Resource Management for Mobile Robots

Abstract: As more functionality and complex systems are combined in limited settings the need for resource management is becoming increasingly important. Resources such as computing capacity, battery power and communication channels need to be divided between different applications and hardware to achieve the highest global performance. Traditionally the computing capacity has been handled by the system scheduler, but in dynamic systems with complex usage scenarios the system analysis that needs to be done off-line in order to guarantee stability does not only grow more unfeasible, but since such an analysis must be done considering the worst case scenario, this may provide an unnecessarily pessimistic result. In the case of mobile robotics, a key feature is the ability for the robot to adapt to a changing environment, the usage scenarios may vary from one time to another, the task to be performed could vary and the surroundings of the robot may change. Heavy computing load may cause specific parts of the system to overheat, thereby limiting the performance. To increase the performance and to make the system more flexible in terms of new hardware configuration, an adaptive resource management framework is needed. An adaptive resource management framework that does the system analysis on-line would reduce the configuration time, and increase the performance and flexibility. In this thesis, an existing resource managemet framwork has been ported to a mobile robot patform. The thesis also proposes a method for controlling the CPU temperature, by using the system utilization and adaptive resource allocation as a mean to keep the CPU temperature bounded.