Rodent-like robots - BIOTACT attempts to mechanically replicate the impressive whiskers of the rat
February 21, 2008 Robotics researchers are taking design tips from one of nature’s less majestic creatures: the rat. The EU funded €5.4 million BIOTACT project, which was launched at the beginning of this year, is seeking to replicate the rat’s highly efficient whisker system, which can accurately determine the shape of objects and help capture prey.
The whiskers of animals like the Norwegian or common rat and the Etruscan shrew greatly surpass the human fingertips in gathering tactile information. By sweeping their whiskers back and forth at high speeds, rats receive detailed, continually updated information about the specifications of three-dimensional objects. The horizontal plane is revealed by the timing of neural signals relative to the whisking motion, the vertical plane is revealed by the vertical spacing of the whiskers, and the radial plane, (or the distance of the object from the whisker), is revealed by the number of times the neurons fire. Furthermore, the signals travel from the whiskers through parallel pathways that function within parallel closed feedback loops, constantly monitoring the signals they receive and changing their responses accordingly. This complex structure gives rats a high level of control over their data gathering whiskers.
'The aim of this research is to help gain a better understanding of the brain on the one hand, and advance technology on the other,' Professor Ahissar explains. 'That is to say, researchers can use robots as an experimental tool, by building a brain-like system, step-by-step, gaining insights into the workings of the brain's inside components. With regard to technological applications, we suggest that it is the multiple closed feedback loops that are the key features, giving biological systems an advantage over robotic systems. Therefore, implementing this biological knowledge will hopefully allow robotics researchers to build machines that are more efficient, which can be used in rescue missions, as well as search missions under conditions of restricted visibility.'
The BIOTACT, or Biomimetic technology for vibrissal active touch, project aims to develop artificial touch technologies that can be integrated into intelligent robots. Researchers from the UK, Germany, France, Italy, Switzerland, Israel and the USA are contributing to the research.
Haptics, or the technology of touch, is becoming increasingly important in the computer and robotic industries. Effective mechanical tactile sensors could pave the way for a new generation of autonomous robots in a wide variety of fields. The BIOTACT researchers envision their artificial whiskers being applied to search and rescue drones that explore dangerous and dark terrain where humans cannot safely go and traditional optics based searching is inadequate. A refined sense of touch could also improve the effectiveness of bots that are required to cover a lot of ground with minimal human oversight, like domestic vacuum bots or automatic lawn mowers.
For robots like ASIMO, and Toyota’s violin playing model, an integrated sense of touch is the next logical step in making them as human-like as possible. The vision for these robots is that they will be able to intelligently learn and adapt to situations, eventually performing complex tasks for human owners. A sense of touch that delivered information on the same scale as human fingertips, or rat whiskers, would drastically enhance their ability to learn and complete tasks.
'The use of touch in the design of artificial intelligence systems has been largely overlooked until now,' says Professor Ehud Ahissar of the Weizmann Institute of Science, Israel, whose research team at the institute's neurobiology department is one of the groups participating in the multinational project. Perhaps it’s true that due to the way we perceive the world, our inventions show an over-emphasis on the sense of sight - in which case this is an impressive reminder that our creations need not share our design constraints.