Specification-Aware Distribution Shaping for Robotics Foundation Models

TL;DR

This paper introduces a framework that ensures robotics foundation models adhere to complex temporal and safety specifications during execution, enhancing reliability without altering the original model.

Contribution

It presents a novel specification-aware distribution shaping method that enforces Signal Temporal Logic constraints in pretrained robotics models without parameter modification.

Findings

Successfully enforced complex STL constraints in simulation

Improved safety and specification satisfaction during robot execution

Demonstrated effectiveness across multiple environments

Abstract

Robotics foundation models have demonstrated strong capabilities in executing natural language instructions across diverse tasks and environments. However, they remain largely data-driven and lack formal guarantees on safety and satisfaction of time-dependent specifications during deployment. In practice, robots often need to comply with operational constraints involving rich spatio-temporal requirements such as time-bounded goal visits, sequential objectives, and persistent safety conditions. In this work, we propose a specification-aware action distribution optimization framework that enforces a broad class of Signal Temporal Logic (STL) constraints during execution of a pretrained robotics foundation model without modifying its parameters. At each decision step, the method computes a minimally modified action distribution that satisfies a hard STL feasibility constraint by reasoning over the remaining horizon using forward dynamics propagation. We validate the proposed framework in simulation using a state-of-the-art robotics foundation model across multiple environments and complex specifications.