Research and development of humanoid robots using physics-based simulation
[2017-11-01] Dr. Shinichiro Nakaoka from AIST visited UMIT Lab on October 31 and gave a talk on the use of physics-based simulation in research and development of humanoid robots.
The National Institute of Advanced Industrial Science and Technology (AIST) in Japan, conduct research and development of humanoid robots for work tasks dangerous to humans, e.g. clearance and rescue work in disaster areas. This requires robots capable of traversing cluttered terrain, manipulating the environment and operating different type of tools. The robots are autonomous or remotely operated by people.
In the talk, Shinichiro Nakaoka described how he and his co-workers at AIST use physics-based simulation for validation of control algorithms and motion patterns before they are transferred to the physical robots. This is important for achieving complex motion patterns, such as to pass obstacles - or even dance - without the risk of the robot falling over or otherwise be damaged.
The Chorenoid software developed at AIST supports high-level programming of humanoid robot movements in a similar way to tools for characters in animated film, i.e. by choreographing selected movements at specific times - manually or using motion capture of actors. The full movement is computed as an interpolation that takes into account the robot's characteristics and physics, and adapts to the surrounding environment.
Physics-based real-time simulation in combination with motion choreography has also proved very useful for remote operation of robots. The operator tests different maneuvers in a virtual environment, created using the data from the robot's vision sensors. The best motion pattern is selected and sent to the robot as a guiding instruction.
Dr. Nakaoka was in Umeå for work with Algoryx Simulation and held a lecture at Umeå University.
From left to right: Martin Servin, Claude Lacoursière, Ikumi Susa (VMC), Shinichiro Nakaoka (AIST), Anders Backman and Kenneth Bodin.
Link to news: