Good experimental work is not novel in robotics: there are many good experimental scientists in our community. However, uniformly good experimental work and reporting has not yet been achieved. We will present the work of good experimental groups and use it to illustrate high-quality experimental work and reporting, using the school format to draw out the strengths (and weaknesses) of presented work. Our pool of prospective lecturers include some of the former workshop participants and speakers that proposed the most promising ideas and results in order to present their methodological proposals and practical examples as well as some of the leading experts coordinating relevant initiatives.
Both replication and benchmarking are needed to foster a cumulative advancement of robotics, and even to correctly appreciate disruptive innovation in the science and technology of robots. The partecipants will learn how experimental results can be replicable and refutable on the one hand, and quantitatively comparable according to community- endorsed metrics, on the other hand. Thus, the partecipants will learn:
- How to implement approaches with exhaustive experimental methodology and apply appropriate benchmarking procedures to compare the actual practical results with reference to standard accepted procedures.
- How to define viable procedures for the replication of robotics research results and the best practice to conduct experiments. As part of the experimental protocols we include the procedures, data, code and hardware description.
- How to enable the replication of experiments by providing the proper kind and amount of data, as a prerequisite to quantitative comparison of capabilities. A key point to allow replication and comparison of results is having adequate data support: all the data necessary to repeat a given experiment, how to achieve it with today's digital media will be addressed.
- What are the issues -from practical to epistemological ones- involved in the suitable reporting process for fully replicable robotics research.
We will share hands-on knowledge and describe in full detail the R-article process, and how to write an R(eproducible)-article and replicate it, with practical examples from already published ones.
We will work out concrete and practical examples of R-articles and how to reproduce them.
We will organize remote experiments.
You will learn how to do remote simulation experiments of ROS systems and also how to use the Python Robotics Toolbox.
You will learn how to use Virtual Machines, Containers and Dockers.
You will be able to test remotely your control algorithms on real hardware (H2Arm), that you may cheaply 3D print at your site.
The school details will be updated. Check back soon!