Introducing Autonomous Parking in Duckietown Cities

Posted on May 27, 2024 | by Duckietown Admin

Introducing Autonomous Parking in Duckietown Cities

Project Resources

Why Autonomous Parking?

Parking is notoriously a hard task to master for many humans. Hence, students of the Autonomous Mobility on Demand course at ETH Zurich wanted to determine to what degree this applied to autonomous parking with Duckiebots.

The goal of the Autonomous Parking project was to design, implement, and test a complete autonomous parking solution compliant with the Duckietown ecosystem.

Duckiebots should be able to enter and exit a parking area, identify viable parking lots, actually park and exit their parking spot safely, and avoid collision with other Duckiebots during the entire process.

The vision is to integrate autonomous charging solutions into the parking area, so Duckiebots can charge themselves when needed.

Autonomous parking in Duckietown: the challenges

Leveraging the Duckietown lane following vision baseline provided a basic infrastructure to build upon.

Some technical challenges specific to this projects were:

Backward Lane Following: Duckiebots must drive backward to exit the parking lots but only have cameras on the front. It is required to adjust the Duckiebot’s control system for stable backward driving, by changing the pose estimation process and re-tuning the PID controller.

Dynamic Color Adaptation: the new parking lot design introduced additional appearance specifications to the Duckietown city setup, such as blue lines identifying parking areas. Modifying the Duckiebots’ native lane detector to recognize blue lines in addition to yellow, red, and white, allows for additional flexibility in lane following based on specified colors.

Time Slot Coordination: Managing the availability of parking spaces is crucial to minimize the probability of collisions between Duckiebots. This project tackled this challenge by implementing a time-slot system to manage parking exits to prevent collisions, using red LEDs for signaling to other Duckiebots.

Project Highlights

Here is a visual tour of the work of the authors.

Check out the documents for more details!

Autonomous parking lot with Duckiebots in Duckietown — Resulting parking lot area configuration within Duckietown. The parking lot is compliant with Duckietown appearance specifications, and features one entry/exit intersection. The parking lot closed-loop and modular configuration With the goal of enabling Duckiebots to autonomously enter, park, and exit parking spaces while avoiding collisions. This design exemplifies the project's focus on creating a functional and adaptable parking solution.

Figure 2 illustrates a conceptual design of a parking area within the Duckietown environment. Comprising various standardized tiles, this depiction showcases a modular layout suitable for accommodating Duckiebots. The design includes six straight tiles, four curved left tiles, one three-way center tile, and two empty tiles. These components are arranged to form a configurable parking space that can be adapted to fit different spatial constraints. Notably, the configuration incorporates a T-intersection positioned such that Duckiebots are required to execute specific maneuvers upon entry and exit, ensuring adherence to the predetermined functionality of the parking area. — Design of a parking area within the Duckietown environment. Comprising various standardized tiles, this example showcases a modular layout suitable for accommodating Duckiebots. The design includes six straight tiles, four curved left tiles, one three-way center tile, and two empty tiles. Notably, the configuration incorporates a T-intersection positioned such that Duckiebots are required to execute specific maneuvers upon entry and exit, ensuring adherence to the predetermined functionality of the parking area.

Autonomous parking in Duckietown, parking lot specifications — Parking spaces are delimited by a blue line with specified geometry. This configuration is functional to allow the Duckietown lane filter to determine the pose of the robot and facilitate precise entrance and exit from the parking space.

Project Parking Results

(Turn on the sound for best experience!)

Project Authors

Trevor Phillips is a former Duckietown student, now a Machine Learning SWE at Apple in Switzerland.

Vincenzo Polizzi is a former Duckietown student, now a Ph. D. student at the University of Toronto, Canada.

Linus Lingg is a former Duckietown student, now the Co-Founder and CTO of bottleplus in Switzerland.

Learn more

Duckietown is a modular, customizable and state-of-the-art platform for creating and disseminating robotics and AI learning experiences.

It is designed to teach, learn, and do research: from exploring the fundamentals of computer science and automation to pushing the boundaries of knowledge.

Dino Claro: a Duckietown journey from project to thesis

Posted on May 24, 2024 | by Duckietown Admin

Dino Claro's Duckietown journey: from project to graduate thesis

Dino Claro, a mechanical and mechatronics engineering graduate from the University of Cape Town, shares his Duckietown journey: with challenges and results.

Cape Town, February 13th, 2024: Dino Claro, Graduate Mechanical and Mechatronics Engineer at the University of Cape Town, shares his experience with Duckietown and the project he developed using Duckiebots for his masters thesis.

Quick links

Duckinator: an Odometry pose-estimation for the Duckiebot robotic car platform

Hello and welcome Mr. Dino Claro! Could you introduce yourself?

My name is Dino Claro and I’m a Graduate Mechanical and Mechatronics Engineer at the University of Cape Town.

Thanks for accepting to share your experience with us. When did you first run into Duckietown?

During vacation work at the University of Cape Town (UCT) Mechatronic Systems Group, I was given the open-ended task of estimating the pose of a robot car. The goal of the vacation work was to solve a problem independently but also free from the
stresses of receiving a mark or grade. There was no expectation for novel work. In fact, the vacation work was only two weeks, and the expected solution would have been straightforward, probably odometry-based. Thus, Duckinator was born.

That's when you decided to use Duckietown?

With two platforms available, a basic Arduino 4WD kit and the Duckiebot, I could simply not resist the Duckies’ pull. The idea of using a Linux-based platform geared toward AI was extremely exciting.

