-Human code for the
upcoming RoboCup 2018 (Montreal, Canada).

We use some of the native B-Human modules including walking, simulation,
whistle recognition, calibration and CABSL.
Instead, we use our own code for perception, coordination, and multi-robot
distributed decision making.

The following parts of the SPQR Team code release are original:
- Visual perception
- Team coordination
- Data fusion among the robots
- Behaviors
- Motion primitives (besides get-ups)

The following parts are modifications of the original B-Human code:
- Localization
- Communication network

A detailed description of the SPQR original contribution can be found at
the end of this email.

We would like to thank the B-Human Team for sharing their code.

Kind regards,
Vincenzo Suriani, on behalf of the SPQR Team




Visual Perception.
>From 2017, SPQR Team is using as replacement of the B-Human Ball Perceptor
a newly developed machine learning approach inspired by our deep learning
work on NAOs [1]. This allows us to play outdoor without color and camera
setting calibrations (see details in [5, 7]). We forked the B-Human 2016
repository adding our Ball Perceptor code (
https://github.com/SPQRTeam/SPQRBallPerceptor). It is worth to be noted
that the code on GitHub is our Ball Perceptor embedded in the B-Human
framework. SPQR Team has its own modules for data fusion in replacement of
the B-Human TeamBallLocator and the TeamPlayersLocator. In particular, our
modules are based on the P-Tracking library [5] to achieve a distributed
multi-robot object tracking.

Team Coordination.
>From RoboCup 2015, we have an algorithm that exploits the high level
information about game situations to obtain a specific behavior in response
of multiple environmental stimuli. The aim is to create a more effective
way of perceiving the World. This coordination module models the context
features of a specific environment, such as RoboCup, and integrates
different coordination techniques for a team of robots. More in detail, our
approach relies upon two well-known methods for coordinating a team of
robots: distributed task assignment and distributed world modeling, by
combining the robustness of them. This work achieved the IROS RoboCup Best
Paper Award in 2016 [4]. To use our coordination in a real playing
scenario, we have developed a tool for testing the network communication in
presence of a large number of packets, which can create delays in the
coordination. This tools has been used to develop our own adaptive
networking module, which is in addition to the B-Human networking
infrastructure. Our communication system can adapt to the external network
conditions and can provide more stability to the coordination module, even
under network unreliability. This enhances the coordination process in high
network traffic conditions. The details about our networking pipeline can
be found in [4, 8]. To validate our approach, we have developed a tool to
influence the network reliability during regular matches in the SimRobot
simulator. This year, we modified the network modules provided by B-Human
2017 code (i.e., TeamData and TeamMessageHandler) and we extended the TCM
in order to plot additional debugging information.

Non-deterministic Decision Making.
In 2016, we presented a method based on a combination of Monte Carlo search
and data aggregation (MCSDA) [3] to adapt discrete-action soccer policies
for a defender robot to the strategy of the opponent team. By exploiting a
simple representation of the domain, a supervised learning algorithm is
trained over an initial collection of data consisting of several
simulations of human expert policies. Monte Carlo policy rollouts are then
generated and aggregated to previous data to improve the learned policy
over multiple epochs and games. At the moment, our team is working on a
extension of these procedures to learn collective team strategies with
MCSDA.

References
[1] D. Albani, A. Youssef, V. Suriani, D. Nardi, D. D. Bloisi. A Deep
Learning Approach for Object Recognition with NAO Soccer Robots. RoboCup
2016: Robot World Cup XX - pages 392-403.
[2] G. Gemignani, M. Veloso, D. Nardi. Language-based sensing descriptors
for robot object grounding. 19th Annual RoboCup International Symposium,
2015. Science - Best Paper Award.
[3] F. Riccio, R. Capobianco, D. Nardi. Using Monte Carlo Search With Data
Aggregation to Improve Robot Soccer Policies. Proceedings of the 20th
International RoboCup Symposium, 2016.
[4] F. Riccio, E. Borzi, G. Gemignani, and D. Nardi. Multi-Robot Search for
a Moving Target: Integrating World Modeling, Task Assignment and Context,
2016. International Conference on Intelligent Robots and Systems (IROS’16).
IROS-RoboCup Best Paper Award.
[5] D. D. Bloisi, F. Del Duchetto, T. Manoni, V. Suriani. Machine Learning
for RealisticBall Detection in RoboCup SPL. arXiv:1707.03628
[6] F. Previtali, D. D. Bloisi, L. Iocchi. A distributed approach for
real-time multi-camera multiple object tracking - Machine Vision and
Applications, pages 1-10,2017.
[7] F. Previtali and G. Gemignani and L. Iocchi and D. Nardi.
Disambiguating Localization Symmetry through a Multi-Clustered Particle
Filtering, 2015, IEEE International Conference on Multisensor Fusion and
Integration for Intelligent
Systems, pages 283-288.
[8] E. Borzi. Multi-robot coordination under unreliable network conditions.
2015,Master thesis.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://mailman.cc.gatech.edu/pipermail/robocup-nao/attachments/20180514/d8fd4f3d/attachment.html>