Designing and developing autonomous robotics systems is quite a challenging and time-consuming task. Besides that, current robotic systems are expensive and mostly designed to perform a single task. Developing software environments to aid the design of low-cost robotic systems will open up development to people and parties which are less connected to the field of robotics.

This thesis proposes a robotic system design environment that is user-friendly, low-cost, and can be used to find solutions for several different tasks. The robotic design system that has been developed controls a Lego Mindstorms robot via a NetLogo model. It makes and enacts decisions by using robot commands and real-time sensor feeds such as demonstrated in NetLogo models (line-following and subsumption architecture roaming). The results have shown that the robotic system design environment is capable of sending commands to robots and getting real-time feed-backs from sensors.

This environment has also been modified to extend its capabilities to include evolutionary techniques using Grammatical Evolution. The code for the NetLogo application was evolved grammatically using simulated agents that were embodied within a virtual environment for the task of maze exploration. The approach was evaluated using a selection of mazes with a robot inserted
into these unknown environments without any internal memory mechanisms.
The solutions found during virtual experiments were extracted to a Lego
Mindstorms robot and validated in the real world.

The robotic system design environment was further developed to discover and validate novel interesting behaviours. A combination of configurations and grammar used for these experiments resulted in producing several unusual behaviours. The behaviours were evolved by using Grammatical Evolution and a novel compression-based metric was used as a fitness function to effectively discover interesting behaviour. These behaviours were validated by real-world experiments by comparing them against the simulated results.

In summary, the research undertaken in this thesis developed a novel robotic system design environment that is user-friendly, low-cost, and capable of performing several different tasks involving complex robotic behaviours such as maze exploration, and was able to discover and validate novel interesting behaviour.