# A Parametric Model for Near-Optimal Online Synthesis with Robust Reach-Avoid Guarantees

arXiv: https://arxiv.org/abs/2504.01006 · HTML: https://mlsanity.com/paper/2504.01006/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Placed in Safe and constrained RL, one of 15 directions in Deep reinforcement learning methods.
- Builds on 8 of its 22 references that are in the map.
- First posted on arXiv 1 Apr 2025.

## Placed in

- Deep reinforcement learning methods / [Safe and constrained RL](https://mlsanity.com/deeprl/safe-and-constrained-rl/)

## Builds on (most cited first, 8 of 8)

- [Safe, Multi-Agent, Reinforcement Learning for Autonomous Driving](https://mlsanity.com/paper/1610.03295/) (2016)
- [Explainability in Deep Reinforcement Learning](https://mlsanity.com/paper/2008.06693/) (2020)
- [Provably Safe Deep Reinforcement Learning for Robotic Manipulation in Human Environments](https://mlsanity.com/paper/2205.06311/) (2022)
- [Formal Verification of Obstacle Avoidance and Navigation of Ground Robots](https://mlsanity.com/paper/1605.00604/) (2016)
- [Verified Synthesis of Optimal Safety Controllers for Human-Robot Collaboration](https://mlsanity.com/paper/2106.06604/) (2021)
- [Safe and Scalable Real-Time Trajectory Planning Framework for Urban Air Mobility](https://mlsanity.com/paper/2306.11647/) (2023)
- [Layered controller synthesis for dynamic multi-agent systems](https://mlsanity.com/paper/2307.06758/) (2023)
- [Solvability of Approximate Reach-Avoid Games](https://mlsanity.com/paper/2502.04544/) (2025)

## Abstract (arXiv)

Objective: To obtain explainable guarantees in the online synthesis of optimal controllers for high-integrity cyber-physical systems, we re-investigate the use of exhaustive search as an alternative to reinforcement learning. Approach: We model an application scenario as a hybrid game automaton, enabling the synthesis of robustly correct and near-optimal controllers online without prior training. For modal synthesis, we employ discretised games solved via scope-adaptive and step-pre-shielded discrete dynamic programming. Evaluation: In a simulation-based experiment, we apply our approach to an autonomous aerial vehicle scenario. Contribution: We propose a parametric system model and a parametric online synthesis.