Recently, the 2019 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS 2019) held in Macau, China officially came to an end. The conference is the most famous and influential in the field of robotics and intelligent systems in the world One of the most powerful top academic conferences, held in Macau, China for the first time this year.
The theme of this year’s IROS conference is robots connecting people, including plenary meetings, keynote speeches, technical conferences, forums, seminars and tutorials, competitions, exhibitions, etc. There are about 4,000 leaders in the fields of robots, automation systems and artificial intelligence from all over the world. People, representatives of top research teams, and business people attended the meeting to jointly explore cutting-edge technologies in the field of intelligent robots and systems, and share and discuss the latest developments in related fields.
Some new products and forward-looking technologies appeared at this IROS conference to showcase the latest research and development results in the field of robotics. Professor Joseph Sung from the Chinese University of Hong Kong, China shared the latest application achievements of artificial intelligence in the medical field, especially the AI-assisted image detection for gastrointestinal cancer, data processing cancer classification, treatment plan recommendations, and automatic and Semi-automated surgical robots realize AI-assisted surgery and other research, which have greatly improved the treatment of patients. In this era of AI revolutionizing medical care, Professor Shen led everyone to examine and discuss the role of medical staff in the future, how to train the next generation of doctors, pharmacists and therapists, and who should be responsible for AI errors. Medicine brings changes and future trends.
Hangzhou Yushu Technology Co., Ltd. has brought the latest self-developed robot dog AlienGo, which integrates VSLAM, which can perform 3D mapping of terrain and obstacles in real time. Compared with lidar, it is more suitable for low-speed mobile robots, especially footed robots that need to observe the terrain. At the same time, AlienGo integrates human body 2D/3D somatosensory recognition and gesture recognition to facilitate human-computer interaction, such as automatic follow, human behavior prediction and other functions. It is reported that AlienGo is currently the world's largest size and largest weight electric-driven four-legged robot that can directly level the ground backflip.
Kuka, one of the world's leading industrial robot manufacturers, has brought two creative robot projects. The picture below shows one of them-the human-machine battle project.
The research vision of the Italian Institute of Technology (Istituto Italiano di Tecnologia, IIT) is to develop human science and technology and integrate the skills in the fields of robotics, nanomaterials, life technology, and computing science. Giorgio Metta, the scientific director of the institute, said on IROS 2019 TV: "Robots will play an important role in our future lives, such as more precise medical assistance, helping humans in factories, and saving humans in dangerous situations. Robots The invention is to empower humans, so the core is still humans. The surgical robots and post-operative rehabilitation robots we have developed will come out soon." The robots produced by the institute can be directly delivered to the market and are currently working to develop smarter robots. The cognitive ability of the robot.
The tactile and virtual reality laboratory of POSTECH in South Korea is guided by Professor SeungMoonChoi. Researchers study basic science, technology and applications related to tactile. Currently, the team is focusing on developing good technologies to help people make tactile content easier and faster. Research topics include tactile perception and multi-modal tactile rendering, with the goal of improving human life.
Sorin Mihai Grigorescu, an artificial intelligence expert in the ROVIS laboratory, introduced NeuroTrajectory (neural trajectory): It is a deep network of perception-planning for automatic vehicle positioning trajectory learning through neuroevolution. Currently, the control of autonomous vehicles is based on the sequence of decoupling perception-planning-behavior operations, or End2End or Deep Reinforcement Learning (DRL) systems, but due to the limitations of deep learning solutions for autonomous driving, NeuroTrajectory can solve this problem , This is a multi-objective neuroevolution method for positioning trajectory learning for autonomous driving, which estimates the desired state trajectory of an automatic vehicle within a limited prediction range through a perceptual planning deep neural network.
The University of California, San Diego (UCSD) laboratory brought robots that move completely on air without any electronic components.
At the seminar on deep probabilistic generative models of cognitive architecture in robotics, researchers from robotics and machine learning shared their knowledge about deep and probabilistic generative models to develop future robotic cognitive architectures. The seminar aims to study challenges and opportunities arising from interdisciplinary research fields including machine learning, cognitive science and robotics. Deep learning technology enables robots to recognize their own environment, such as visual and speech recognition, and can effectively learn behaviors, such as reinforcement and imitation learning. However, most of the success of deep learning depends to a large extent on the labeled data that needs to be prepared before the learning process or the hand-made reward function. The goal of this seminar is to share knowledge about the most advanced machine learning methods that will help model language-related capabilities in robots.
During this IROS conference, a drone competition was held. The competition is an engineering and computer science challenge. It is necessary to understand computer vision, develop the ability to include gate detection algorithms and programming logic so that the drone can understand when completing tasks . The participating teams need to program the drone to realize the autonomous navigation function of the drone. The drones of the participating teams need to automatically pass through a set of doors and complete as many times as possible within 5 minutes. The UAV of the University of Tsukuba in Japan completed 15 flights in 4 minutes and 47 seconds and won the championship.
