Assistive Teaching of Motor Control Tasks to Humans

Authors: Megha Srivastava, Erdem Biyik, Suvir Mirchandani, Noah Goodman, Dorsa Sadigh

NeurIPS 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Through an extensive mix of synthetic and user studies on two motor control tasks parking a car with a joystick and writing characters from the Balinese alphabet we show that assisted teaching with skills improves student performance by around 40% compared to practicing full trajectories without skills, and practicing with individualized drills can result in up to 25% further improvement.
Researcher Affiliation Academia Megha Srivastava Stanford University megha@cs.stanford.edu Erdem Bıyık UC Berkeley ebiyik@berkeley.edu Suvir Mirchandani Stanford University suvir@cs.stanford.edu Noah D. Goodman Stanford University ngoodman@stanford.edu Dorsa Sadigh Stanford University dorsa@cs.stanford.edu
Pseudocode Yes Algorithm 1 Diverse Scenario Selection Input: Skill labels M e ξ for each ξ Ξe Input: Number of scenarios to be selected N s ... Algorithm 2 Individual Expertise Identification ... Algorithm 3 Individualized Drill Creation
Open Source Code Yes Our source code is available at https://github.com/Stanford-ILIAD/teaching.
Open Datasets Yes WRITING: We introduce a novel task of writing Balinese character sequences from the Omniglot dataset [41] ... PARKING: We use the Parking environment from Highway Env [42]
Dataset Splits Yes Students in both environments follow a sequence of pre-test scenarios, practice sessions (including skills or drill-based practice), and evaluation scenarios. ... We measure and report Reward Improvement, or the difference in average reward across 5 random evaluation scenarios and 2 random pre-test scenarios.
Hardware Specification No The main paper does not specify any particular hardware details such as GPU models, CPU models, or specific cluster configurations used for running experiments. While the checklist indicates this information is in the Appendix, it is not present in the main body of the paper.
Software Dependencies No The paper mentions 'Stable Baselines [43] implementation of Soft Actor-Critic' but does not provide specific version numbers for this or any other software dependencies. The full software details are likely deferred to the Appendix, but not included in the main text.
Experiment Setup Yes Drill Creation. We create one drill for each latent skill identified by SKILLEXTRACTOR for both PARKING (n = 3, Nrep = 1, Ntarget = 7, Ndrills = 1) and WRITING (n = 2, Nrep = 3, Ntarget = 8, Ndrills = 1) tasks.