Dataset of a forwarder operating in rough terrain

View PDF of the article FORWARD: A Badland Freight Forwarder’s Dataset, by Mikael Lundup\”ack and 10 other authors

View PDF HTML (beta)

a summary:We present FORWARD, a high-resolution multimodal dataset of a cut-to-length shipping company operating in rough terrain at two harvest sites in the central part of Sweden. The freight forwarder is a large Komatsu model equipped with vehicle information sensors, including GPS satellite navigation, motion sensors, accelerometers, and engine sensors. The vehicle is additionally equipped with cameras, operator vibration sensors and multiple IMUs. The data includes time-records of events recorded at 5 Hz of driving speed, fuel consumption, centimeter-accurate vehicle location, and jack use while the vehicle is operating in forested areas, and is laser scanned at a resolution of approximately 1,500 dots per square metre. Production log files (StanForD standard) with timestamped machine events, extensive video material, and terrain data in various formats are also included. About 18 hours of regular timber extraction work over three days was annotated from 360 video material into individual work items and included in the dataset. We also include scenario specifications for experiments performed on forest roads and terrain. Scenarios include repeatedly driving on the same tracks with or without steel tracks, different payload weights, and different target driving speeds. The dataset aims to develop models and algorithms for passability, perception and autonomous control of forest machines using artificial intelligence, simulation and experiments on physical testbeds. We focus in part on freight forwarders traversing terrain, avoiding or handling obstacles, and loading or unloading logs, while taking into account efficiency, fuel consumption, safety and environmental impact. Other benefits of the open dataset include the ability to explore automatic generation and calibration of forest machinery simulators and descriptions of automation scenarios using data recorded in the field.