The ability to jointly understand the geometry of objects and plan actions for manipulating them is crucial for intelligent agents. This ability is referred to as geometric planning. Recently, many interactive environments have been proposed to evaluate intelligent agents on various skills, however, none of them cater to the needs of geometric planning. PackIt is a virtual environment to evaluate and potentially learn the ability to do geometric planning, where an agent needs to take a sequence of actions to pack a set of objects into a box with limited space.
The SheetCopilot dataset contains 28 evaluation workbooks and 221 spreadsheet manipulation tasks that are applied to these workbooks. These tasks involve diverse atomic actions related to six task categories (i.e. Entry and manipulation, Formatting, Management, Charts, Pivot Table, and Formula). Dataset statistics: 1. Each task possesses one or more ground truth solutions. 2. The lengths of the task instructions range from 20 to 530 characters, with most tasks between 80 and 110 characters. 3. The number of atomic actions required by each task ranges from 1 to 9. Evaluation metrics: 1. Execution success rate, pass rate, and the number of used actions are evaluated to judge the functional correctness and efficiency of a method. 2. A submitted solution is considered correct if the properties to be checked match those of any of the GT solutions of the corresponding task. Please download the full datasets in our Github Repo: https://github.com/BraveGroup/SheetCopilot Thanks for using our dataset!
The dataset concerns toy tasks that a human should teach to a robot. The number of task repetitions is limited in the dataset since the human should demonstrate the task to the robot only a few times. The data contains images (collected with a Kinect camera) and objects' positions (acquired with a motion capture system and shared through CSV files) over time. The collected data concerns tasks such as (i) assembling the legs of a toy table and (ii) building a stack of objects. Both tasks have variations where the human should validate each step through a pen (e.g., to simulate a quality check procedure). Please, check our repository for more information.
This dataset involves a 2D or 3D agent moving from a start to goal pose while interacting with nearby objects. These objects can influence position of the agent via attraction or repulsion forces as well as influence orientation via attraction to object's orientation. This dataset can be used to pre-train general policy behavior, which can be later fine-tuned quickly for a person's specific preferences. Example use-cases include: - self-driving cars maintaining distance from other cars - robot pick-and-place tasks with intermediate subtasks (ie: scanning factory items before dropping them off) Overall, pre-training initial policy behavior to be fine-tuned later is a powerful paradigm and is arguably essential for robots to handle changing environments and user preferences. This is compared to the paradigm of training on massive amounts of data and remaining fixed at test time, hoping that generalization alone will help the agent handle new scenarios.
The Fields2Benhmark dataset is a collection of 350 agricultural fields in vector format manually selected to test agricultural coverage path planning algorithms. The files in this dataset are organized into two folders: - imgs/: contains the satellite images of the fields. - wkt/: has the vector data of the fields. A file with the same name in both folders corresponds to the same field. The first two letters in the name of the field indicate the country the field belongs to. The fields have been extracted from the EuroCrops dataset --specifically from the Netherlands, Estonia and Lithuania --, transformed each field into its own file, and converted them to Well-Known text (wkt). The satellite images were also provided, and they were created using the Mapping Toolbox from Matlab. ## Funding This dataset is provided as part of the project "Fields2Cover: Robust and efficient coverage paths for autonomous agricultural vehicles" (with project number ENPPS.LIFT.019.019 of the research programme Science PPP Fund for the top sectors which is (partly) financed by the Dutch Research Council (NWO). @article{Mier_Fields2Cover_An_open-source_2023, author={Mier, Gonzalo and Valente, João and de Bruin, Sytze}, journal={IEEE Robotics and Automation Letters}, title={Fields2Cover: An Open-Source Coverage Path Planning Library for Unmanned Agricultural Vehicles}, year={2023}, volume={8}, number={4}, pages={2166-2172}, doi={10.1109/LRA.2023.3248439} }