The following blog is a transcript of Brian Douglas’ (Control Systems Lectures) YouTube video (2015) on the Braitenberg vehicles. He says the following:
“A Rube Goldberg machine is a series of convoluted and complicated actions that in the end performs a relatively simple task. The concepted was invented by cartoonist and inventor Rube Goldberg who would often depict such machines in his cartoons.
A famous example of his work is Professor Butz and the self operating napkin. The machine would automatically wipe your face after you would raise your soup spoon to your mouth. The picture to the left gives you a sense of the machine’s absurdity.
Building these types of machines can be a fun engineering challenge, but if you are tyring to develop a system that performs a complex task, for example a robot that can mimic human or animal psychology, then trying to take the complicated Rube Goldberg approach probably isn’t the best solution.
This is because as the designer you’d have to think of every situation that your robot could get into and then through a series of premeditated processes generates the appropriate response and this can get out of hand for systems that are really complicated.
In general, a robot that produces a complex outcome does not need to be a complex robot.
The question is how can you build a simple robot that exhibits complex behaviours?
A celebrated example of very simple robots are the thought exercises proposed by Valentino Braitenberg in his book “Vehicles, Experiments in Synthetic Psychology”. In his book, Braitenber theorises simple robots that can create surprisingly complex behaviour. As Braitenberg says himself
We will talk only about Machines with very simple internal structures, too simple in fact to be interesting form the point of view of mechanical or electrical engineering.
He begins by describing the absolute simplest vehicle: a single sensor and a single motor and then he builds in complexity from there. A sensor is drawn as a half circle on a stock and a motor is drawn as a rectangle. The type of sensor is not terribly important. You really just need to know that it is capable of measuring its environment in some way. This can mean it measures temperature, the amount of light it sees, the air pressure or so on.
The same goes for the motor. It can be a wheel that drives a car or a propellor that drives a boat or even something just like a speaker that generates sound. We can connect them with a line that represents that this sensor is directly driving this motor. In this case as the sensor measures more of the quantity it is tuned to, the motor speed increases proportionately.
We can also add a a minus here, indicating that it inverts the relationship. As the sensor measures more, the motor speed decreases. And with set of symbols we can start to lay out any number of really complicated interactions, but let’s go back to the first vehicle that has a single sensor in front and a single motor in back and the sensor is connected directly to the motor.
Let’s imagine what this vehicle will do if the sensor detected light and the motor was connected to a wheel. Whenever the vehicle was in the shade, it would drive very slowly and as the vehicle got to brighter areas, it would drive faster. This in itself doesn’t seem all that impressive, but keep that thought in mind for just a bit while we take a slight detour.
Braitenberg devised all of his vehicles just as thought exercises, but the really cool thing about them is that we can build them and test them out very easily.
Imagine that you came across a but that would run faster whenever it was exposed to light and is actively finding a safe dark spot. Similarly, let’s build a vehicle that prefers the dark, because it tends to spend more time there and tries to run away faster when exposed to light.
So, we are building a vehicle that’s quite complex, as least as complex as a survival instinct of a cockroach. We can build a vehicle with a light sensor simply connected to its motor. So, we might say that this vehicle is really quite dumb and since it doesn’t understand anything, it can’t possible prefer something like darkness over light. But this is one of the central themes in Braitenberg’s book:
Synthesis is easier than analysis.
Or in other words, creating something that acts complex is easier than analysing what looks like a complex system. So, you can see that as an outsider looking at our vehicle, it would be very easy to apply concepts to it about what it likes and dislikes based on the behaviour that they observe.
Let’s take our first vehicle and invert the relationship between sensor and actuator and see how it infact its behaviour.
Starting from our first vehicle, we can adjust it by adding a second sensor and a second motor. You could think of this as a mutation where during the building process two vehicles got inadvertently stuck together. Each sensor-actuator pair still has the same behaviour as before, but now the behaviour of the system changes and becomes very interesting.
The vehicle still gets excited by light with both wheels moving faster in bright areas causing the vehicle to accelerate towards the source, but now if the light gets off to one side, the motor on that side will drive faster than the other causing our vehicle to turn away from the light source, eventually running away.
As an alternative, if the sensor lines were crossed, during the mutation, then our vehicle would still accelerate towards the light, but now if our vehicle isn’t pointing directly at the light, the mismatch in motor speeds is self-correcting and that will continue to point our vehicle towards the source until it crashes into it as fast as it can.
Braitenberg referred to these two vehicles as fear and aggression and although their behaviour is decidedly more complex than the first vehicle, they are exhibiting traits that are less than desirable. Who would like to have a robot that only knows how to be afraid or how to be angry.
We can create a more loving and caring robot, by switching the signs between the sensors and motors which will slow the motor down when the sensor is stimulated and give our robot a more calming demeanour. Our crossed-lined vehicle will initially drive towards the light and this fairly quickly as the source is far away and dim, but as it gets closer, it will begin to slow down and enjoy the company of our light source.
Once the light gets off to one side, the motor speed mismatch will cause the vehicle to continue turning and ultimately begin speeding up in search of another friend. You would say that vehicle likes the source just fine, but is always looking to explore the area for new friends.
For the straight-line vehicle, the light source is its one true love. When it is far away, it speeds up in search of it and once it spots the light, the mismatch in motor speeds points the vehicle is at its destination and as it gets closer, it slows down and eventually comes to rest right at the source forever.
In his book, Braitenberg is building more complex behaviours from seemingly simple electromechanical relationships. He goes on to describe how our simple machines could develop memory, the ability to predict and have foresigth and develop an ego. According to Braitenberg all these complexities of behaviour can arise from a simple neuron.
When building and programming robots, instead of always trying to define a structure that would exhibit a specific behaviour, try building simple things and see if you can replicate the behaviour you are trying to understand. Building something from the bottom up with simple rules can lead to a wonderful result that you might not have thought of.
If you would like to know more about Braitenberg’s vehicles, take a look at this review of Braitenberg’s vehicles by Dr Michael Dawson at the University of Alberta, Edmonton, Canada (Dawson, n.d.).
Rube Goldberg Machines — Boom Box Post (2016). Boom Box Post. [online] Boom Box Post. Available at: https://www.boomboxpost.com/blog/2016/4/22/rube-goldberg-devices.
Wikipedia Contributors (2019). Rube Goldberg machine. [online] Wikipedia. Available at: https://en.wikipedia.org/wiki/Rube_Goldberg_machine.
The Braitenberg Vehicles. (2015). YouTube. Available at: https://www.youtube.com/watch?v=A-fxij3zM7g [Accessed 31 Jan. 2020].
Thomas Schoch – www.retas.de [CC BY-SA (https://creativecommons.org/licenses/by-sa/3.0)]