ASIC implementation of a Gaussian Pyramid for use in autonomous mobile robotics

Mobile robots evolved significantly in the last decade. Reasons for this are the increase in computing power and the advances in sensory. A state-of-the-art robot uses many different sensors to perceive its environment. Among these are visual sensors - from simple photo detectors to sophisticated camera systems. For the evaluation of visual sensors´ data, a significant amount of processing power is needed. The necessary processors consume some power. But power consumption is an important factor, limiting the operating radius and operating time of mobile robotics. Minimization of power consumption is important to extend the operating time, especially for battery powered robots. For specific applications, it seems to be appropriate to implement certain algorithms in dedicated hardware, to save processing power and power consumption. In this work, the authors present an ASIC implementation of a low-level image processing algorithm, the Gaussian Pyramid, which is widely used in robotics. It is implemented in a 0.15 μm process and is now in testing. After functional and parametric tests, the ASIC will be tested in the autonomous mobile robots AMOR and PSYCHE from the Institute of Real-Time Learning Systems of University of Siegen, Germany.