At the end of the two-week vacation work, I was still ploughing through Duckietown documentation, the EdX: Self-Driving Cars with Duckietown MOOC, and ROS tutorials. My pose estimation solution seemed very far down the road. At that point, I should have realised that the DB (besides its cute exterior) is nuanced, to say the least.

Could you describe us your project?

The early phase of my project was extremely rudimentary. I had only had a couple of weeks during the vacation work to play with the DB [Duckiebot]. I planned to continue with the EdX MOOC [Self-Driving Cars with Duckietown, 2023 edition] while researching Docker and ROS on the side for the first couple of weeks and then begin development. A pitfall with this technique was completing a section of the MOOC or some other tutorial and believing I could implement it myself. My initial thinking was that if the MOOC could be completed in 10 weeks or so and given that I have already a couple of weeks’ headstart due to vacation work, I should be able to implement my standalone autonomous solution for the DB in the 12-week frame.

Spoiler alert, Duckinator did not rival Tesla. I made the realisation about 4 weeks into the project. At that stage, I was in the Object Detection activity of the MOOC. With the world in a frenzy over AI and ML, I was itching to dip my hands in some of this mysterious ML stuff.

Dr. Pretorius obliged, and my plan from this point was to implement my own standalone Duckietown-compliant Docker image for the YOLOv5. Charged with the excitement of the new project direction, I began researching ML, computer vision algorithms and YOLO itself. Implementing the YOLOv5 model was relatively smooth sailing and I loved learning computer vision. In all honesty, my YOLOv5 model was just organising the Object Detection MOOC into a standalone Docker image as the MOOC hides the Docker image from the student. I obtained the training data using the MOOC helper files and then trained the YOLOv5 model using a very similar Google Colab script as provided by the MOOC.

I slightly extended the YOLOv5 model from the MOOC by training the model to detect DBs, which proved to be sort of successful. As I only had one Duckiebot, I tested the model by parking Duckinator in front of a mirror or putting it in front of my laptop showing photos of other DBs. Due to this shabby testing, I left this extension out of my write-up. This was all completed after week 7.

With the world in a frenzy over AI and ML, I was itching to dip my hands in some of this mysterious ML stuff.

Dino Claro

Did you meet your objectives?

Completing the Object Detection model effectively meant that my revised project brief had been met but as I still had some time, I needed to extend the model in some way.

Duckinator had eyes but I wanted to make it move … autonomously. I had the idea of creating a safety controller where the distance of objects from the duckiebot could be inferred using the predicted bound box and perspective geometry.

My theory went as follows: knowing the real-world size of all the objects the DB could detect and comparing this to the dimensions of the bounding box provided by the YOLO model, it would be possible to infer the depth of the object and this depth could then be used to base autonomous controller commands. This led me to research autonomous vehicle safety architectures/controllers and modern depth estimation algorithms. I soon realised that much more advanced autonomous architectures existed. For instance, modern autonomous vehicles fuse camera feeds, object detection models, kinematic models and various other sensors to generate vector or depth maps. The creation of these depth maps is extremely complex and a field of intense research.

What where the challenges you encountered during your project?

After coding up my algorithm for the projective projection algorithm, I obtained unexpected results. Negative in most cases. Describing my algorithm in more clarity to Dr. Pretorius, he made it clear that a simple projection perspective would not work in this case.

I was projecting everything from the camera image to the ground plane but of course, the duckies and any other objects do not exist solely on the ground plane. This being week 10 of the project, I had simply run out of time and had to scrap the projective perspective and had no time to implement any of the more complex algorithms out there.

I was devastated at the fact that Duckinator was not going to move.

Upon some reflection though, the YOLOv5 model was working quite well, and I had all this research about autonomous architectures and depth estimation. One of the autonomous architectures I researched was Braitenberg vehicles acting as, possibly, the simplest autonomous architecture. A basic Braitenberg controller was simple enough to implement and would mean once again Duckinator could move.

Bounding boxes were populated onto a black image and then divided into left and right region maps. These maps were then element-wise multiplied with a weight matrix to provide a scalar value which can be used for wheel commands. Using the ‘fear’ Braitenberg vehicle the DB would then steer away from any detected objects. Another realisation was that my project was experimental with one of the main goals being for it to act as a stepping stone for future projects.

At this stage (two weeks before my project was due) I was satisfied with a newly engineered aim: Evaluating the viability of the Duckietown platform at the undergraduate level by implementing an ML object detection model. The key outputs being the YOLOv5 model (Duckie Detector) as well as possible future projects and trajectories for students.

The real learning occurs when getting your hands dirty, experimenting and troubleshooting.

Dino Claro

What are your final considerations?

Reflecting on my journey from a complete beginner to a slightly more competent beginner, here’s my advice for those on a similar journey:

Begin with a blank Duckietown-compliant Docker image and dive into coding a demo, whether it’s based on my solution or another. Ultimately, the goal is to first understand the code and then attempt to recreate it without directly copying.

While documentation and EdX activities are useful in providing broad overviews and points of contact for debugging, relying solely on them may create a deceptive sense of competency.

The real learning occurs when getting your hands dirty, experimenting and troubleshooting.

Thank you very much for taking the time, we appreciated your story very much! Is there anything else you would like to add?

Yes, I would say that embracing the hands-on experience is key to understanding the platform and being immersed in the infectious ethos surrounding Duckietown.

On that note, I would like to express my gratitude to Dr. Pretorius for granting me the freedom to experiment and the opportunity to work with the Duckiebot. I eagerly await future projects and the growth of the Duckietown community

Learn more about Duckietown

Duckietown enables state-of-the-art robotics and AI learning experiences.

It is designed to help teach, learn, and do research: from exploring the fundamentals of computer science and automation to pushing the boundaries of human knowledge.

Tell us your story

Are you an instructor, learner, researcher or professional with a Duckietown story to tell?

Reach out to us!

Massachusettes Institute of Technology: first BCI hackathon

Posted on May 20, 2024 | by Duckietown Admin

Cambridge, MA, USA – McGovern Institute, 24-25 February 2024: Over 100 participants took part to the first MIT BCI hackathon, competing in teams to control Duckiebots using brain computer interfaces.

Quick links

Controlling Duckiebots using brain computer interfaces

Over 100 participants gathered at the Massachusetts Institute of Technology for the first BCI hackathon organized by Dr. Federico Claudi. The participants tried to control a Duckiebot using only brain computer interfaces, and competed in a series of tasks.

BCI is the field of research that studies how to measure, amplify, filter and utilize electrical signals from the brain to interact with external devices.

What made this hackathon distinctive was the hands-on challenge, where participants were tasked with controlling a physical robot. This not only tested participants’ technical skills but also showcased their ability to tackle real-world problems through innovative BCI applications.

The task teams competed on was having Duckiebots (DB21-J4) navigate a road loop as fast as possible while avoiding Duckies. Here is an example:

The hardware used in this competition was an X.on EEG headset, and Duckiebots for control. Also, the winning team’s solution will be soon made available as a reproducible Learning Experience with Duckietown – stay tuned!

The Duckiebot is a DIY, Raspberry Pi-based robot powered by Nvidia and designed for introducing learners to autonomous technologies.

If you would like to contribute in developing accessible BCI LXs with Duckietown, and support the dissemination of BCI research, e.g. reach out to us at [email protected].

Learn more about Duckietown

The Duckietown platform enables state-of-the-art robotics and AI learning experiences.

It is designed to help teach, learn, and do research: from exploring the fundamentals of computer science and automation to pushing the boundaries of human knowledge.

Safe Reinforcement Learning (RL) Thesis Project

Posted on May 17, 2024 | by Duckietown Admin

Safe Reinforcement Learning (Safe-RL) in Duckietown

Project Resources

Safe-Reinforcement Learning (Safe-RL): Project Description

Safe-RL Duckietown Project – In his thesis titled “Safe-RL-Duckietown“, Jan Steinmüller used safe reinforcement learning to train Duckiebots to follow a lane while keeping said robots safe during training.

Safe Reinforcement Learning involves learning policies that maximize expected returns while ensuring reasonable system performance and adhering to safety constraints throughout both the learning and deployment phases. Reinforcement learning is a machine learning paradigm where agents learn to make decisions by maximizing cumulative rewards through interaction with an environment, without the necessity for training data or models.

The final result was a trained agent capable of following lanes while avoiding unsafe positions.

This is an open source project, and can be reproduced and improved upon through the Duckietown platform.

Safe Reinforcement Learning: Project Highlights

Here is a visual tour of the work of the author.

Check out the documents for more details.

Implementation of the safe reinforcement learning (Safe-RL) Duckietown project — Implementation of the Safe-RL Duckietown project

safe reinforcement learning in Duckietown project: Safety Layer Description — Safety Layer Description

Process diagram of action selection and safety layer in the safe reinforcement learning project using Duckietown — Process diagram of action selection and safety layer

Results of the safe reinforcement learning (Safe-RL) Duckietown Project — Results of the Safe-RL Duckietown Project

Safe Reinforcement Learning: Results and Conclusions

Based on the results, it can be concluded that there is no disadvantage to using a safety layer when doing reinforcement learning since execution time is very similar. Moreover, the dramatically improved safety of the vehicle is helpful for the robot’s training as fewer actions with lower or even negative rewards will be executed. Because of this, reinforcement learning agents with safety layers learn faster and reduce the number of unsafe actions that are being executed.

Unfortunately, manual observation and intervention by the user were still necessary, however, the frequency was clearly reduced which further improved learning as the robots in testing did not know if an outside intervention was done which could result in an action being rewarded incorrectly.

It was also concluded that this project did not reach perfect safety with the implementation. Therefore a fully autonomous reinforcement learning training without any human intervention has not yet been achieved. A lot of improvement factors have been found that can further improve the safety and recovery rate. Additionally, some major problems which are not direct results of the reinforcement learning or safety layer have been identified.

These problems could be attempted to be fixed in different ways like improving the open source implementations of lane filter nodes or adding more sensors or cameras to the robot in order to extend the input data to the agent. Another area that was untouched during the research of this project was other vehicles inside the current lane. The safety layer could potentially be extended to also include safety features that should keep the robot safe from hitting other vehicles.

Read the full report here.

Project Author

Jan Steinmüller is a computer science student working in the computer networks and information security research group at Hochschule Bremen in Germany.

Dr. Amr Alanwar is an Assistant Professor at the Technical University of Munich (TUM).

Learn more

Duckietown is a modular, customizable and state-of-the-art platform for creating and disseminating robotics and AI learning experiences.

It is designed to teach, learn, and do research: from exploring the fundamentals of computer science and automation to pushing the boundaries of knowledge.

Vision-Based DRL Autonomous Driving Agent with Sim2Real Transfer

Posted on May 6, 2024 | by Duckietown Admin

General Information

Vision-based DRL Autonomous Driving Agent with Sim2Real Transfer
Dianzhao Li and Ostap Okhrin
Universität Dresden, Dresden
D. Li and O. Okhrin, "Vision-Based DRL Autonomous Driving Agent with Sim2Real Transfer," 2023 IEEE 26th International Conference on Intelligent Transportation Systems (ITSC), Bilbao, Spain, 2023, pp. 866-873, doi: 10.1109/ITSC57777.2023.10422677.

