arXiv subjects

Émiland Garrabé

Publications and source records attributed to Émiland Garrabé.

7 recordsLinked to original sources

arXiv2026

Acceptable Misalignment

There is a specific challenge designers face in developing autonomous robots capable of open-ended learning. This challenge is how one can develop such systems in an ethically sound way when there are many unknowns during design time about their manner of deployment, what they can and will learn, and how these systems will act on the basis of what they have learned. One response to this challenge is to align, as far as possible, the activities and goals of autonomous robots with human values, raising the question of how to make this alignment possible, a question known as the value alignment problem. We argue that the value alignment problem faces a number of challenges that show that full alignment is both theoretically and practically impossible, meaning that misalignment incidents will keep occurring. This undermines current alignment approaches, as the attempt to maximize alignment cannot properly deal with the possibility that the risk of misalignment that remains after alignment methods have been implemented might still be unacceptable to society. To overcome this difficulty, we argue that a shift in perspective is required from a focus on the maximization of alignment to the conditions under which misalignment incidents become acceptable. In this paper, we explore what this shift in perspective entails. We introduce the concept of acceptable misalignment, formulate normative requirements for alignment mechanisms aimed at bringing misalignment incidents to an acceptable level, and we conclude by presenting a new methodological framework that consists of alignment mechanisms that can meet these requirements.

cs.AI↗

arXiv2026

Autonomously Acquiring Robot Manipulation Skills with Language-Driven Quality-Diversity

Quality-diversity (QD) algorithms have been gaining traction in robot learning, where diverse motion primitive libraries allow robots to adapt zero-shot to constraints at deployment time. However, such methods typically require expert designers to write the success condition, fitness and diversity metrics, and this strongly limits the robot's autonomy. On the other hand, existing LLM-based reward-shaping techniques allow robots to learn autonomously but only output single high-performing solutions, limiting the robot's adaptability. In this paper, we propose an approach designed to output diverse motion primitive archives by autonomously leveraging quality-diversity algorithms, only requiring a free-form description of the task in common language. To address the difficulty of designing relevant fitness and diversity metrics, we propose an autonomous exploration mechanism able to reliably output sets of functionals covering the fitness and behavior descriptor (BD) space. First, we pose policy exploration as a functional design problem, where the functional spaces are lower-dimensional than the full BD and fitness spaces, and propose an LLM-based exploration scheme to sample from these low-dimensional spaces without any task-specific prompts, fine-tuning or expert intervention. We adapt a multi-BD variant of the MAP-Elites success (MES) algorithm, designed to leverage the heterogeneous BD samples. Finally, through experiments based on the genesis simulator, we show that our method effectively generates archives of diverse motion primitives, outperforming classical QD algorithms with inferred and hand-written parametrizations on a set of $4$ robotic manipulation tasks.

cs.RO↗

arXiv2025

Free-Gate: Planning, Control And Policy Composition via Free Energy Gating

We consider the problem of optimally composing a set of primitives to tackle planning and control tasks. To address this problem, we introduce a free energy computational model for planning and control via policy composition: Free-Gate. Within Free-Gate, control primitives are combined via a gating mechanism that minimizes variational free energy. This composition problem is formulated as a finite-horizon optimal control problem, which we prove remains convex even when the cost is not convex in states/actions and the environment is nonlinear, stochastic and non-stationary. We develop an algorithm that computes the optimal primitives composition and demonstrate its effectiveness via in-silico and hardware experiments on an application involving robot navigation in an environment with obstacles. The experiments highlight that Free-Gate enables the robot to navigate to the destination despite only having available simple motor primitives that, individually, could not fulfill the task.

math.OC↗

arXiv2025

Enhancing Robustness in Language-Driven Robotics: A Modular Approach to Failure Reduction

Recent advances in large language models (LLMs) have led to significant progress in robotics, enabling embodied agents to better understand and execute open-ended tasks. However, existing approaches using LLMs face limitations in grounding their outputs within the physical environment and aligning with the capabilities of the robot. This challenge becomes even more pronounced with smaller language models, which are more computationally efficient but less robust in task planning and execution. In this paper, we present a novel modular architecture designed to enhance the robustness of LLM-driven robotics by addressing these grounding and alignment issues. We formalize the task planning problem within a goal-conditioned POMDP framework, identify key failure modes in LLM-driven planning, and propose targeted design principles to mitigate these issues. Our architecture introduces an ``expected outcomes'' module to prevent mischaracterization of subgoals and a feedback mechanism to enable real-time error recovery. Experimental results, both in simulation and on physical robots, demonstrate that our approach significantly improves task success rates for pick-and-place and manipulation tasks compared to both larger LLMs and standard baselines. Through hardware experiments, we also demonstrate how our architecture can be run efficiently and locally. This work highlights the potential of smaller, locally-executable LLMs in robotics and provides a scalable, efficient solution for robust task execution.

cs.RO↗

arXiv2023

On a Probabilistic Approach for Inverse Data-Driven Optimal Control

We consider the problem of estimating the possibly non-convex cost of an agent by observing its interactions with a nonlinear, non-stationary and stochastic environment. For this inverse problem, we give a result that allows to estimate the cost by solving a convex optimization problem. To obtain this result we also tackle a forward problem. This leads to the formulation of a finite-horizon optimal control problem for which we show convexity and find the optimal solution. Our approach leverages certain probabilistic descriptions that can be obtained both from data and/or from first-principles. The effectiveness of our results, which are turned in an algorithm, is illustrated via simulations on the problem of estimating the cost of an agent that is stabilizing the unstable equilibrium of a pendulum.

math.OC↗

arXiv2022

CRAWLING: a Crowdsourcing Algorithm on Wheels for Smart Parking

We present the principled design of CRAWLING: a CRowdsourcing Algorirthm on WheeLs for smart parkING. CRAWLING is an in-car service for the routing of connected cars. Specifically, cars equipped with our service are able to {\em crowdsource} data from third-parties, including other cars, pedestrians, smart sensors and social media, in order to fulfill a given routing task. CRAWLING relies on a solid control-theoretical formulation and the routes it computes are the solution of an optimal control problem where cars maximize a reward capturing environmental conditions while tracking some desired behavior. A key feature of our service is that it allows to consider stochastic behaviors, while taking into account streams of heterogeneous data. We propose a stand-alone, general-purpose, implementation of CRAWLING and we show its effectiveness on a set of scenarios aimed at illustrating all the key features of our service. Simulations show that, when cars are equipped with CRAWLING, the service effectively orchestrates the vehicles, making them able to react online to road conditions, minimizing their cost functions. The code implementing our service and to replicate the numerical results is made openly available.

math.OC↗

arXiv2021

On the design of autonomous agents from multiple data sources

This paper is concerned with the problem of designing agents able to dynamically select information from multiple data sources in order to tackle tasks that involve tracking a target behavior while optimizing a reward. We formulate this problem as a data-driven optimal control problem with integer decision variables and give an explicit expression for its solution. The solution determines how (and when) the data from the sources should be used by the agent. We also formalize a notion of agent's regret and, by relaxing the problem, give a regret upper bound. Simulations complement the results.

math.OC↗