Since 1985, we have been working on biped robots. Our approach is characterized by following points.

First,	we don't stick to human type legs but mechanically optimal leg system.
Second,	patterns of leg motion are not prepared before walking but are generated in real-time.

So far, we have developed biped robots: Meltran I, II and III. Especially, Meltran II realized dynamic walk on flat floor, stairs walk with prior knowledge, sensor based stair walk without prior information. If you are interested in a biped robot which has wheels at the ends of its legs, please take a look here.

Meltran I This robot was developed at the very beginning of our project. Although it was designed for 3D dynamic walk, it had many mechanical problems and could not perform reliable walk.	Meltran II Our second biped robot was built after learning a lot from a trouble of Meltran I. In this page, I will introduce experiments of this robot.	Meltran III This is a robot to study lateral stability. It can perform side walk, but that's all Meltran III can do :-) A rate gyro is equipped on its body.

Highlights of walking experiment

	[Walk on a flat floor: m2mov.mov(835KB)] Walk of Meltran II. One step is 12cm and average walking speed is about 0.5km/h. The strategy of control is to keep body height constant at every moment of walking. This makes walking control simpler than former walking control algorithms.
	[Walk over a stair: us_step.mpg(337KB)] Meltran II is equipped with ultrasonic sensor so that the robot can detect the height of engaging floor. Without giving any prior information about stairs (location, height, steps etc.), the robot walk over stairs.
	[Walk over a obstacle us_obs.mpg(464KB)] Walk over a small square pipe. The robot decides walking pattern by itself. Please observe the robot adjust its steps in the front of the object.

Structure of Meltran II

Meltran II is a small robot for scientific experiment. The power source and computer is placed on the floor and connected to robot via umbilical cable. The robot has wide feet to obtain lateral stability. The robot can walk only forward and backward. It cannot do side walk, nor change its walking direction. You can see a finer drawing here.