Abstract:
Experimental evaluation of navigation algorithms requires physical robots as well as position sensing devices. The common alternative is to use simulations to run the exp...Show MoreMetadata
Abstract:
Experimental evaluation of navigation algorithms requires physical robots as well as position sensing devices. The common alternative is to use simulations to run the experiments. However, simulation often does not provide an accurate prediction of real-world behavior. Therefore, in this paper, we present an innovative approach towards evaluation of navigation algorithms, which does not need physical robots and position sensors to be present at the experimenter's site, but relies on a special remote internet-accessible physical testbed, the “Teleworkbench”, which can be used in order to evaluate as well as uniformly cross-compare algorithms with no need of spending money on hardware or simulation software. More specifically, in this paper we are using the Teleworkbench to evaluate three different path planning algorithms, and compare it with simulation. Different metrics are proposed, such as the path execution time, smoothness and path clearance deviations. Our results clearly illustrate the superiority of the Telework-bench as an evaluation platform in comparison to simulation, which does not provide an accurate prediction of actual physical performance, and thus illustrate both the viability as well as the power of our novel approach.
Published in: 2013 European Conference on Mobile Robots
Date of Conference: 25-27 September 2013
Date Added to IEEE Xplore: 02 January 2014
Electronic ISBN:978-1-4799-0263-7