Sometimes the best way to figure out how to make a robot do something is to find an animal that can do it and then build a robot that imitates it. Here are some examples of robots that do just that.
Stickybot is a robot designed to imitate a gecko. It can climb smooth vertical surfaces and is being developed at Stanford University
Currently Stickybot is limited to operating in a very controlled environment but eventually it will be able to climb a much wider range of surfaces.
Snakebot is a robot similar in design to a snake which allows it to travel in places that would be too small for people are other machines. It is being developed at Carnegie Mellon University.
Snakebot works pretty well already but one important drawback is it currently requires a tether to operate which limits its usefulness.
The following is one of several examples of a robotic fish. This one is being developed at Essex University.
Robotic fish could be effectively used for underwater exploration in places or situations that would be too dangerous for people.
In the last two posts I have looked at examples of autonomous vehicles that can be used on public roads. The two examples I have discussed are the DARPA Urban Challenge and Google’s autonomous vehicles. Both of these involve fully autonomous vehicles either driving in a simulated environment or with human supervision on public roads. Although the technology needs to be perfected it still raises some interesting questions.
One of the biggest that comes to mind is how good is good enough? How reliable and safe will these vehicles need to be before they are allowed to be used without close supervision? The most logical answer is when robots are better drivers then people but that may not be good enough. I would see this more as the minimum requirement where the real answer depends on the public’s perception of the technology. Even if on average it’s safer than human controlled cars if people don’t see it as safer then it can’t be used.
Another possibility which could help with this issue is that the technology being slowly implemented using semi-autonomous vehicles. In other words, the car and the driver would be working together. Examples of this can already be seen today with features such as automatic parking and lane drift warning systems. Even cruise control and automatic transmission could be considered examples. This would provide time for people to get used to the idea of automated vehicles being on the road. However, even with a gradual transition there is still the jump from the driver being ultimately responsible to the car being fully autonomous and a driver no longer needed.
Or perhaps we will never be willing to give up control.
My previous post discussed Google’s autonomous robotic cars which it has been testing on the roads in California. One thing this post did not highlight is how difficult it is to build an autonomous vehicle. This example might give you a little bit better idea of both the difficulty and how few people/organizations are capable of rising to the challenge.
In 2004 DARPA hosted the first of a series of competitions for fully autonomous vehicles known as the Grand Challenge. This first competition was a 150 mile race through the Mojave Desert with a winner’s purse of $1 million. Of the 21 teams which entered only 7 were able to make it through the preliminary test which consisted of a 1 mile obstacle course that each entrant had to navigate successfully to be allowed to compete. In the final race not a single vehicle was able to complete the 150 mile course and the furthest any team got was 7.4 miles. All the vehicles either ran off the road or suffered mechanical failure.
Not to be deterred, DARPA tried again the next year. The race was similar to the previous one but used a different 132 mile course, which followed a dirt road and contained numerous sharp turns, several tunnels and a mountain pass. This time, 23 out of the initial 43 teams made it through the preliminary test and were allowed to race. In the race, 5 vehicles successfully completed the course but only 4 were within the 10 hour time limit.
The third race, commonly referred to as the Urban Challenge, was held in 2007 and used a 60 mile course through a simulated urban driving environment. The course included intersections, stop signs, multilane roads, parking lots and other challenges commonly encounter in an urban driving environment. The vehicles were required to obey all California traffic laws and complete a number of different challenges along the way. Of the 11 vehicles allowed in the final race 6 were able to complete all of the challenges and finish the race.
The following video gives some more information about Urban Challenge. The highlights from the race start 3min 20sec into the video.
If you want more information about these races, Wikipedia is a good place to start.
A few months ago Google disclosed that it has outfitted six Priuses and an Audi TT to run autonomously and has been testing them on the roads in California. This article from the New York Times provides some good information about this project.
The cars have covered over 1,000 miles on public roads while operating completely autonomously and an additional 140,000 miles with only limited human control. They navigate using a series of GPS waypoints. Additionally, a system of sensors and cameras are used throughout each car to find surrounding vehicles, stay within the lane lines and detect things such as stop signs, traffic signals and cross walks. The technology is still not perfect so during the tests the cars had a human sitting behind the wheel monitoring the situation and ready to take over if anything went wrong.
Eliminating the need for a driver could help to make driving safer since most automobile accidents are caused at least in part by human error. A computer doesn’t get tired or distracted (like in this video) and because of the multiple sensors it would literally have eyes in the back of its head.
One interesting point that the article makes is the current legal code assumes that cars are being controlled by a human driver so the eventual introduction of autonomous vehicles would probably require the addition of a whole new set of traffic laws. The most optimistic estimates say autonomous vehicles won’t be available on the market for another eight to ten years so it will be awhile before this is a serious concern.
