Learning Contractive Dynamical Representations for Composite Adaptive Control
基於收縮動力學表徵學習的複合自適應控制框架
This paper presents a representation-learning framework for composite adaptive tracking control under coupled disturbances. By linking classical disturbance-accommodating control (DAC) with adaptive disturbance-rejection, the authors introduce a hard-EM procedure with a Kalman smoother to identify contractive dynamical representations of latent disturbances. Validated on a slippery ground vehicle with sloshing liquid and a pendulum, the method achieves superior tracking and predictive performance over LTI and PD baselines.
Key points
Bridging DAC and Adaptive Learning
Connects classical disturbance-accommodating control (DAC) with modern last-layer adaptive disturbance rejection methods.
Statistically Principled Contractive Learning
Uses a hard-EM procedure with a Kalman smoother to identify latent disturbance representations that are uniformly contractive.
Provable Exponential Convergence
Combined with Bayesian filtering, the controller guarantees provable exponential convergence of tracking errors to a bounded neighborhood.
Robust Physical Validation
Experimentally validated on a slippery ground vehicle with sloshing liquid/pendulum loads and a coupled Duffing oscillator system.
How it works
Why it matters
This research bridges control theory and machine learning for physical systems. While pure machine learning lacks safety guarantees and classical control struggles with complex nonlinear disturbances (e.g., fluid sloshing), this framework offers both adaptive predictive capabilities and mathematically provable stability. It has highly practical implications for safety-critical applications like autonomous driving, aerial transport, and robotics.
Who it affects
- AI Researcher
- AI Developer
- Enterprise Leader
How to use it
- 1Trajectory control for autonomous vehicles on slippery or icy roads
- 2Drone or robotic manipulation transporting sloshing liquids or suspended payloads
- 3Precision industrial positioning systems experiencing highly coupled nonlinear disturbances
Limitations & caveats
- The expectation-maximization (EM) optimization and Kalman smoothing pose high computational overhead during representation learning.
- Convergence is only guaranteed to a bounded neighborhood rather than achieving zero asymptotic tracking error.
Related
RPG Framework: Guided Self-Improvement Boosts Embodied Agent Success to 95% Without Weight Updates
自主學習免微調！「RPG」框架藉由虛擬練習與自我診斷，將機器人任務成功率提升至 95%
The RPG framework enables autonomous robot improvement without updating model weights, boosting task success from 28.6% to 95.0% through simulation practice, failure diagnosis, and real-world transfer.
DynaHarness: A Dynamic Physical Harness for Self-Evolving Robot Agents
DynaHarness：具備自我演化能力的機器人代理動態實體約束框架
DynaHarness is a dynamic physical framework for self-evolving robots that bridges semantic reasoning and execution through a contract, transforming failures into capability updates.
Skill-Space Shooting: Autonomous Robot Policy Improvement Guided by Foundation Models
技能空間射擊演算法：利用基礎模型引導機器人自主進行策略優化
This research introduces 'skill-space shooting,' a method that leverages foundation models to guide exploration in a reusable skill space, enabling robots to autonomously correct and improve their policies.