Communications and Localisation for Cooperating Autonomous Mobile Robots
Ref: CISTER-TR-181121 Publication Date: 18, Feb, 2016
Communications and Localisation for Cooperating Autonomous Mobile RobotsRef: CISTER-TR-181121 Publication Date: 18, Feb, 2016
Heterogeneous teams of cooperating robots are ideal candidates for applications where the presence of humans is impossible or should be avoided, not only ensuring the safety of the people they replace, but also allowing the execution of tasks otherwise impossible. Moreover, if the team is heterogeneous, i.e. robots have dierent sensors/actuators, it is possible to optimise the use of such components reducing associated costs while maintaining full functionality. However, one the most attractive reasons for using such cooperating teams is the possibility of maximising the utility of the whole system; e.g. increasing the eectiveness of surveillance by performing cooperative sensing, improving the rate of coverage in search and rescue missions, and performing motion coordination for the transport of large parts. There are several key factors that enable such cooperation; in this work we will focus on two of them: a)exchanging information; b)tracking relative positions. For factor a), we propose a new solution for communicating data amongst a team of heterogeneous robots, which due to mobility requirements and dynamic team composition, with nodes joining and leaving on-line, is done through a wireless medium. Thus, we developed a new wireless communication protocol with support for dynamic group membership, based on fully decentralised proximity (ad-hoc) communications. However, to improve the eciency of using the channel bandwidth the protocol uses controlled transmissions, synchronized in a circular temporal framework of the Time Division Multiple Access (TDMA) type. One of the main innovations of this protocol is achieving distributed synchronization without using a global clock while tolerating uncontrolled trac that is external to the team. Resorting to a consensus technique we show that this method, based on the reception instants of the messages exchanged among the team members, assures synchronization in a vast range of operational scenarios. Moreover, the developed protocol supports multi-hop forwarding using a new algorithm that reduces the end to end delivery delay and, most importantly, drastically reduces the variability of such delay, isolating the communication requirements of each node throughout the whole network, significantly improving its temporal behaviour and analysability. For the factor b) above, we developed new methods to derive relative positions from local RF communications. In particular, we use both the signal strength and time of flight paradigms and with them we propose an innovative hybrid solution capable of improving the performance of any of them alone. These ranging paradigms are then used to produce pairwise distance measurements with which we derive the robots positions, relative to each other. Finally, we introduce another innovation that allows us determining the confidence of the relative positions determined with this method as a function of the confidence of the pairwise distances between robots.
PhD Thesis, FEUP.