Analogue, Mixed Signal, LSI


Biologically inspired software being developed for the 21st Century

1 March 2000 Analogue, Mixed Signal, LSI

Software which enables computer systems to adapt to change automatically, think for themselves and identify problems before they occur, it is being developed by the British Telecommunications Laboratories using programming techniques which mimic biological evolution. Ian Gordon-Cumming, Country Manager for British Telecommunications South Africa, explains further.

This 'thinking' software has an intuitive problem-solving ability which is already helping to run the next generation of intelligent networks. When applied to business and technical applications, it will increase flexibility and enable British Telecommunications to deliver better value for money and more advanced products and services to its customer.

As Professor Peter Cochrane, Head of Advanced Applications and Technologies at the British Telecommunication Laboratories, puts it: "So far, we have tended to treat computer programs as electronic servants. We humans have kept the real thinking to ourselves and only used the machine to try out the 'what if' scenarios we dreamt up. "But what if we were able to ask the thing on our desk 'Where's the problem? What is it? Please solve it.' At the moment, they're not smart enough for that, they can only respond to our input and instructions. Here at the British Telecommunications Laboratories we decided to see if they could think for themselves."

BT has targeted biologically-inspired software because of its awareness that software productivity is substantially lagging behind that of hardware. While chip power is doubling every 18 months, software only increases its 'usefulness' by some 5% per year. As a result, the company is looking for ways of speeding-up software development to provide the increasingly complex and busy communications and computer networks it will need in the 21st Century.

Current developments

BT is currently using biologically inspired software technology to develop:

* Self-healing network algorithms, which help maintain trouble-free, self-correcting telecommuni- cations links.

* Programs which help Internet users find the right documents from the thousands on offer, or act as filers to weed out unsuitable material for particular audience groups. p Dynamic visualisation tools, also for use on the Internet.

* Programs enabling users to participate in virtual 'shared spaces' or virtual worlds on the Internet.

It also has the potential to be used to develop systems which understand speech, recognise a face, improve business and economic modelling and even assist in the development of newer, more complex computer programs.

Principles

The principle behind biologically-inspired software is to see how biological concepts can change the way computer systems are designed and used. A great deal has been learned by modelling evolutionary processes and using them to design more intelligent programs which automatically adapt to change, as nature does and which use intuitive reasoning to anticipate problems or mistakes.

This has become necessary because today's software is brittle and inflexible having no fault tolerance. Just small errors can make the entire program fail, as programs are designed to be accurate rather than robust. However, the prospect of programs which work by giving inexact but acceptable answers opens up a whole new approach to computing. When existing software is used for tasks where speed or robustness is more important than accuracy, it has severe limitations. Yet these are the areas where biologically evolved organisms perform best of all.

The solution is to create progenitor programs which each product 'offspring,' which in turn can breed to produce more intelligent generations. This is the foundation for future generations of intelligent, thinking software, where only the fittest programs survive.

A demonstration program has been written to illustrate software elements combining de-combining, mutating, selecting and dying off. The algorithm enables each program to produce two offspring programs, note their characteristics and then use the information to produce the perfect program. This is being used to design further software and it can be done much faster and more cheaply than by human programmers.

For example, BT is designing a business model to show how, using this technique, service delivery programs which should almost exactly meet a specific customer's needs can be developed very quickly. It also means that as new competitive elements within a particular market are factored in, the process can be quickly repeated to regain an optimum operating efficiency.

As computers are sequential, local thinkers, the biological approach will introduce randomness and an ability to adopt 'soft' criteria, like dynamic modelling, which will adapt itself to changing circumstances. Applied to something as complex and extensive as a telecommunications network, an ability to have integral fault tolerance is essential, as more and more applications are embedded in the network.

Peter Cochrane explained: "BT is moving quickly towards a telecommunications network that will soon have a powerful, embedded computing capability. I expect this within the next three to five years. Subscribers who are today downloading software to their PC from the Internet maybe won't even need a computer. They will simply be dialling one up on demand and getting access to much more computing power than they could ever imagine."

He added: "Using this technology, computer programs are being produced that we didn't design - the machines have done it for themselves. It seems that computers are beginning to think for themselves and with our help and under our control, it is possible that soon they could even begin to be asked to make subjective judgements."

British Telecom can be contacted on (011) 807 6760.





Share this article:
Share via emailShare via LinkedInPrint this page

Further reading:

High-speed PIN diode
Altron Arrow Analogue, Mixed Signal, LSI
Vishay’s new high-speed Silicon PIN diode is able to detect both visible and near infra-red radiation over a wide spectrum range from 350 to 1100 nm.

Read more...
Analogue front end for sensor measurements
Electrocomp Analogue, Mixed Signal, LSI
The NJU9103 AFE from Nisshinbo is a tiny analogue front end, with a 16-bit resolution ADC and up to 512 x signal amplification from the programmable gain amplifier.

Read more...
Single-channel software configurable I/O
Altron Arrow Analogue, Mixed Signal, LSI
These use cases of the AD74115H include analogue output and input, digital output and input, resistance temperature detector (RTD), and thermocouple measurement capability.

Read more...
Current-sense amplifier with PWM rejection
Altron Arrow Analogue, Mixed Signal, LSI
Analog Devices’ AD8410A is a high voltage, high bandwidth current-sense amplifier that features an initial gain of 20 V/V and a 2,2 MHz bandwidth.

Read more...
Combining a LPF and ADC driver for a 20 Vp-p signal
Altron Arrow Analogue, Mixed Signal, LSI
A mixed-signal ADC driver circuit’s optimum performance depends on multiple variables: the driver’s settling time, the RC filter’s time constant, driving impedance, and the ADC sampling capacitor’s kickback current, all interact during acquisition time and contribute towards sampling errors.

Read more...
Pressure sensor with NextNav certification
EBV Electrolink Analogue, Mixed Signal, LSI
STMicroelectronics’ LPS22DF pressure sensor has received a NextNav certification, which guarantees performance and reliability for geolocation and other types of applications.

Read more...
Reference design for Raspberry Pi analogue I/O
Analogue, Mixed Signal, LSI
Analogue Devices has published a reference design for ±10 V analogue input and ±15 V analogue output for Raspberry Pi platforms.

Read more...
Multi-zone distance sensor with 90° field of view
Avnet Silica Analogue, Mixed Signal, LSI
STMicroelectronics has revealed a new FlightSense multi-zone distance sensor, with 90° field of view, to bring lifelike situational awareness to applications like home automation, computers, robots, and smart equipment.

Read more...
Six-axis IMU with Qvar sensing channel
Altron Arrow Analogue, Mixed Signal, LSI
The LSM6DSV16X from STMicroelectronics is a high-performance, low-power six-axis IMU, featuring a three-axis digital accelerometer and a three-axis digital gyroscope.

Read more...
Bidirectional current-sense amplifier with PWM
RS South Africa Analogue, Mixed Signal, LSI
The MAX49925 from Analogue Devices is a bidirectional current-sense amplifier (CSA) with an input common-mode range that extends from -40 to 76 V, making it suitable for 48 V HEV applications where there are large automotive transients.

Read more...