Research map / Label-efficient and robust learning / Knowledge distillation
A Virtual Driving Instructor That Generates Personalized Driving Lessons Based on Student Skill Level
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Presents a Lesson generator module for a virtual driving instructor that procedurally builds personalized driving lessons (maps, tiles, events, dynamic traffic, paths) from a Bayesian student skill model, evaluated via a simulated-student case study.
Problem
Driver education relies on costly, inefficient one-on-one instructor lessons and constrained real-world traffic situations; the paper targets automatically tailoring driving lessons to each student's skill level to speed learning and cover needed skills.
Approach
A stochastic, heuristic procedural map generator that draws skill and tile utilities (weighted toward low-proficiency skills from a Bayesian skill network) to assemble chunks into a personalized route, with dynamic traffic difficulty scaled to skill level and adjustable online.
Key results
- Case study with four simulated student profiles shows generated routes differ as expected across profiles.
- For a high-proficiency student with weaknesses (highway, queue, tunnel), generated lessons consistently train those weak skills most across ten consecutive lessons.
- Lesson generator has not yet been tested on human students; evaluation is purely simulation-based.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Generate lessons using student skill levels to preferentially select tiles for low-proficiency skills | Lesson generator · Case study: high-proficiency student with weaknesses | prioritization of weak skills for training | Helped |
Limitations
- Evaluated only on synthetic students, not human learners.
- Only offline generation tested; online adaptation not evaluated.
- Tiles are not production quality and only one tile per type was designed.
Experimental setup
Datasets: expert-generated dataset from driving instructors for conditional probability tables (used for the skill network)
Metrics: skill distribution of trained skills per lesson alignment of generated lesson with student weaknesses (radar
Baselines: Ropelato et al. 2018 adaptive tutoring system
Builds on: Ropelato et al. 2018 adaptive tutoring for VR driving Greitzer et al. 2007 serious game design guidelines Togelius et al. 2011 procedural content generation classification
Closest papers by meaning
- The Virtual Driving Instructor: Multi-Agent System Collaborating via Knowledge Graph for Scalable Driver Education (AAAI 2024 · similarity 0.88)
- Mimicking the Maestro: Exploring the Efficacy of a Virtual AI Teacher in Fine Motor Skill Acquisition (AAAI 2024 · similarity 0.78)
- Generating Traffic Scenarios via In-Context Learning to Learn Better Motion Planner (AAAI 2025 · similarity 0.77)
- Knowledge Distillation from Single-Task Teachers to Multi-Task Student for End-to-End Autonomous Driving (AAAI 2024 · similarity 0.77)
- LASIL: Learner-Aware Supervised Imitation Learning For Long-term Microscopic Traffic Simulation (CVPR 2024 · similarity 0.77)
- SubjectDrive: Scaling Generative Data in Autonomous Driving via Subject Control (AAAI 2025 · similarity 0.76)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.