United States

Reinforcement learning is an active research area with a vast number of applications in robotics, and the RoboCup competition is an interesting environment for studying and evaluating reinforcement learning methods. A known difficulty in applying reinforcement learning to robotics is the high number of experience samples required, being the use of simulated environments for training the agents followed by transfer learning to real-world (sim-to-real) a viable path. This article introduces an open-source simulator for the IEEE Very Small Size Soccer and the Small Size League optimized for reinforcement learning experiments. We also propose a framework for creating OpenAI Gym environments with a set of benchmarks tasks for evaluating single-agent and multi-agent robot soccer skills. We then demonstrate the learning capabilities of two state-of-the-art reinforcement learning methods as well as their limitations in certain scenarios introduced in this framework. We believe this will make it easier for more teams to compete in these categories using end-to-end reinforcement learning approaches and further develop this research area.

RoboCup 2021

rSoccer: A Framework for Studying Reinforcement Learning in Small and VerySmall Size Robot Soccer

robot soccer

openai gym

simulation

continuous control

reinforcement learning

[![](https://assets.underline.io/uploads/markdown_image/1/image/21d13e03491d7e9acf6b8c46c7979c25.png)](https://docs.google.com/spreadsheets/d/1SCDDfkQk7es4J6RULKHVx1RzbB0Mutla--Sle5SVfb0/edit?usp=sharing)

**RoboCup 2021 will be Worldwide!**

We are excited to announce that RoboCup 2021 is a fully remote event with RoboCup competitions and activities taking place all over the world. Everyone is invited to participate in, or observe, the [competition leagues](https://2021.robocup.org/leagues) and the [symposium](https://2021.robocup.org/symposium)!

**RoboCup Symposium**

The 24th RoboCup International Symposium will be held in conjunction with RoboCup 2021 in a purely online setting. Scientific papers will be presented reporting innovative, original research with relevance to robotics and artificial intelligence. Within the described scope of topics we also encouraged submissions of high-quality overview articles, papers describing real-world research, and papers reporting theoretical results. In addition to the main track with regular research papers, there will be a special track focused on systems, data, and benchmarks featuring papers describing novel, open-source hardware and software systems, as well as datasets and benchmarks relevant to the community. In addition to research paper presentations, the RoboCup Symposium 2021 includes keynote talks by Dieter Fox, Jean-Paul Laumond, and Stefanie Tellex, and several interactive sessions.

**RoboCup Introduction Video**
<iframe width="995" height="516" src="https://www.youtube.com/embed/Cnh43cCsDKU" title="YouTube video player" frameborder="0" allow="accelerometer; autoplay; clipboard-write; encrypted-media; gyroscope; picture-in-picture" allowfullscreen></iframe>


Registration is required for entrance into RoboCup 2021. <br><br>
**Fees:** 
* Major Faculty or Symposium Faculty - US$50
* Major Student, Junior Mentor, Junior Student, or Symposium Student - US$20 

***Members of qualified teams:** Please contact your league for your registration password and proceed to register with your team. *<br>
***All non-team members:** Please register as "Symposium Only".*

You need to be registered to explore this event!

The 24th RoboCup International Symposium will be held in conjunction with RoboCup 2021 in a purely online setting. Scientific papers will be presented reporting innovative, original research with relevance to robotics and artificial intelligence. Within the described scope of topics we also encouraged submissions of high-quality overview articles, papers describing real-world research, and papers reporting theoretical results. In addition to the main track with regular research papers, there will be a special track focused on systems, data, and benchmarks featuring papers describing novel, open-source hardware and software systems, as well as datasets and benchmarks relevant to the community. In addition to research paper presentations, the RoboCup Symposium 2021 includes keynote talks by Dieter Fox, Jean-Paul Laumond, and Stefanie Tellex, and several interactive sessions.

