Enhanced Precision Time Synchronization for Modular Robots

As in all distributed systems, having an accurate access to a global notion of time is vital for a modular robot's modules to coordinate their activities and accomplish their goal. In this paper, we present and compare two methods for clock skew compensation to enhance the performance of network-wide time synchronization in modular robots with neighbor-to-neighbor communication. The first one, Adaptive Rate Search (ARS), uses a light weight adaptive search method to adapt the drift rate of local clocks. The second one, combines Linear Regression with Bayes estimation (LR+) to reduce the accumu-lative error induced during the propagation of synchronization messages on large number of hops. We evaluate both methods with Blinky Block robots: using a real modular robots system and simulation. The results show that both methods LR+ and ARS present a significant error reduction compared to least-square linear regression used in previous state of the art synchronization protocol for modular robots with neighbor-to-neighbor communication.