So it was time for the Robotics Society of Japan’s annual meeting again. This time it was right in the middle of Tokyo, not a very scenic location. The event was hosted at Shibaura institute of technology as shown to the right. Not very big but good enough for the conference.
There are two items of research that caught my eye at the humanoid session. The first one by Mr. Urada Junichi focused on generating trajectories which would keep the robots body from rotating under large disturbances as shown below where the robot got kicked and had to step back but amazingly the body hardly rotated at all.
The second item by Yoshida Yuuki was research into adjusting the null space of a robot such that the COM stayed within the support polygon while the robot was getting pushed around. The figure below shows a commercially available Fujitsu HOAP robot moving one of its feet to prevent the COM from moving even though the robot is being pushed. Once the disturbance is removed the robot resumes its original position.
There were several demonstrations which I liked. One were these tinny motors with encoders. If you ever considered making a tiny robot, these might do the trick. Just make sure the power to weight works in your favor.
I wasnt aware that the robots from KAWADA industries could be controlled directly from OpenHRP but judging from their close relationship with the makers of OpenHRP, General Robotix, I shouldn’t be surprised. Here is a small video of the HIRO robot performing a simple motion while being controlled by OpenHRP. The communication between the robot and the simulator goes both ways so if you manually change the robots position the simulator will show the updated state.
Here is a small humanoid robot from Kondo the servo motor maker. Kondo’s specialization in low cost actuators makes it easy enough for them to make a humanoid robot. Its a simple machine but then again so is NAO. Nice thing about this robot is that its fully controllable from a remote control so actions like walking and turning can be controlled just as easily as controlling a remote control toy car. Neither robot has ground reaction sensors, thereby making them toys in my book but at least this robot is cheaper. If your looking for a serious research robot with a low cost I recommend the HOAP robot from Fujitsu, its got force sensors and cameras on its open architecture sensor network plus its made out of machined aluminum. Finally HOAP’s price tag is very roughly the same as NAO’s.
Ok here is a MicroMouse Contender from RT-Corp. MicroMouse is a event where wheeled robots are put inside a random maze and have to find their way to the end position. The robot shown here has three stages of learning, with each phase the robot become progressively faster. Unfortunately due to a error on my part I didn’t record the final stage but during this stage the robot is very fast it literally drifts around the corners. From what I understand the robot runs a SLAM algorithm on a standard processor board from Akihabara with a SH-72XX processor.
Here is a robot arm thats made from blocks. Its like lego for robots. Each block costs roughly $4000. Unfortunately the software doesn;t automatically update forward and inverse kinematics once a block is attached so its not quite plug and play.
Finally here is one of NAO’s friends called Darwin. It claims to be open architecture so thats a plus.