Vision-Based DRL Autonomous Driving Agent with Sim2Real Transfer

One way to obtain quick and cheap training data is to use simulation instead of real-world experiments. The question remains if the learnings of a simulation-trained agent apply to the real world. Sim2Real transfer is the field of research that studies this problem.

The challenge is particularly meaningful when using vision as the primary sensing capability for robots. Vision-based deep reinforcement learning (DRL) refers to a technique where ML agents, typically modeled as multi-layered neural networks, learn to “make decisions” directly from visual input.

The essence of RL is training robotic agents based on policies that reward desirable outcomes. This family of techniques typically leads to increased adaptability to operational scenarios.

To learn about RL and its place in the larger context of robot autonomy, check out the resources below.

Abstract

To achieve fully autonomous driving, vehicles must be capable of continuously performing various driving tasks, including lane keeping and car following, both of which are fundamental and well-studied driving ones. However, previous studies have mainly focused on individual tasks, and car following tasks have typically relied on complete leader-follower information to attain optimal performance.

To address this limitation, we propose a vision-based deep reinforcement learning (DRL) agent that can simultaneously perform lane-keeping and car-following maneuvers.

To evaluate the performance of our DRL agent, we compare it with a baseline controller and use various performance metrics for quantitative analysis. Furthermore, we conduct a real-world evaluation to demonstrate the Sim2Real transfer capability of the trained DRL agent.

To the best of our knowledge, our vision-based car following and lane-keeping agent with Sim2Real transfer capability is the first of its kind.

Highlights - Sim2Real transfer results

Here is a visual tour of the work of the authors. For all the details, check out the paper link.

Fig. 1. The proposed DRL framework for vision-based multi-task autonomous driving agents. The perception module leverages camera images to produce impact attributes regarding the environment, then the DRL control module utilizes the information to control the agent with enhanced generalization.

Duckiebot DB21 — Fig. 2. Robot car [Duckiebot] used during the real-world evaluation. (a) Side view of the robot car which equipped with front-view camera and Jetson Nano 2GB. (b) Back view of the car with a pattern of circles.

Fig. 3. An example velocity trajectory for the leading vehicle generated by the Ornstein-Uhlenbeck process for the training process.

Fig. 4. Training results of PPO agent with ten independent seeds over one million steps.

Fig. 5. Example trajectories of DRL agent (red) and baseline controller (green) following random leader trajectory (orange) during the evaluation.

Fig. 6. Distribution of TTC and time headway for DRL agent (red) and baseline controller (green).

Fig. 7. Example trajectories of DRL agent (red) and baseline controller (green) following a self defined external leader trajectory (orange) during the evaluation.

Conclusion

This study proposes a vision-based DRL agent that can simultaneously perform lane-keeping and car-following tasks.

The overall system is divided into two modules: the perception module and the control module. The perception module extracts task-relevant attributes of the surroundings, while the control module is a DRL agent that takes these attributes as input. To evaluate the performance of the DRL agent, we compare it with a baseline algorithm in both simulation and real-world environments.

In the simulation, we compare the car following and lane-keeping capabilities of the DRL agent and baseline controller using various performance metrics. In the real-world environment, we demonstrate that the DRL agent can follow the leading vehicle while maintaining lane-keeping ability.

In future work, we plan to enhance our DRL agent by incorporating a comfort factor to address unstable driving behavior. Additionally, we aim to deploy more advanced algorithms for improved generalization.

Project Authors

Dianzhao Li is a research assistant at the Technische Universität Dresden, Dresden, Germany.

Ostap Okhrin is Chair of Statistics and Econometrics at the Institute of Economics and Transport, School of Transportation, Technische Universitat Dresden in Germany.

Learn more

Duckietown is a platform for creating and disseminating robotics and AI learning experiences.

It is modular, customizable and state-of-the-art, and designed to teach, learn, and do research. From exploring the fundamentals of computer science and automation to pushing the boundaries of knowledge, Duckietown evolves with the skills of the user.

Anatidaephilia: centralized city-based SLAM (cSLAM)

Posted on April 26, 2024 | by Duckietown Admin

Anatidaephilia: centralized city-based SLAM (cSLAM)

Project Resources

cSLAM Project Description

Project cSLAM – Simultaneous Localization and Mapping (SLAM) is a successful approach for robots to estimate their position and orientation in the world they operate in, while at the same time creating a representation of their surroundings.

This project, centralized SLAM (or cSLAM), enables a Duckiebot to localize itself, while the watchtowers and Duckiebots work together to build a map of the city. The task is achieved by using the camera of the Duckiebot, together with watchtowers located along the path, to detect AprilTags attached to the tiles, the traffic signs, and the Duckiebot itself.

P. S. Anatidaephilia, is Latin for loving, and being addicted to, the idea that somewhere, somehow, a duck is watching you.

Project Highlights

Here is a visual tour of the work of the authors.

Check out the documents for more details!

Duckietown cSLAM algorithm physical architecture for localizing Duckiebots — cSLAM algorithm physical architecture for localizing Duckiebots. Watchtowers are traffic lights without lights, and are used to transform Duckietown in Autolabs.

Duckietown cSLAM logical architecture for merging sensor measurements from robots and the city.

