# White-Box Neural Ensemble for Vehicular Plasticity: Quantifying the Efficiency Cost of Symbolic Auditability in Adaptive NMPC

arXiv: https://arxiv.org/abs/2602.01516 · HTML: https://mlsanity.com/paper/2602.01516/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 11 of its 17 references that are in the map.
- First posted on arXiv 2 Feb 2026.

## Builds on (most cited first, 11 of 11)

- [On the Spectral Bias of Neural Networks](https://mlsanity.com/paper/1806.08734/) (2018)
- [Neural-Fly Enables Rapid Learning for Agile Flight in Strong Winds](https://mlsanity.com/paper/2205.06908/) (2022)
- [acados: a modular open-source framework for fast embedded optimal control](https://mlsanity.com/paper/1910.13753/) (2019)
- [Learning for CasADi: Data-driven Models in Numerical Optimization](https://mlsanity.com/paper/2312.05873/) (2023)
- [Deep Dynamics: Vehicle Dynamics Modeling with a Physics-Constrained Neural Network for Autonomous Racing](https://mlsanity.com/paper/2312.04374/) (2023)
- [BayesRace: Learning to race autonomously using prior experience](https://mlsanity.com/paper/2005.04755/) (2020)
- [Gaussian processes for dynamics learning in model predictive control](https://mlsanity.com/paper/2502.02310/) (2025)
- [Composing MPC with LQR and Neural Network for Amortized Efficiency and Stable Control](https://mlsanity.com/paper/2112.07238/) (2021)
- [S2GPT-PINNs: Sparse and Small models for PDEs](https://mlsanity.com/paper/2506.15687/) (2025)
- [AltNet: Addressing the Plasticity-Stability Dilemma in Reinforcement Learning](https://mlsanity.com/paper/2512.01034/) (2025)
- [Position: Certifiable State Integrity Should Be Built from Local Validity, Not Global Scale](https://mlsanity.com/paper/2601.21249/) (2026)

## Abstract (arXiv)

We present a white-box adaptive NMPC architecture that resolves vehicular plasticity (adaptation to varying operating regimes without retraining) by arbitrating among frozen, regime-specific neural specialists using a Modular Sovereignty paradigm. The ensemble dynamics are maintained as a fully traversable symbolic graph in CasADi, enabling maximal runtime auditability. Synchronous simulation validates rapid adaptation (~7.3 ms) and near-ideal tracking fidelity under compound regime shifts (friction, mass, drag) where non-adaptive baselines fail. Empirical benchmarking quantifies the transparency cost: symbolic graph maintenance increases solver latency by 72-102X versus compiled parametric physics models, establishing the efficiency price of strict white-box implementation.