In the previous article, most of the projects we introduced came from the familiar industry giants, technology unicorns, academic and technical double ceilings... dazzling and seemingly out of reach. But don't forget that the development of the technical field is driven by the contributions of all participants in the community. Various robotics academic summits, such as ICRA and IROS, will be held in 2020, and amazing drone projects will be one after another. In addition, the power of civil society and enterprises cannot be ignored either. The drone industry is experiencing its second growth period. Open source and modularity are the focus of everyone’s attention; perhaps in the next few years, drones will become indispensable smart devices in people’s lives like smart phones. , And who can deny that we are not in the first year of an industry?
6. Prometheus-United Kingdom University Enterprise Joint Project
Some people will mistakenly believe that scientific research developers or technical researchers are dismissive of science fiction movies. The fictional technologies that they show for visual effects are too advanced, dreamy, and impractical. But the significance of science fiction movies to technologists is different from that of historians to historians. They are not only entertainment or talks, but sometimes they can be lights and catalysts.
In the 2012 science fiction film "Prometheus", there is an impressive scene. When astronauts were exploring unknown alien relics, they released multiple spherical drones. These drones used autonomous flight exploration and laser scanning to build a three-dimensional model of the cave in the Prometheus spacecraft in real time. The audience who watched the movie for the first time at that time might not have realized that this would be a prequel to Alien. Director Ridley Scott probably didn't realize that this scene will inspire many drone developers.
Prometheus, born at an innovation seminar, is such a project. When the chief engineer of the British railway company Network Rail was talking about hoping to have a device that could drill into the ground and map the cave, an engineer from the University of Manchester blurted out: "Hey, isn't this the one in Prometheus?" The two people who have been through this film hit it off and started the first step of the project. Today, Prometheus is already a large-scale project led by Headlight AI, involving the University of Manchester, University of Bristol and Network Rail. It lasted for two years and cost 2.2 million pounds. Expect to revolutionize underground inspections in railways, utilities, construction and mining.
Prometheus-UK University Enterprise Joint Project
Today, Prometheus is already a large-scale project led by Headlight AI, involving the University of Manchester, University of Bristol and Network Rail. It lasted for two years and cost 2.2 million pounds. Expect to revolutionize underground inspections in railways, utilities, construction and mining.
Of course, using drones to explore caves seems nothing new. DARPA (United States Defense Advanced Research Projects Agency) also launched an "underground" challenge for this type of activity. However, it is not common for the scale and investment to reach such a scale, which may stimulate the dependence of other public projects on drone technology. For people in science and engineering, nothing is more exciting than realizing the technology in science fiction movies.
Project introduction: https://www.youtube.com/watch?v=x1fWbWDQWBc&t=11s
https://www.bbc.com/news/business-55191623
7. Prometheus autonomous drone open source project-Amu Lab
Coincidentally, just as more than one person entered NASA because of their love of "Star Wars", "Prometheus" certainly inspired more than just a group of people. Friends who have followed our Amu laboratory for a long time know that Amu's UAVs are divided into JCV series (basic frame), Z series (simple development) and P series (advanced development). However, it seems that few people ask me what the initials of these series represent, and I rarely have the opportunity to talk to others: "The P of the P series is of course the P of Prometheus!"
Prometheus is an open source project designed to provide mature platform solutions and development interfaces for the intelligent and autonomous flight of drones. The project is mainly maintained by the postdoctoral team of Beijing Institute of Technology, and the Amu Lab provides hardware matching and functional implementation. At the place where the project was established, the original goal was "to make Prometheus' drones come true!". As we all know, with the blessing of SLAM and RGBD, it is not uncommon to use drones to build maps of formal environments.
Prometheus autonomous drone open source project-Amu Lab
However, the development of the Prometheus project completely exceeded the original vision of the founder. Under the basic concept of open source and modular design, the Prometheus open source project is iteratively optimized with the contributions of community participants all the time. And this year, the Prometheus project not only successfully realized the drone cluster formation under GPS, UWB and motion capture devices, but also comprehensively updated and upgraded the current P series drones. While maintaining the existing expansion space, the new P300/P600 focuses on enhancing the load and endurance, and provides the hardware foundation for the next new functions and new components.