Duckietown cSLAM graph optimizer representation — AprilTag detections from Watchtowers and Duckiebots are harmonized by solving a minimization problem.

RViz visualization of Autolab virtual twin in Duckietown cSLAM project — RViz reconstruction of experimental localization outcomes.

cSLAM Autolab watchtowers network diagnostic

cSLAM Project Results

(Turn on the sound for best experience!)

This work developed into a paper, check the article here.

Project Authors

Rohit Suri is a former Duckietown student, now a Roboticist at Venti Technologies in Singapore.

Aleksandar Petrov is a former Duckietown student, now a Ph. D. student at the University of Oxford.

Amaury Camus is a former Duckietown student, now a Lead Robotics engineer at Hydromea SA in Switzerland.

Francesco Milano is a former Duckietown student, now a Ph. D. student at ETH Zurich in Switzerland.

Benson Kuan is a former Duckietown student, now a Senior Robotics Research Engineer at DSO National Laboratories in Singapore.

Learn more

Duckietown is a modular, customizable and state-of-the-art platform for creating and disseminating robotics and AI learning experiences.

It is designed to teach, learn, and do research: from exploring the fundamentals of computer science and automation to pushing the boundaries of knowledge.

Learning Skills to Navigate without a Master: A Sequential Multi-Policy Reinforcement Learning Algorithm

Posted on April 4, 2024 | by Duckietown Admin

General Information

Learning Skills to Navigate without a Master: A Sequential Multi-Policy Reinforcement Learning Algorithm
Ambedkar Dukkipati, Rajarshi Banerjee, Ranga Shaarad Ayyagari, Dhaval Parmar Udaybhai
Indian Institute of Science
A. Dukkipati, R. Banerjee, R. S. Ayyagari and D. P. Udaybhai, "Learning Skills to Navigate without a Master: A Sequential Multi-Policy Reinforcement Learning Algorithm," 2022 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), Kyoto, Japan, 2022, pp. 2483-2489, doi: 10.1109/IROS47612.2022.9981607.

Learning Skills to Navigate without a Master: A Sequential Multi-Policy Reinforcement Learning Algorithm

Reinforcement learning (RL) is a rising star approach for developing autonomous robot agents. The essence of RL is training agents based on policies that reward desirable outcomes, which leads to increased adaptability to operational scenarios. Through iterations, robots refine their decision-making, optimizing actions based on rewards and penalties. This method provides robots with the flexibility to handle unpredictable situations, enhancing their efficiency and effectiveness in real-world tasks. To learn about RL with Duckietown, check out the resources below.

Abstract

Solving complex problems using reinforcement learning necessitates breaking down the problem into manageable tasks, and learning policies to solve these tasks. These policies, in turn, have to be controlled by a master policy that takes high-level decisions. Hence learning policies involves hierarchical decision structures. However, training such methods in practice may lead to poor generalization, with either sub-policies executing actions for too few time steps or devolving into a single policy altogether. In our work, we introduce an alternative approach to learn such skills sequentially without using an overarching hierarchical policy. We propose this method in the context of environments where a major component of the objective of a learning agent is to prolong the episode for as long as possible. We refer to our proposed method as Sequential Soft Option Critic. We demonstrate the utility of our approach on navigation and goal-based tasks in a flexible simulated 3D navigation environment that we have developed. We also show that our method outperforms prior methods such as Soft Actor-Critic and Soft Option Critic on various environments, including the Atari River Raid environment and the Gym-Duckietown self-driving car simulator.

Highlights

Here is a visual tour of the work of the authors.

For all the details, check out the paper link!

(a) An illustration of policies learned in our approach. (b) A representation of the state space partitioned by the termination functions of the options. Each oval corresponds to the set of states classified as non-termination states by the corresponding nested termination function. — Fig. 1. (a) An illustration of policies learned in our approach. Different policies learn different skills out of necessity to traverse the environment and not terminate the episode. The trajectory spawned by the sequence of policies that correctly learns to avoid terminating the episode is shown in green. The policy π1 is used when the agent is in the middle of the corridor. When the end of the corridor is reached, π2 is selected to take a turn and avoid a collision. (b) A representation of the state space partitioned by the termination functions of the options. Each oval corresponds to the set of states classified as non-termination states by the corresponding nested termination function.

Learning Skills to Navigate without a Master: A Sequential Multi-Policy Reinforcement Learning Algorithm image — Fig. 2. (a) & (b) The input to the neural network in our 3D navigation environment. These images obtained from the simulation are scaled (and also transformed to grayscale in the case of (a)) and appended with K previous images, and sent to the agent as input. (c) A view of the Duckietown environment.

Autonomous robot Navigation performance results obtained on various environments: SAC, SOC, SAC with HER — Fig. 3. The results obtained on various environments.

Outputs of the three learned policies when fed the corresponding input shown on the left. πω1 : orange, πω2 : blue, πω3 : green. Each policy outputs a Gaussian distribution, the active policy is the one filled with color. A positive value for the output action corresponds to turning right, and a negative value indicates a left turn. — Fig. 4. Outputs of the three learned policies when fed the corresponding input shown on the left. πω_1: orange, πω_2 : blue, πω_3 : green. Each policy outputs a Gaussian distribution, the active policy is the one filled with color. A positive value for the output action corresponds to turning right, and a negative value indicates a left turn.

Conclusion

In this paper, the authors proposed an algorithm called “Sequential Soft Option Critic” that allows adding new skills dynamically without the need for a higher-level master policy. This can be applicable to environments where a primary component of the objective is to prolong the episode. We show that this algorithm can be used to effectively incorporate diverse skills into an overall skill set, and it outperforms prior methods in several environments.

