Robot applications are autonomous systems that operate in highly dynamic surroundings. Consequently, they are susceptible to changing environmental characteristics and unanticipated resource breakdowns. Furthermore, they are often required to operate for extended periods. Dynamic reconfiguration provides a powerful mechanism to enhance robot adaptability, and to allow the software to be maintained at run-time. It allows "on-the-fly" reconfiguration of robot software components to change algorithms and hardware drivers. In this paper we present an approach towards dynamic reconfiguration for robot software. The paper addresses requirements and design details for carrying out changes safely and efficiently at run-time. Further, dynamic reconfiguration is applied to two robot navigation systems with different design properties in a case study. Results show efficiency has a heavy dependence upon the choice of interaction technique between processing entities (objects). Applications built from objects that communicate asynchronously can be reconfigured more efficiently than alternative applications where inter-object communication is via explicit method invocation