Project homepage:
https://github.com/amov-lab/Prometheus
8. EVDodgeNet uses event cameras to avoid dynamic obstacles-University of Zurich
Although the University of Zurich feels more "literary" than ETH Zurich, they also have a super-powerful drone team. After setting up the laboratory in 2012, Professor Davide Scaramuzza has been working on the new field of "Event Camera" and has produced papers every year. When the event camera works, it is not recording a scene, but recording a change in the scene, and responds to changes in light and darkness of each pixel, which means that it is more capable of responding to fast-moving objects and sudden times than traditional cameras. reaction. More importantly, as a way to provide dynamic response to drones, its cost is very low.
The latest result published by Professor Davide Scaramuzza at ICRA2020 is to use event cameras to realize drones against dynamic obstacles. For example, the real-time obstacle avoidance of a fast flying baseball. After deep learning using a series of shallow neural networks, drones can avoid dynamic obstacles through onboard computers and event cameras. Its dynamic effect is so silky and witty...More importantly, it is hard to imagine how low its cost is.
Video address: https://www.bilibili.com/video/BV1EX4y1T7uk/
The application of event camera to drones is a new thing. Since it does not output traditional image information, the various algorithms of Open CV are not related to it, which is a limitation. However, from an academic point of view, if FAST feature points, optical flow, VIO, SLAM, and semantic segmentation are all pulled to the event camera and run, wouldn’t it be a new article after another?
Then again, Professor Davide Scaramuzza’s laboratory has really done a lot of new activities this year, from deep learning-driven drone acrobatics and autonomous racing, to new application directions for event cameras... Limmat Bank (UZH and Where ETHZ is located) is really a treasure of drones.
Article link: https://arxiv.org/abs/1906.02919
Project link: http://prg.cs.umd.edu/EVDodgeNet
Homepage of the Robotics and Perception Group of Zurich University: http://rpg.ifi.uzh.ch/index.html
9.CIMON-2 Space Station Smart Assistant-European Space Agency
Some people may remember that NASA sent a robotic assistant "Astrobee" to the space station last year to help astronauts handle some daily observation and maintenance tasks. This cube propelled by an electric fan seems more in line with our vision of future science fiction: a robot assistant that can freely float and move.
Since they all work in one place, ESA is naturally unwilling to lag behind. After sending the smart assistant CIMON to the space station to complete the test last year, I couldn't wait to send CIMON-2 too. The task of the intelligent assistant is very clear, providing task assistance and status monitoring for human astronauts. At the same time, CIMON seems to be more focused on human-computer interaction, it can be controlled by voice, and it has a big...big face that can chat with astronauts to relieve mental stress.
CIMON-2 Space Station Intelligent Assistant-European Space Agency
Some people say it is not a drone...Hey! It is also driven by fan blades! Not to mention that its voice interaction is the direction in which the drone can be propelled, although the noise of the drone on the ground is still a bit loud when it is running. The future application of drones in microgravity environments is also an interesting direction, although few people have come into contact with this direction yet...Oh yes, CIMON is running on Ubuntu, although there is no information showing that it and Robonaut 2 The same is developed based on ROS.
Astrobee introduction: https://www.nasa.gov/astrobee
Introduction to CIMON: https://en.wikipedia.org/wiki/Cimon_(robot)
10. Autonomous and Resilient Exploration Project of UAV Open Source Underground Environment-Norwegian University of Science and Technology
If the British Prometheus project described above is only in the planning stage (a lot of money has been raised), then the Autonomous Robotics Laboratory of the Norwegian University of Science and Technology has moved Prometheus' drones into reality. And even more ruthlessly, they have already open sourced it on ICRA2020.
This project is the Norwegian University of Science and Technology's response to the DARPA Underground Challenge. As the latest achievement, Charlie aerial robot scouts integrate and fuse LiDAR, vision, thermal and inertial data to perform self-positioning and draw environmental maps in a multi-modal manner.
Video address: https://www.bilibili.com/video/BV1AK411g7GA/
Two algorithms are used for autonomous path planning of UAVs, namely image-based exploration path planning (GBPlanner) and motion primitive-based exploration path planning (MBPlanner), both of which are released in the form of ROS function packages.
Now that Prometheus' drones are already there, it shouldn't be too far to find the alien ruins, right?
Project homepage: https://www.autonomousrobotslab.com/subtplanning.html
GBPlanner homepage: https://github.com/unr-arl/gbplanner_ros
MBPlanner homepage: https://github.com/unr-arl/mbplanner_ros