PID Control

Control theory looks at a lower level of intelligence than the deliberative and reactive architectures that we have looked at before. An example of when it may be needed in robotics is as follows: you need a robot to travel at some specific speed and that is all the input you give the robot, as the terrain becomes more challenging the motors will have to work harder to keep the robot at this specified speed. A controller can be used that is given required speed and some sensor feedback about the actual speed to figure out how much power the motor needs. PID control is a simple example, PID stands for Proportional, Integral and Derivative. These are all error terms calculated in the controller, in a P control the error is found between the required speed and the current speed, this error is multiplied but some factor and the output is sent to the motors. If you are still interested at this point, (and I think you should be, it's quite important stuff) please check out this article by John Holland from the EE Times and you will learn all about it.

Here is a quick video showing the difference between P control and PID control:

Useful Resource for Embedded Systems

I would like to recommend the following e-book from Byte Craft Limited - First Steps with Embedded Systems. This book covers the fundamentals of embedded systems and the C programming language. Robots are a type of embedded systems and the C language is often used to program them. This is a very useful introduction but it is also worth looking at lectures on YouTube such as the following (if you can get past the opening titles!): 

Reactive Robotics

In order to increase the responsiveness of the early robots the plan element of the sense, plan, act sequence was dropped in what became known as 'reactive robotics'. The paper Intelligence without representation (1991) by Rodney A. Brooks has an introduction which explains the ideas behind the concept and then goes on to explain one of the most well known examples of the reactive architectures - the subsumption architecture. 

Subsumption works by defining the robot's behaviours and organising them on to the 'subsumption stack'. Each layer has access to the sensor input and can specify the action to be taken. The higher layers are more abstract and work to achieve the robot's overall goal whereas the lower layers are simpler actions which need to respond immediately to environmental stimulus. The details of the architecture are well worth looking into. Reactive robots are still in use especially in swarm robotics.     

Check out YouTube for some examples, I found this one illustrating an avoid behaviour: 

Shakey the Robot

Shakey is a robot created at SRI International in California in the 1960's. It is a very well known early example of a robot that has the ability to see and model its 'world', plan, plot a route and perform simple tasks. It is also an important example of an hierarchical or deliberative control architecture. This type of control architecture organises the robot's 'brain' using the sequence sense, plan and act. Planning normally involves reading, creating or updating a world map and finding and following a route. This step in the sequence can be quite slow and given the computing power in the 1960's the reason for Shakey's name becomes apparent. 

There is a lot of information about Shakey the robot on the SRI International's website, which includes video footage and some papers. It is important to learn about early robots because they give us an idea of where robotics has come from and helps us to see where it is heading, perhaps allowing us to revisit older ideas given newer technology.

Marshall Astor from San Pedro, United States, took this photo of Shakey which is now in the Computer History Museum, California, USA.

Looking at Shakey we can see that there are several important fields that immediately come together in the field of robotics that we will be interested in learning more about, hopefully some of which in this blog! Primarily we have to look at the construction of the robot - the field of mechatronics and the brain of the robot, which will cover both hardware - electronics and embedded systems and software - the field of artificial intelligence. We have a long way to go.

Introduction to Robotics

I found a series of 16 lectures from Stanford University today. In the first one - Introduction to Robotics the lecturer, Oussama Khatib, shows examples of recent robot development and introduces some of the many areas of robotics that we should know about robotics; foundations in kinematics, dynamics, control, motion planning, trajectory generation, programming and design.


The lectures are all about an hour long so make some notes and if you don't understand something read about it elsewhere. Simple, right?

Introduction

In this blog I hope to learn about robotics with you guys, both the history and the technologies currently in use. I also want to keep us all up to date with the latest developments in the field.


I am thinking of just posting on the blog whenever I have time and posting news links and such on twitter so you might want to follow zestengineering just in case I come up with something interesting. 


I have been reading various books and papers on robotics and surrounding fields so I will recommend any that I think are useful.


Ok, so let's see how it goes - wish me luck!