poster

A long term goal in robotics is lifelong learning, in which a robot learns to perform new tasks without human intervention. For this a promising alternative is reinforcement learning (RL), in which an agent learns how to solve sequential decision-making problems by interacting with its environment. Despite the effectiveness of RL, learning in a trial-and-error way can be infeasible in robotics applications, in which long training periods can significantly deteriorate hardware. For robots that face long sequences of tasks, transfer learning represents an alternative to reduce training time, by transferring knowledge from previously learned tasks. However, in domains such as service robotics, identifying reusable pieces of knowledge and transferring them across tasks can be challenging, due to the mismatch between the state-action spaces (heterogeneous tasks). Thus, in this article we present an inter-task similarity measure and a transfer learning method for heterogeneous tasks. Through experimental evaluations, we show that the proposed measure is able to rank tasks in a way that prioritizes those that cause a larger increment of the learning performance in a target task. Additionally, results show that for some tasks the similarity measure can be computed with few data, enabling significant speedups in the learning process by transferring knowledge at an early stage.

Inter-Task Similarity Measure for Heterogeneous Tasks

The task of 6D object pose estimation from RGB images is an important requirement for autonomous service robots to be able to interact with the real world.
In this work, we present a two-step pipeline for estimating the 6 DoF translation and orientation of known objects. Keypoints and Part Affinity Fields (PAFs) are predicted from the input image adopting the OpenPose CNN architecture from human pose estimation.
Object poses are then calculated from 2D-3D correspondences between detected and model keypoints via the PnP-RANSAC algorithm.
The proposed approach is evaluated on the YCB-Video dataset and achieves accuracy on par with recent methods from the literature.
Using PAFs to assemble detected keypoints into object instances proves advantageous over only using heatmaps. Models trained to predict keypoints of a single object class perform significantly better than models trained for several classes.

6D Object Pose Estimation using Keypoints and Part Affinity Fields

Pose estimation commonly refers to computer vision methods that recognize people's body postures in images or videos. With recent advancements in deep learning, we now have compelling models to tackle the problem in real-time. Since these models are usually designed for human images, one needs to adapt existing models to work on other creatures, including robots. This paper examines different state-of-the-art pose estimation models and proposes a lightweight model that can work in real-time on humanoid robots in the RoboCup Humanoid League environment. Additionally, we present a novel dataset called the HumanoidRobotPose dataset. The results of this work have the potential to enable many advanced behaviors for soccer-playing robots.

Real-time Pose Estimation from Images for Multiple Humanoid Robots

The design and implementation of behaviors for robots operating in dynamic and complex environments are becoming mandatory in nowadays applications. Reinforcement learning is consistently showing remarkable results in learning effective action policies and in achieving super-human performance in various tasks -- without exploiting prior knowledge. However, in robotics, the use of purely learning-based techniques is still subject to strong limitations. Foremost, sample efficiency. Such techniques, in fact, are known to require large training datasets, and long training sessions, in order to develop effective action policies. Hence in this paper, to alleviate such constraint, and to allow learning in such robotic scenarios, we introduce SErP (Sample Efficient robot Policies), an iterative algorithm to improve the sample-efficiency of learning algorithms. SErP exploits a sub-optimal planner (here implemented with a monitor-replanning algorithm) to lead the exploration of the learning agent through its initial iterations. Intuitively, SErP exploits the planner as an expert in order to enable focused exploration and to avoid portions of the search space that are not effective to solve the task of the robot. Finally, to confirm our insights and to show the improvements that SErP carries with, we report the results obtained in two different robotic scenarios: (1) a cartpole scenario and (2) a soccer-robots scenario within the RoboCup@Soccer SPL environment.


Improving Sample Efficiency in Behavior Learning by Using Sub-optimal Planners for Robots

This proposal addresses the creation of educational tools to help
students with the concept of ‘localization’ of a mobile robot using
the extended Kalman filter (EKF) and landmarks in the
environment.

