\index{MARS!history} MARS has an almost 2 year history, whose basic objectives have not changed since its first conception. (The planet Mars is a lot older) The implementation of the system has, however, changed drastically. Initially it was considered that the project will fall within the \$500 budget, but as we all know computer related projects are never within budget! It was initially considered that the UNSW Micromouse \cite{MICROMOUSE} project was going to be used as the platform, and the processor on board the mouse was going to handle the vision processing. It, however, quickly became apparent that the Micromouse would not have the payload carrying capability to carry the camera and its associated power supply. Another reason why the Micromouse was considered unsuitable was because the Z-80 processor on board was not going to be able to provide enough computing power, other than the most basic of vision processing functions. The project could have been implemented on the Micromouse, but the research value of the project would have been quite minimal.

So, a new platform had to be chosen, one that could easily carry the camera and power supply. There weren't many platforms already available through the school, so it was decided to use a robot, that I had constructed 2 years earlier (ROBI-6). This had the carrying capability for both the camera and power supply. One drawback was that ROBI-6 did not have an onboard processor board but rather was attached to an IBM PC via an umbilical cord. For an autonomous project, as this thesis is, that was unacceptable. So a processor board had to be found or built that could be placed inside ROBI-6. This was not the final platform used for the project though.

Choosing the processor board was just as eventful. At first it was considered, building a dedicated Z-80 board that could manage around 1 MIPS. But this was soon dismissed, because the amount of time required to develop the board would have been far too great, not to mention the software development tools, and Z-80 assembly language that would have had to be tackled. Also, the board would have been a non-standard board, making it difficult for further updates. It was next considered using an off-the-shelf single chip computer board. This would have been better in some respects, but still the computation power would not be there. So finally, I decided I would use my old 80286 mother board. This had many advantages, I was already familiar with the system, as well as the development tools, and it provided a very convenient environment to develop and test the software for the system.

Since this is a vision system, some sort of image capture hardware was required. Again, it was first considered building a frame grabber. I bought a few parts, but because at the time I decided to change the processor board to a 286, it was again considered buying an off-the-shelf frame grabber. This was finally done, and the very powerful Video Blaster\footnote{Video Blaster is a registered trademark of Creative Labs}was purchased, that is capable of capturing live video, (24-bit images) up to resolutions of 800x600. The VideoBlaster comes with its own capture software, and simply plugs into one of the slots on the PC bus. This has considerably reduced the implementation time.

Not being satisfied with the 286 motherboard, and a new 486 motherboard was desperately seeking a new case. It was decided to make the case for the 486 the robot itself. However, ROBI-6 was not designed to carry that much payload, the 486 and associated peripherals as well as a power supply and battery. So we were back to square one of trying to find a platform for the project. Finally it was decided that a new platform would be built, that was big enough to carry everything. And thus was born the MARS platform.

The implementation has changed from an initially considered tiny robot, with a very simple processor, to quite a large and powerful one.

It is more practical, because it offers a great potential for future research. It has the capacity for adding perhaps colour stereoscopic cameras, as well as other sensors, all easily connected to a powerful computer. The Z-80 Micromouse would have been little more than a toy, not easily upgradable for future work.