Learn more

Duckietown is a platform for creating and disseminating robotics and AI learning experiences.

Duckietown is modular, customizable and state-of-the-art. It is designed to teach, learn, and do research: from exploring the fundamentals of computer science and automation to pushing the boundaries of knowledge.

Successful failures: learning robotics with Prof. Bruegge

Posted on January 25, 2024 | by Duckietown Admin

Prof. Bruegge and his colleagues about his experience using Duckietown and the new 4GB Duckiebots.

Munich, April 3rd: Prof. Bruegge (Brügge) studied computer science at the University of Hamburg and Carnegie Mellon University (CMU), where he also earned his doctorate. Appointed as Professor at the Technical University of Munich (TUM) in 1997, and also associate professor at CMU, Prof. Brügge was on the research committee of Deutsch Telekom and the Munich district. He has served on the board of directors of the Center for Digital Technology and Management, and acted as liaison professor for the Max Weber Foundation and the German National Academic Foundation (2000–2017).

Quick links

Failing successfully, an interview with Prof. Bruegge

Prof. Bruegge and the JASS 2023 summer camp

Prof. Bruegge, Bernd, is an established computer scientist researcher and Professor, having (co)authored hundreds of peer reviewed publications with over 8800 citations at the time of this writing.

Emeritus Professor since 2021, Bernd regularly contributes to the organization and running of the Joint Advanced Student School (JASS).

In this interview we capture his experience from the summer 2022 edition.

We are thrilled to have this chat with him!

Good morning! Could you introduce yourself?

My name is Bernd Bruegge. I was a Full Professor of Computer Science at the Technical University in Munich, I was at Carnegie Mellon before I joined the Technical University. I was for 20 years at the Computer Science Department at CMU in Pittsburgh, Pennsylvania.

What can you tell us about this JASS school event that took place in Cyprus?

The recent [2023] school was quite an adventure!

My colleague, Prof. Kirill Krinkin, was in St. Petersburg. We had done all the JASS projects with Kirill and support activities by JetBrains in St. Petersburg. And as you know, it’s a total disaster there, there’s nothing left.

So we organized this school in Cyprus with 10 students from Munich and 10 from Cyprus and St. Petersburg. And once there, we built an advanced Duckiebot system, with focus on context sensitivity.

We looked not only at intersections and turns, but we also had times when the Duckiebots would have to slow down, or stop at repair sites where we had traffic lights. There were not only OCR codes, we used Thread and Matter. Are you familiar with Thread and Matter?

I call this a "successful failure". The challenge in using advanced technology in the real world is that when you explore edge cases, unexpected situations may arise that one would not have considered by just simulating the same scenarios.

Prof. Bernd Bruegge

Is this a new functionality you added to the system?

Yes, this is actually an add-on we made to Duckietown. It’s a new IP standard that allows to save energy much, much more than Bluetooth. It’s actually better by a factor of 10. We added Thread and Matter capabilities to our Duckietown environment.

That's fascinating! When did you first encounter Duckietown?

Well, I first heard of Duckietown a long time ago. As you know, I have been at Carnegie Mellon, which competed with MIT and Stanford. And when Kirill told me he would be a visiting professor at MIT, I asked: “What are you doing there?” And he said, “I’m using Duckietown!”

He was interested in robots and robotic technology and observed that this new technology is actually more like an environment and ecosystem than just a robot.

We then came up with a few ideas: first in St. Petersburg, we used a combination of drones and Duckiebots. Our scenario included two airports and a (duckie!) passenger. The drone had to pick up the duckie from one Duckiebot, and the students, divided into three teams, had to develop their own pickup mechanisms.

Each of them with different approaches. One used magnets, the other one used scoops, and so on. They had to then transport the duckie from airport A to airport B using indoor navigation GPS. There was our first really impressive demo.

Credits: Special thanks to Andreas Jung and Ruth Demmel, and the multimedia team supporting Prof. Bruegge.

Prof. Bruegge, do you feel you achieved your objectives?

I’d say we didn’t fulfill our technical objectives. The main challenge with Duckiebots was the third, omnidirectional, wheel, which had too much attrition. So, for instance, if we came to drive near a church, the idea would be to slow down the speed. But then we discovered it was difficult to control the Duckiebot at low speed, since when it comes down to a crawl, friction may cause it to stop. So operating at low speeds was not a practical possibility.

But I call this a “successful failure”. The challenge in using advanced technology in the real world is that when you explore edge cases, unexpected situations may arise that one would not have considered by just simulating the same scenarios.

I think the Duckiebot is good at lane-following and following traffic rules. When looking at context sensitivity though the challenges are trickier. So, e.g., we have a construction site with two traffic lights, how to coordinate them so to minimize, e.g., traffic? If you have one Duckiebot coming from one side and one coming from the other, then one traffic light should go green, and the other one should be red.

But then we also have the idea that you have one traffic site that consists of multiple repair sites. In that situation the Duckiebots, lining up at the red traffic light, should have a green wave to go through each of the traffic lights until they exit the last traffic light.

So what we had in mind was coordination, making sure the Duckiebots were talking to each other. We used Python and the Duckiebots with 4GB NVIDIA Jetson Nanos, and this happened during a five-day course.

Is there anything else you would like to add?

We plan to prepare a special room for students from primary schools, to show them the interplay of autonomous driving and context-sensitive or ubiquitous computing.

We would like to use the room in multiple ways. So we would like to have the Duckietown road sections, the tiles, interconnected in such a way that that allows us to switch between topographies basically in no time.

