**7. Conclusions and Future Work**

In this paper, we presented a framework designed for fast prototyping of autonomous navigation systems. This framework reduces dramatically the amount of work required to implement a whole application, making easier to test and to compare different algorithms in real-world conditions. The proposed architecture permits us to focus the efforts in the desired research topics, while the provided basic set of tools enables the users to generate fully operative autonomous navigation systems to perform experimental tests. To validate the described approach we used the framework and the provided tools to implement an initial basic system that was able to complete successfully several laps around a building autonomously in a challenging outdoor scenario. To demonstrate the easiness of module substitution, we developed a novel algorithm that relies in a Kalman filter to fuse 2D SLAM and GNSS positioning and used it to replace the localization module of the initial basic system. This new module was incorporated just changing a single line of the launch file. Extensive

tests for this new localization module were performed navigating autonomously through mixed on-map/off-map trajectories in the University of Alicante campus. This new localization module has shown interesting properties on its own, and thus has become one of the contributions of the present work. The experimental sessions covered more than twenty kilometers in two absolutely different outdoor environments. All the software is publicly available in a GitHub repository (https: //github.com/AUROVA-LAB/aurova\_framework) with the aim of being a useful tool for research groups interested in any of the fields related to autonomous navigation.

As future work, we want to extend provided set of tools, adding re-planning capabilities, redundant safety modules and terrain analysis algorithms. In the localization part, we plan to integrate a graph SLAM system implementing a tight integration of GNSS raw observables and to use unsupervised learning techniques for landmark detection.

**Author Contributions:** Conceptualization, M.A.M. and I.d.P.; methodology, M.A.M. and I.d.P.; software, M.A.M. and I.d.P.; validation, M.A.M. and I.d.P.; formal analysis, M.A.M. and I.d.P.; investigation, M.A.M., I.d.P., F.A.C. and F.T.; writing—original draft preparation, M.A.M., I.d.P., F.A.C. and F.T.; writing—review and editing, M.A.M., I.d.P., F.A.C. and F.T.; supervision, F.A.C. and F.T.; project administration, F.T.; funding acquisition, F.A.C. and F.T.

**Funding:** This work has been supported by InterregV Sudoe and FEDER programs of European Commission through the COMMANDIA project SOE2/P1/F0638, and by the Spanish Government through the FPU grant FPU15/04446 and the research project RTI2018-094279-B-I00.

**Conflicts of Interest:** The authors declare no conflict of interest.
