The previously discussed engineering discipline extends the ideas of electronic engineering to the world of biology. However a question could be raised, “Is it possible to do the exact opposite?” Not only can we make use of the ideas of electronic engineering in the field of biology, but we can also employ biology to electronic engineering. Evolvable hardware is such an intriguing application that mimics the power of nature described by Charles Darwin’s theory of evolution to develop electronic hardware systems that evolve themselves.

In modern day’s electronics design, it becomes essential to construct integrated circuit chips which are increasingly capable because of the demands. However this eventually leads to two difficult situations. One is that the average circuit density in a chip doubles every 18 months for embedded systems according to Moore's Law. The other one is that because of the complexities of the circuitry increases considerably over the last few decades, the costs of design also rose significantly, resulting long development time for products to reach the market. Hence, on the brink of a major breakthrough it is crucial that we investigate ways to develop new theories and methods to overcome those problems.

One solution proposed by researchers is evolvable hardware. Evolvable hardware is a kind of hardware that are auto-transformed from some basic circuit elements. This functionality is made possible by using evolutionary algorithms which was inspired by Darwin's theory of evolution. In these algorithms, one most used category of evolutionary algorithms is genetic algorithm (GA) which we discussed in the previous section.

For many years in the embedded hardware or circuit designing domain, designers always have to consider the following problems: The increasing demand of higher processing power while they need to keep the power consumption low. The most obviously example is our mobile phone which is more powerful in terms of its functionality while it becomes slimmer than its parent. Meanwhile, it seems that it is increasingly harder to shrink its size and lower down its purchasing value due to all kind of limits. An ordinary phone running faster than a current desktop computer with 3 GHz dual-core CPU is still a dream. Also noteworthy is that, some special-tasked systems are expensive, time consuming for the design and to maintain. Such as the Hubble space telescope or the satellite to be sent to the Mars. Therefore, something must to kick in to improve the quality of the hardware.

Evolvable Hardware is based on evolvable algorithms, and implanted on reconfigurable hardwares. There are always two evolving modes to choose from, intrinsic and extrinsic1. Intrinsic evolution means that the circuit evolution only on the board without interference with the host PC, apart from the initial setup period. This mode is suitable for computational intensive application. Extrinsic evolution on the other hand can be designed on computer with specific software and download strings of data onto the board. This method is preferred when the design is the main focus and the time is not a major consideration. As we can see, the intrinsic mode is much more difficult to do mainly because we just cannot control the whole evaluating process on the board yet it is very useful for some tasks which we will discuss later. For all evolvable hardwares, they must have 3 kinds of units to work with. Some memory units are used to store the chromosomes, the fitness data processing unit to renew the fitness for every offspring of the population, and another unit to perform the algorithm with operators. If they are not appearing on the board then it must exist on a host PC. Here, we first compare the evolvable hardware to other solutions in order to achieve better performance, and then we discuss the architecture used to build such system and their tradeoffs.

Comparisons

We focus on other 2 kinds of systems comparing to the Reconfigurable Computing System (RCS). The first one contains high-end processor. The main advantage it has is that it has great information processing power such as the Intel i7-960 which runs at 12.8 GHz for the whole unit but the power required is about 130 Watts and it is out of reach for normal customer... (Read more)

Architectures of FPGA

Evolvable systems were often built upon commercial reconfigurable Field Programmable Gate Array (FPGA) considering most signals are converted into bit-stream fashion to enhance the design ability by software. Device from the company such as Xilinx and Altera’s are used more often. We learnt from a survey on reconfigurable architecture... (Read more)

References

1 Zebulum, R. S., Pacheco, M.A., Vellasco, M.: Analog circuits evolution in extrinsic and intrinsic modes. Proceedings. Evolvable Systems: from biology to hardware 1478, pp.154-165 (1998)

New developments

In this section, we will have a look on systems that EWH has developed so far, and means that have recently used to improve the qualities of the systems. First, we will look at error detection... (Read more)