Gecko-Inspired Window Washing Robot is Powered Entirely by Water
Batteries and motors are heavy and inefficient in that they expend a significant percentage of their power just moving their own mass. This is especially apparent in climbing robots, which spend most of their time hoisting themselves vertically upward. Researchers from Zhejiang University in China have developed a robot that's capable of sticking to smooth surfaces, climbing vertically, and washing windows, relying almost entirely on water pressure:
To function, the robot gets connected to a faucet with a loop (a really long loop, if necessary) of hose. As water flows through the hose, its pressure accomplishes several things. First, the water passes through fluidic vacuum generators, which use that same Bernoulli principle that those supersonic jet grippers take advantage of to turn the motion of a fluid into a vacuum. This allows the bot's feet to stick to any smooth surface.
Then, the water is routed through a solenoid valve to a piston that's attached to the "spine" of the robot. The inspiration for this design was the gecko, arguably the best wall-climber in existence, and the upshot of it is that the robot can climb relatively quickly (constrained only by the time it takes to establish a solid vacuum) and turn in either direction with just one single spinal actuator. And of course lastly, the water is squirted out at the end of the robot's arm to do the actual washing.
The robot does currently use a very small battery to power the wireless communication system and to trip the servo to control the direction of motion, but it's certainly possible that a small turbine could run all that stuff instead. The present design is able to lift twice its body weight in payload using just standard tap water pressure, and future versions might be able to conduct inspections, fight fires, paint, or even perform repairs.
This robot was presented in an ICRA paper entitled "A Gecko Inspired Fluid Driven Climbing Robot," by Jilin Liu, Zhangqian Tong, Jinyuan Fu, Donghai Wang, Qi Su, and Jun Zou of the Institute of Mechatronic Control Engineering at Zhejiang University, China.