We have not been able to do this yet in an easy way, it’s always a scramble. Even if we use the same layout, the yellow tape is breaking or the white tapes are out of sync, and then we have to repair them.

So it would be great if we managed to keep the set up time for schools, especially for children (10 to 12 years), when doing exhibitions here, to under three minutes.

That's a great suggestion, thanks! Do you think Duckietown can be useful for younger learners too?

Yes, I think Duckietown can be very persuasive with young kids. They’re used to Lego or Fischer products, so when they meet the Duckiebot, that’s their first exposure to robots.

We actually have another project in mind, a series of Summer Schools. I’m a retired professor now, I don’t have to teach so I have time for Summer Schools! One will take place in the Dolomites, where the challenge will be a logistic one taking place in a factory. It will include robots moving packages around a warehouse.

Learn more about Duckietown

Duckietown enables state-of-the-art robotics and AI learning experiences.

It is designed to help teach, learn, and do research: from exploring the fundamentals of computer science and automation to pushing the boundaries of human knowledge.

Tell us your story

Are you an instructor, learner, researcher or professional with a Duckietown story to tell?

Reach out to us!

Spearheading Autonomy: Giulio Vaccari and the amazing AEA

Posted on January 25, 2024 | by Duckietown Admin

Giulio Vaccari tells us how he spearheaded Duckietown at the Milan Polytechnic through the AEA student association, to support autonomy research and networking.

Milan, February 29th: Giulio Vaccari, master student at the time of this interview, tells us how and why he spearheaded the use of Duckietown at the Polytechnic of Milan, one of the most important Italian engineering universities.

Quick links

Spearheading Autonomy: Giulio Vaccari and the amazing student Automation Engineering Association (AEA)

An interview with Giulio Vaccari from the Polytechnic University of Milan

Hello and welcome Mr. Giulio Vaccari! Could you introduce yourself?

My name is Giulio Vaccari, and I’m a master student at the Politecnico di Milano (“Polimi”). I am studying automation and control engineering and I am right now in the process of finishing my thesis with Duckietown. During my five years here, I founded the student Automation Engineering Association, focused on automation engineering. At the “Polimi” the course of control and automation engineer is not super big, now it’s getting bigger. More and more people are joining, but it’s still pretty small compared to, for example, computer science or other bigger courses. The idea of our association was to help with networking, help to understand what are future possible career paths as automation engineers. That’s why together with some friends we decided to found the association some years back.

Very interesting. How does Duckietown fit in all this?

It all started when I met and talked with Jacopo Tani and Vincenzo Polizzi I believe three years ago.

I had some friends, older friends, at ETH Zurich that introduced us to Duckietown as a possible project for our association, that at the time was still pretty small. We were newborn and we liked the idea of start working with Duckietown right away, because we were looking for a project that we could treat like a competition. It’s easier to get more people to know each other and network, when they have a common goal, as there are common topics to discuss.

I wanted something that would help us develop teamwork skills, something around which to build a strong team. Then this idea evolved a little bit and now we have a very focused and skilled team that works with aim of doing good at the Artificial Intelligence Driving Olympics (AI-DO). We did our first competition in 2021, in December, and it went well. Now we are improving. We are still working on it.

Could you tell us more about your team's experience with the AI-DO competition?

Sure! We ended up being finalist, I’m very proud of that! In the meantime, I talked with some professors to try to get some help to improve what we were doing with Duckietown, to have some suggestions on what we should focus on. I was lucky enough to meet a Professor, now my thesis advisor, who had already heard about Duckietown and liked it. I was very happy that we were able to bring Duckietown to the Politecnico of Milano. Although the AI-DO are over now, I’m still doing related work, so I’m able to help the team do better while working on my thesis. Duckietown is now part of the laboratory for autonomous driving here at PoliMi, and I feel like managing to get Duckietown to be adopted by the Politecnico di Milano is a pretty big thing that I’m still proud of!

And you should be! Do you use Duckietown also in the activities of the association?

When it comes to the association, we are focused on AI. We have a team of around 20 people working on different things. We use Duckietown also to do outreach when we have, for example, “associations days”, which are open days when students from high schools come and visit the University. That’s when we showcase Duckietown to show both what we do as an association but also what a control engineer is supposed to do, because it’s very difficult to explain otherwise.

What do you like about Duckietown? Why do you think it is useful for what you do?

Through Duckietown, it’s very easy to implement ideas in the real world. Usually if you need to build something like that from scratch, it will be super difficult. There are a lot of failure points, while with Duckietown we can just deploy it and it just works. This is super cool. We can also test a lot of different control strategies with very low effort. I mean, just the fact of writing the control algorithm, deploying it and it just works is huge. So this makes things super easy. Duckietown is very friendly and easy to use. It’s not something you often see in robotics, which is instead usually very complicated and intimidating. Yes, very intimidating. And Duckietown is even very friendly looking, and this is definitely a big plus!

Do you think people that are using Duckietown, are satisfied with it?

Yes, we had really positive feedback. I mean, everyone is really happy to work with Duckietown and we had so many people applying that we had to limit the team size because that would’ve made it very difficult to manage all those people. So I would say it had a very positive impact in general on the association. People are generally enthusiastic to just see the robots going around.

There are a lot of failure points in robotics, while with Duckietown we deploy our algorithms and it just works. This is super cool. We can test a lot of different control strategies with very low effort.

Giulio Vaccari

Would you suggest it to colleagues or other student associations?