Learning tools for mobile robots localization using visual landmarks and the extended Kalman filter

Multi-robot coordination is one of the fields where RoboCup has consistently provided conditions that promote state-of-the-art advances. Still, it is common to see many teams achieving autonomy by applying hand-crafted rule-based plans. Formal approaches have produced optimistic results when addressing this issue but currently lack available software tools. In this project, we developed a MATLAB® toolbox that provides, through a user-friendly API, methods to model multi-robot coordination problems as generalized stochastic Petri nets with rewards (GSPNRs). This formalism provides a compact way of capturing: action selection, uncertainty on the duration of action execution, and team goals. This toolbox also implements algorithms that allow extracting optimal policies for the GSPNR model, while reasoning over uncertainty and optimizing team-level objectives. Furthermore, it integrates with ROS middleware enabling the execution of real multi-robot systems. We strongly believe that this could significantly benefit the RoboCup initiative as it facilitates the use of state-of-the-art formal methods that enable the improvement of autonomous behaviors in aspects such as efficiency, robustness, reliability, safety, and predictability.

Software Package to Model, Plan, and Execute Multi-Robot Tasks

Nowadays, the advancement of the mobile robot is also factored in the development of a world surrounded by technologies. Reinforcement Learning is used to enable the system to operate automatically, thus the robot will learn the next movement based on the interaction between the agent and the environment. The main contribution of this paper is to implement Soft Actor-Critic (SAC) to control mobile robot navigation in a 2D simulation. To evaluate the system developed using a 2D simulation with a trained SAC agent. According to this simulation, the process shows that SAC is successful to navigate the mobile robot from an initial to the target location and also can avoid obstacles in the environment.

Soft Actor Critic Reinforcement Learning for Autonomous Mobile Robot Navigation

Deep learning approaches have become the standard solution to many problems in computer vision and robotics, but obtaining sufficient training data in high enough quality is challenging, as human labor is error prone, time consuming, and expensive. Solutions based on simulation have become more popular in recent years, but the gap between simulation and reality is still a major issue. In this paper, we introduce a novel method for augmenting synthetic image data through unsupervised image-to-image translation by applying the style of real world images to simulated images with open source frameworks. The generated dataset is combined with conventional augmentation methods and is then applied to a neural network model running in real-time on autonomous soccer robots. Our evaluation shows a significant improvement compared to models trained on images generated entirely in simulation.

Closing the Reality Gap with Unsupervised Sim-to-Real Image Translation

When producing a model to object detection in a specific context, the first obstacle is to have a dataset labeling the desired classes. In RoboCup, some leagues already have more than one dataset to train and evaluate a model. However, in the Small Size League (SSL), there is not such dataset available yet. This paper presents an open-source dataset to be used as a benchmark for real-time object detection in SSL. This work also presented a pipeline to train, deploy, and evaluate Convolutional Neural Networks (CNNs) models in a low-power embedded system. This pipeline was used to evaluate the proposed dataset with state-of-art optimized models. In this dataset, the MobileNet SSD v1 achieves 44.88% AP (68.81% AP50) at 94 Frames Per Second (FPS) while running on an SSL robot.

Dataset and Benchmarking of Real-Time Embedded Object Detection for RoboCup SSL

We show the advantages of using Swift as the programming language for behaviour on the Pepper and Nao robots as used with the RoboCup Standard Platform Leage and the RoboCup@Home - Social Standard Platform. We show that Swift is not only incoporating modern features of Object-Oriented programming and functional programming, but is also now a stable systems programming language that enables both high-level development as well as fine hardware control.

Premium content

rSoccer: A Framework for Studying Reinforcement Learning in Small and VerySmall Size Robot Soccer

Downloads

Next from RoboCup 2021

Inter-Task Similarity Measure for Heterogeneous Tasks

Similar lecture

Soft Actor Critic Reinforcement Learning for Autonomous Mobile Robot Navigation