Spinoza has evolved, with new communication, computation, and sensing capabilities added over time. We have pursued an incremental strategy, extending its abilities to handle new tasks. To limit the degree of low-level programming, as well as to reduce the impact of changing hardware, a stable software interface to robot services was required.
The actual robot hardware is varied, and cannot easily be mastered, thus a certain level of abstraction from the actual implementation was necessary.
The separation between abstraction and implementation, argued for by these considerations, leads to a model where the robot is isolated from the program development system by an interface. Controllers can be designed and tested in isolation from the actual robot; for example the development of controllers for the robot soccer players was facilitated by realistic simulation of the soccer players and their environment. This dovetails nicely with our desire to include teleoperation, the human intelligence and decision-making capability, in the robotic system.