Yes, for sure. But I see that it’s not only our team; also the Professor that I’m working with is very happy to be able to use it, also as a benchmark for different controls. There are other universities that have their very specific tracks to run their robots, while with Duckietown, it’s super easy to test like, the same strategy in different environments, but still with the same basics. This is a very positive thing also for the university.

Thank you very much for taking the time, we appreciated your story very much! Is there anything else you would like to add?

Right now I’m working on a different subject. So I’m using still Duckietown, but not in a city environment, but in a racing environment. My objective is to set up a racing track and have two robots compete against each other. They are still not competing, but we are getting close. It’s very nice that together with the Duckietown team, we are working on the urban environment, using different kind of materials to create these racing environments. I think it’s very cool.

There are also some issues every now and then. Maybe something that is not super easy to find, something that is very in depth, some APIs. If you want to use them, you need to dive deep. There is a support Slack channel though, and it’s super easy to ask for help. This too I believe is very nice and just adds to the positive experience.

Special Giulio Vaccari content!

We found in our archives the video submission of Giulio Vaccari and his team at AEA on the occasion of the 2021 AI Driving Olympics finals. In this fun, short, video team PoliMi explains the rationale and challenges faced in developing an ML-based agent for the 2021 competition!

Learn more about Duckietown

The Duckietown platform enables state-of-the-art robotics and AI learning experiences. It is designed to help teach, learn, and do research: from exploring the fundamentals of computer science and automation to pushing the boundaries of human knowledge.

Tell us your story

Are you an instructor, learner, researcher or professional with a Duckietown story to tell?

Reach out to us!

Driving autonomy education at BFH with Prof. Affolter

Posted on November 24, 2023 | by Duckietown Admin

Bern, Switzerland: Prof. Peter Affolter, Head of the automotive department at Berner Fachhochschule (BFH), shares his experience creating a curriculum in autonomy education with Duckietown. Quick links Bern University of Applied Science Prof. Peter Affolter @ BFH Prof. Peter Affolter Driving autonomy

Bern, Switzerland: Prof. Peter Affolter, Head of the automotive department at Berner Fachhochschule (BFH), shares his experience creating a curriculum in autonomy education with Duckietown.

Quick links

Driving autonomy education with Duckietown

Prof. Affolter at the Bern University of Applied Sciences introduces autonomy education in the automotive curriculum

Hi! Thank you for taking the time to chat with us. I'd like to start by asking you to introduce yourself, and tell us who you are and what you do.

Certainly! My name is Peter Affolter. I am at the head of the automotive department at the Bern University of Applied Science. I work as a lecturer in autonomous driving, vehicle communications, and related areas.

How did you get to know Duckietown?

The first time I saw it was probably two or three years ago. I came across it by chance while Googling on the internet. I was looking for information about lessons and classes on autonomous driving, a very complicated topic. I was searching for an easy platform, especially hardware, to simulate a simple vehicle. Then, I stumbled upon Duckietown. Initially, it looked a bit childish, but I was surprised by its concept, quality, and comprehensive documentation, including lessons slides for this complex theory. I decided to focus on Duckietown, especially on the massive online open course hosted on edX [Self-Driving Cars with Duckietown]. I also really enjoyed the great support from all the stakeholders and universities involved in this initiative.

Did you find the "Self-Driving Cars with Duckietown", the massive open online course (MOOC) on edX, helpful?

Yes, at that time, there was only the archive available on edX.org, and I registered there. I couldn’t participate in real-time, but I went through the materials myself. I also contacted the Duckietown staff, and they provided me with support and missing documents. I adapted the content for my own class, focusing on bachelor-level students, parallel to the massive online open course from edX.

Very interesting, so you adapted the materials for your course. How did you set up the class, did you find any challenges?

It was clear to me that I could introduce our students, future automotive engineers, to autonomous driving. I wanted to give them an overview of the potential, limits, obstacles, and areas needing progress. Since they were not programmers or IT technologists, I needed to simplify it. I decided to conduct whole-class courses using video material from the MOOC. I adapted the quizzes and provided them with virtual machines, avoiding Linux and network issues. They could use the vehicles hosted on the BFH server farm through a browser terminal, even utilizing the Duckiebots. It became an infrastructure for exercises from simulation to reality, based on Duckietown classes.

How did the teaching experience go?

The first class had potential for improvement, but we adopted a flipped classroom approach with four lessons a week for 16 weeks. Students studied theory at home, and class time was for exercises, discussions, and additional explanations. The students enjoyed it, and while they didn’t grasp every detail, they felt proud and motivated. There’s still plenty of room for improvement in the next classes. They grasped the feeling and had the chance to work on neural networks, image manipulations, and more. I made good progress, and there’s still plenty of material to explore and implement in future classes.

Honestly, I haven't found another hardware platform as good as Duckietown for my needs. It's a simple platform with a fun approach, well-documented, and with a really reactive community. Even compared to other commercial products, Duckieown stands out. It fulfills all the needs from beginners to experts.

Prof. Peter Affolter, Head of Automotive Dept. at BFH

What were your overall impressions introducing autonomy education to your students with Duckietown?

Honestly, I haven’t found another hardware platform as good as Duckietown for my needs. It’s a simple platform with a fun approach, well-documented, and with a really reactive community. Even compared to other commercial products, Duckietown stands out. It fulfills all the needs from beginners to experts.

Thank you again for sharing your experience with us. Best of luck in your next classes!

Learn more about Duckietown

Tell us your story

Are you an instructor, learner, researcher or professional with a Duckietown story to tell?

Reach out to us!