An environmental complexity analysis for robot-environment system design

Recently, researchers have begun to investigate intelligent environments for robot applications. In this work, a robot and its work environment can be considered to be an integrated system. Currently, such robot-environment systems are designed on an ad-hoc basis, with the final performance of the system greatly dependent on the experience and preferences of the designer. This dissertation investigates a way to improve on this situation by looking at the complexity of an environment from the perspective of a robot. The objective of this research is to develop a method to evaluate environmental complexity and then use this information to help design a robot-environment system.; The author introduces an approach to quantitatively estimate environmental complexity as seen by robots with particular sensing capabilities. Based on regression analysis, an empirical equation has been established to estimate the complexity of an environment using two environmental parameters: the entropy value of the open spaces and the compression level of all sensing patterns in the environment. This regression model has a high R-square value of 0.97 that indicates a successful complexity prediction. The quantified complexity estimation can act as a yardstick to evaluate different regions in the environment. The result can be used to determine high-value locations in an environment. An experiment result demonstrates that placing environmental information at high-value points can improve a given robot's performance. This gives engineers a way to make a deliberate design to improve the performance of the robot-environment system.; In addition, an innovative RFID landmark navigation auxiliary system is developed in order to embed knowledge into the intelligent environment. The developed RFID system can help the robot estimate its position and pose. The average positioning error of the system is less than 100 mm and the orientation error of the system is less than 5°. The system can not only determine a mobile robot's pose, but also provide other useful information to robots.