Approximate and Distributed Algorithms for a Collaboratively Controlled Robotic Webcam

Introduction

The recent development of low-power networked robotic cameras provides low-cost interactive access to remote sites. Robotic pan-tilt-zoom cameras can cover a large region without using excessive communication bandwidth. With applications in natural environment observation or surveillance, a single robotic camera is often concurrently controlled by many online users and networked in-situ sensors such as motion detectors. Since there are multiple simultaneous requests in a dynamic environment, an optimal camera frame needs to be computed quickly to address the resource contention problem and hence to achieve the best observation or surveillance results. In last chapter, this is proposed as a Single Frame Selection (SFS) problem when requests are rectangular regions. However, a majority of requests are not necessarily rectangular in many applications. The shapes of requests are usually determinated by factors such as the shapes of objects in the scene and the coverage of in-situ sensors. On the other hand, the existing algorithms give exact optimal solutions and are not scalable due to their high complexity. A new class of fast and approximate algorithms are favorable for this generalized SFS problem that prefers speed to accuracy.

This chapter was presented in part at the 2003 IEEE/RSJ International Conference on International Conference on Intelligent Robots and Systems (IROS), Las Vegas, NV [170] and in part in the IEEE Transactions on Robotics [168].