Visual Motion Sensors and Neuromorphic Obstacle Avoidance System

Abstract

In this work, we present motion detection algorithms that are based not only on biological models but also on the computational properties of motion perception. First, we describe monolithic implementations of hysteretic winner-take-all and nonlinear differentiator based algorithms. Second, we explain multi-chip implementations of biomimetic intensity-based models, namely Adelson-Bergen model. In addition, we describe an obstacle avoidance algorithm that is realized by incorporating a multi-chip version of the Adelson-Bergen algorithm with centering behavior and time-to-collision computation. In this way, the overall system can successfully acquire clues about the structure of its environment so that collisions can be effectively avoided. This system might be employed in building a robot that can navigate in complex cluttered environments. I. Introduction In this we describe an obstacle avoidance algorithm based on the system level implementation of the Adelson-Bergen multi-chip sensor integrated with centering and escape behaviors. As with biological systems, perception of motion information occupies a vital role in behavioral tasks achieved by artificial systems. It is utilized for tracking, collision avoidance, object recognition, time-to-flight computation, guidance, balance, and postural control.CCC In building biologically inspired architectures, it always has to be taken into consideration that biological models are by their nature continuous-time systems and they operate by employing a massively parallel processing strategy. These biological principles can be efficiently utilized in visual computations. In contrast to other sensory computations, optic flow computation is a very intensive process that is constrained by power consumption, and employing such technology definitely improves the performance of built systems. In addition, conventional design approaches in image processing that are implemented by employing a CCD camera together with a DSP processor cause problems such as an image transfer bottleneck, temporal aliasing and high power consumption. These systems work at high frequencies to deliver image information in a timely way and in achieving this they have to consume considerable amounts of power. Whereas, an integrative approach by analog VLSI technology and neuromorphic design principles makes intensive visual computations possible to be realized in power and space efficient systems. In contrast to conventional designs, these systems are data driven, that is, the output is sampled when there is a demand. Hence they consume little power and yield temporal aliasing-free computation. Inspired by biological models in the visual pathways of organisms and by computational properties of visual motion perception, a variety of visual motion sensors, which consume little power and work in real time, have been developed to solve problems faced in optical flow computation.