Distribution-Agnostic Robust Trajectory Optimization via Chance-Constrained Reinforcement Learning

2026-06-11 · Source: Machine Learning · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Robotics & Autonomous Systems · Depth: Expert, quick

Summary

A new distribution-agnostic robust trajectory-optimization framework is introduced, leveraging chance-constrained reinforcement learning to manage uncertainty from initial conditions and process noise. The approach first computes a deterministic nominal trajectory offline, then uses reinforcement learning to robustify this baseline through a structured affine closed-loop correction law, incorporating feedforward control adjustments and time-varying feedback gains. Probabilistic feasibility is empirically enforced via rollout-based upper-tail quantiles, while terminal dispersion is controlled using covariance-feasibility penalties. The framework was assessed on two distinct problems: a three-dimensional multi-impulse Earth-Mars transfer and a stochastic atmospheric pinpoint rocket landing. Results demonstrate competitive upper-tail fuel cost and preserved probabilistic feasibility, highlighting the robustification scaffold's portability across diverse spacecraft trajectory planning scenarios without core stochastic-control structure redesign.

Key takeaway

For Aerospace Engineers designing complex spacecraft trajectories, this framework offers a robust method to manage uncertainty without redesigning core stochastic-control structures. You can compute a nominal path and then efficiently robustify it using reinforcement learning, ensuring probabilistic feasibility and controlled terminal dispersion even with unseen disturbances. Consider applying this approach to multi-impulse transfers or pinpoint landings to enhance mission reliability and reduce fuel costs under uncertain conditions.

Key insights

A chance-constrained reinforcement learning framework robustifies nominal trajectories against sampled uncertainty using affine closed-loop corrections.

Principles

• Robust trajectory optimization can correct a nominal baseline.
• Probabilistic feasibility is enforceable via rollout-based quantiles.
• A single robustification scaffold can span diverse problems.

Method

Compute a nominal trajectory offline, then use RL to learn an affine closed-loop correction law with feedforward adjustments and time-varying feedback gains, enforcing probabilistic feasibility via upper-tail quantiles and regulating terminal dispersion with covariance penalties.

In practice

• Design robust policies for Earth-Mars transfers.
• Adapt for stochastic rocket landing problems.
• Evaluate policies under diverse uncertainty distributions.

Topics

• Trajectory Optimization
• Reinforcement Learning
• Chance-Constrained Control
• Robust Control
• Spacecraft Trajectory Planning
• Stochastic Control

Best for: Research Scientist, AI Scientist, Robotics Engineer

Related on AIssential

• Distribution-Agnostic Robust Trajectory Optimization via Chance-Constrained Reinforcement Learning — A chance-constrained reinforcement learning framework robustifies spacecraft trajectory optimization against diverse uncertainties using an affine closed-loop correction.
• Robust Exploratory Stopping under Ambiguity in Reinforcement Learning — A robust RL framework for optimal stopping under ambiguity balances exploration and exploitation using $g$-expectation and BSDEs.
• Transit Network Design with Two-Level Demand Uncertainties: A Machine Learning and Contextual Stochastic Optimization Framework — Integrating machine learning with contextual stochastic optimization improves transit network design by modeling two levels of demand uncertainty.
• Model-Driven Policy Optimization in Differentiable Simulators via Stochastic Exploration — MDPO enhances differentiable planning by adaptively injecting stochastic noise for improved exploration in complex optimization landscapes.
• Rule-based High-Level Coaching for Goal-Conditioned Reinforcement Learning in Search-and-Rescue UAV Missions Under Limited-Simulation Training — A hybrid rule-based and RL framework enhances UAV mission safety and efficiency in limited-simulation SAR.
• WOMBET: World Model-based Experience Transfer for Robust and Sample-efficient Reinforcement Learning — WOMBET unifies uncertainty-aware model-based data generation with adaptive online fine-tuning for efficient RL experience transfer.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Machine Learning.