RoCKIn benchmarks require that the robot logs some of its own internal data. This document explains the modalities of this logging. ***WHAT DATA MUST BE SAVED The data that the robot must save are specified by the Competition wiki. They differ from benchmark to benchmark. Please note that some data streams (those with the highest bitrate) must be logged only in the time intervals when they are actually used by the robot to perform the activities required by the benchmark. In this way, system load and data bulk are minimized. For instance, whenever a benchmark includes object recognition activities, video and point cloud data must be logged by the robot only in the time intervals when it is actually performing object recognition. ***WHAT WE DO WITH THE DATA These data are not used during the Competition. In particular, they are not used for scoring. The data are processed by RoCKIn after the end of the competition; they are used for in-depth analyses and/or to produce datasets to be published for the benefit of the robotics community. ***WHERE AND WHEN THE ROBOT MUST SAVE THE DATA Robots must save the data specified by the Competition wiki on a USB stick provided by RoCKIn. The USB stick is given to the team immediately before the start of the benchmark, and must be returned (with the required data on it) at the end of the benchmark. ***DIRECTORY NAMES Each time a team's robot executes a benchmark, the team must: 1. Create, in the root directory of the USB stick, a new directory named NameOfTheTeam_FBMx_DD_HH-MM (for Functionality Benchmarks) or NameOfTheTeam_TBMx_DD_HH-MM (for Task Benchmarks) where x is the number of the Functionality or Task Benchmark DD is the day of the month (28, 29 or 30 for the 2014 Competition) HH, MM represent the time of the day (hours and minutes) 2. Configure the robot to save the data files in such directory. All files produced by the robot that are associated to the execution of the benchmark must be written in this directory. Please note that a new directory must be created for each benchmark executed by the robot. This holds true even when the benchmark is a new run of one that the robot already executed.