Research Robots: Khepera III

For the last month I have had the opportunity, through my internship in DISAL, to get familiar with the Khepera III robot. The 130mm differential drive, produced by the K-TEAM Corporation, is the descendent of the smaller (5.5cm) Khepera robot developed in the mid 90s, in the LAMI laboratory of EPFL, by a team which included among others Francesco Mondada and André Guignard.

The original Khepera based around a Motorola 68331 CPU was widely used by research universities for over 10 years and was featured on the cover of the August 31st, 2000 issue of Nature.

Its latest incarnation, the Khepera III, offers a much higher computing power than its predecessors. With the KoreBot extension with an 400MHz XScale CPU and Linux OS, and many other stackable modules available, it presents a great platform for the implementation of truly autonomous complex mobile robot algorithms.

The basic sensing on the Khepera III is formed by 9 infrared sensors forming a ring around the robot, 2 additional infrared sensors facing down (that can be used for navigation on a table top or line following) and 5 ultrasonic sensors, arranged on the front semicircle of the robot. Also optical encoders for odometry are present on each of the wheel axes.

The infrared sensors have two functioning modes: an active mode, in which the time of flight of an emitted light signal is measured, and a passive mode, for measuring the amount of infrared light present in the environment.

Ultrasonic sensors have had a very bad reputation because of their so-called lack of directionality. This is due to the fact that an ultrasonic sensor tipically has a significantly large beam angle. In fact, the sensors on the Khepera III have a very large beam angle which might discourage their use for navigation algorithms. Nevertheless, with a proper understanding of the physics of such a sensor, a functional sensor model for indoor navigation applications is feasible without great difficulty.

With its basic capacities and also its flexibility in adding new modules, the Khepera III robot is a very good option for research in swarm and evolutionary robotics.

Elephants Don’t Play Chess

In 1986, Rodney Brooks introduced the idea of the Subsumption Architecture into the world of robotics changing its landscape forever since. His view on artificial intelligence went against the trend of the time, of centralizing symbolic processing, which was not delivering the kind of intelligent behavior that would be equivalent to even the simplest of living organisms.

The idea behind the Subsumption Architecture was to decentralize intelligence in a way that is very similar to what we know to be true about the nervous system. It decomposes the intelligence of the system into many simpler modules organized in layers. The lower layers take care of basic tasks like obstacle avoidance and the higher ones take care of more complex tasks like environment exploration. Lower layer modules work quicker, like reflexes, but higher layers have the posibility of inhibiting the lower layers.

The result of this new philosophy was the emergence of a new generation of robots much simpler in design that were able to achieve much more interesting real environment behavior. The first one, Genghis, was built in the mid 80s at MIT by Rodney Brooks and looked like a large insect. It can be seen in the first images of the above youtube clip.

What Brooks did was to divide robotics between two basic paradigms: the  deliberative paradigm and the reactive paradigm. The deliberative paradigm represents the classical approach in which the robot senses the world, plans the next action and then acts. The subsumption architecture is representative for the second paradigm, also known as the sense/react paradigm. A third and more recent trend is to try to get the best out of the two philosophies, resulting into a hybrid paradigm.

In 2002, Rodney Brooks was present on BBC’s HARDtalk speaking about his views on the future of robotics. You can download the show here. One year later, he gave a very interesting talk at TED called How robots will invade our lives, which you can see here.

A New Leaf

A few days ago I got inspired by a news feature in Nature to transform what really was a pointless and rather boring blog into a science oriented one. To give credit where it’s due, the catalyst for this reaction was a very captivating lecture given by Prof. Floreano, during the breaks of which I was reading the feature.

The ThreadBag will mainly deal with mobile robotics, bio-inspired robotics, evolution and artificial intelligence, but will certainly also veer into other science and technology domains that might strike my fancy. I will try to keep it as accessible to the uninitiated reader as possible so as to enable anyone with a real interest in science to read it with ease.