Lamarck's Driving School: Discovering Autonomous Driving Training Strategies through Evolutionary Competition

2026-10-08

AI Summary

This study addresses the resource inefficiency caused by predefined proxy metrics in autonomous driving training by proposing a Lamarckian evolutionary framework. Abandoning conventional proxy metrics, the method employs a multi-stage bilevel optimization architecture that integrates Darwinian crossover, mutation, and selection with Lamarckian inheritance mechanisms to achieve co-evolution of training scenarios and model capabilities. Furthermore, it reveals the evolutionary dynamics of staged challenges to distill lightweight, reusable strategies. Experimental results demonstrate that the complete framework reduces performance degradation by 25.07%, while the lightweight strategy achieves a 19.13% reduction, significantly outperforming existing baseline methods.

Abstract

Autonomous driving capabilities depend strongly on the distribution of scenarios encountered during training. Existing methods commonly construct or dynamically adapt training scenario distributions using surrogate criteria such as realism, difficulty, or risk. However, these predefined surrogates may misrepresent training value, leading to inefficient use of training resources. To address this limitation, we propose a Lamarckian evolutionary framework that replaces surrogate-based guidance with competition among candidate distributions. We formulate training strategy discovery as a multi-stage bilevel optimization problem and use Lamarckian evolution algorithm to approximate its solution. At the outer level, Darwinian crossover, mutation, and selection explore the scenario distribution space; at the inner level, policy learning acquires new capabilities, and Lamarckian inheritance transfers them to subsequent stages, allowing scenario distributions and policy capabilities to co-evolve. The resulting evolutionary trajectories reveal recurring stage-wise regularities among high-value distributions, characterized by capability accumulation through stage-wise challenge rotation. We further distill these regularities into a lightweight, reusable Lamarckian Training Strategy. Experiments show that, compared with the baseline, the complete framework reduces performance loss by up to 25.07%, while the lightweight strategy still achieves a 19.13% reduction. These results demonstrate that evolutionary competition can both discover effective training strategies and reveal reusable stage-wise patterns in how the value of training distributions changes with policy capability. Code is available on GitHub.

Problem

• Autonomous Driving
• Training Strategy
• Scenario Distribution
• Surrogate Criteria
• Training Efficiency

Innovation

• Lamarckian evolution
• bilevel optimization
• autonomous driving
• scenario distribution
• evolutionary competition