# Nine textbook driving planners outscore a learned planner in simulation, study says

Andreas Look says the best textbook planner scored 41.8 in the HUGSIM driving simulator, up from 32.0 for UniAD’s learned planner with the same perception system.

Robotics · By ITANU Magazine newsroom · Published 2026-11-20T17:33:15.000Z · 1 min read

## Key points

- Look changed only the planner, keeping UniAD’s perception system fixed.
- Look found the largest gains in encounters with stationary or oncoming cars.
- The best textbook planner finished the route in 31 of the 182 runs where UniAD’s planner hit a car, Look reported.

Andreas Look reported in an arXiv preprint submitted on 18 November that nine of 16 textbook driving planners outscored UniAD’s learned planner in the HUGSIM simulator [1]. All used the same system to perceive their surroundings [2].

UniAD’s developers say it combines perception, prediction and planning in one network [4]. Look replaced only the planner, the component that chooses the vehicle’s path and how to move along it [1]. He gave the textbook planners no high-definition map [1].

Look reported a driving score of 41.8 for the best textbook planner, compared with 32.0 for UniAD’s planner [2]. HUGSIM’s developers say they calculate the score by multiplying a vehicle’s average score over a run by the fraction of the route it completes [5].

Look found the largest gains when the test vehicle encountered stationary or oncoming cars, and no gain when other cars steered toward it [1]. He reported that UniAD’s planner hit a car in 182 runs and that the best textbook planner completed the route in 31 of those runs [1].

## Sources

1. [Old-School Planners Drive (Sometimes) Better Than Learned Ones](https://arxiv.org/abs/2611.25138), arXiv (Primary)
2. [Old-School Planners Drive (Sometimes) Better Than Learned Ones](https://arxiv.org/html/2611.25138v1), arXiv (Primary)
3. [HUGSIM: A Real-Time, Photo-Realistic and Closed-Loop Simulator for Autonomous Driving](https://xdimlab.github.io/HUGSIM/), XDimLab (Primary)
4. [Planning-oriented Autonomous Driving](https://arxiv.org/abs/2212.10156), arXiv (Primary)
5. [HUGSIM: A Real-Time, Photo-Realistic and Closed-Loop Simulator for Autonomous Driving](https://arxiv.org/html/2412.01718), arXiv (Primary)

ITANU’s newsroom system produced this article with AI and checked it against the sources listed above.

Canonical: https://itanu.org/magazine/textbook-driving-planners-outscore-learned-planner-in-simulation-qlohx26l/
