arXiv subjects

Francisco Cruz

Publications and source records attributed to Francisco Cruz.

At least 19 recordsLinked to original sources

arXiv2026

Initiation Safety: A Missing Dimension in Generalist-Robot Safety

Safety for generalist robots is usually discussed in terms of motion or dialogue. We argue a third question is missing: should the robot take its first hard-to-undo social action at all, such as a greeting, an uninvited grasp, or stepping into someone's space? We call this initiation authorization. Current frameworks rarely treat it as a separate safety layer. Today's stacks often skip this step: a high engagement score or a confident VLA rollout is treated as permission to act. But seeing a person is not the same as having their consent to be addressed. We frame initiation authorization within generalist-robot safety and contrast it with post-plan VLA guardrails, implementing PAS (probe-authorize-speak) on a doorway humanoid, comparing it with direct-init on logged traces, and proposing a three-condition user study, with open questions on metrics, governance, and where initiation ends and foundation-model generation begins.

cs.RO↗

arXiv2026

Memory Merge DQN: Sensitivity Weighted Target Updates for Stable Value Learning

Deep Q-networks use target networks to stabilise bootstrapped value learning, but the standard hard copy update also introduces a tradeoff. Holding the target network fixed, improves short term stability, yet each hard update abruptly replaces the target parameters with the newest online network and discards recent parameter history. This can produce sudden changes in the bootstrap target and may remove value function structure that remains useful later in training. This paper introduces Memory Merge DQN, a target network update mechanism that maintains a short memory of recent historical online network copies and constructs the target network by merging network parameters based on the Q-value sensitivity rather than copying only the newest online network. Memory Merge gives greater influence to parameters that remain locally important for current Q-value behaviour, while using a recency prior to keep the merged target close to the latest online parameters. The method is inspired by Fisher Weight Model Merging, but uses Q-value sensitivity rather than Fisher information as the weighting signal. This paper evaluates Memory Merge DQN on Atari environments against DQN, Averaged DQN, DQN with layer normalisation, and PQN (with gradient clipping). The results show that Memory Merge DQN is highly competitive and it achieves the largest number of first place final performance results among the evaluated methods, beats DQN, Averaged DQN, and PQN (with gradient clipping), and produces substantial gains in several games where preserving useful value-function parameters appears beneficial. These findings suggest that selectively merging recent parameter weights and history can improve the stability and final performance of DQN agents, and that target network design is an important mechanism for preserving useful value function structure during long horizon value learning.

cs.LG↗

arXiv2026

Generating Natural and Expressive Robot Gestures through Iterative Reinforcement Learning with Human Feedback using LLMs

Expressive gestures are essential for natural and effective communication, complementing speech when verbal cues alone are insufficient (e.g., pointing). For social robots such as the humanoid Pepper, producing natural and expressive movements is critical for improving human-robot interaction (HRI) and long-term acceptance. However, generating gestures remains challenging due to reliance on expert-authored animations, resulting in rigid behaviors that are impractical for dynamic and diverse environments. Alternatively, machine learning approaches often struggle to capture perceived naturalness, becoming increasingly challenging with more degrees of freedom. Consequently, producing expressive robot gestures requires a system that can adapt to the environment while adhering to social norms and physical constraints. Recent advances in large language models (LLMs) enable dynamic code generation, offering new opportunities for runtime gesture synthesis from natural language. In this paper, we integrate ChatGPT into the humanoid robot Pepper to generate co-speech gestures aligned with conversational output. While this baseline enables flexible gesture generation, the resulting motions are often perceived as stiff and unnatural. To address this limitation, we introduce an iterative reinforcement learning with human feedback (RLHF) system that finetunes gesture generation based on user evaluations, leveraging an iterative user study to compare Pepper's generated gestures. Our results show that RLHF improved the LLM's co-speech generative capabilities, producing more expressive, relevant and fluid movements.

cs.RO↗

arXiv2026

Infant Spontaneous Movement Noise Improves Exploration in Deep RL

Exploration in deep reinforcement learning (RL) is commonly implemented as temporally uncorrelated white noise. However, recent works show that temporally correlated colored noise can improve exploration efficiency by producing smooth trajectories with better coverage of the state space. We inquire whether action noise inspired by infant spontaneous movements can also improve exploration in deep RL. We find that the power spectral densities of babies' end-effector velocities follow a colored noise process where the spectral exponent increases with age. Inspired by this developmental pattern, we introduce a mechanism that progressively increases the temporal auto-correlation of exploration noise during RL training, matching the infant statistics. Experiments across several RL environments show that infant-inspired noise produces structured exploratory behavior and can improve learning efficiency compared to conventional exploration strategies. These findings suggest that human motor and cognitive development can provide useful guidance for designing learning mechanisms in artificial agents. Our code is available at https://github.com/trieschlab/baby-noise-rl.

cs.LG↗

arXiv2026

Self-Predictive Representation for Autonomous UAV Object-Goal Navigation

Autonomous Unmanned Aerial Vehicles (UAVs) have revolutionized industries through their versatility with applications including aerial surveillance, search and rescue, agriculture, and delivery. Their autonomous capabilities offer unique advantages, such as operating in large open space environments. Reinforcement Learning (RL) empowers UAVs to learn intricate navigation policies, enabling them to optimize flight behavior autonomously. However, one of its main challenge is the inefficiency in using data sample to achieve a good policy. In object-goal navigation (OGN) settings, target recognition arises as an extra challenge. Most UAV-related approaches use relative or absolute coordinates to move from an initial position to a predefined location, rather than to find the target directly. This study addresses the data sample efficiency issue in solving a 3D OGN problem, in addition to, the formalization of the unknown target location setting as a Markov decision process. Experiments are conducted to analyze the interplay of different state representation learning (SRL) methods for perception with a model-free RL algorithm for planning in an autonomous navigation system. The main contribution of this study is the development of the perception module, featuring a novel self-predictive model named AmelPred. Empirical results demonstrate that its stochastic version, AmelPredSto, is the best-performing SRL model when combined with actor-critic RL algorithms. The obtained results show substantial improvement in RL algorithms' efficiency by using AmelPredSto in solving the OGN problem.

cs.RO↗

arXiv2025

Towards Senior-Robot Interaction: Reactive Robot Dog Gestures

As the global population ages, many seniors face the problem of loneliness. Companion robots offer a potential solution. However, current companion robots often lack advanced functionality, while task-oriented robots are not designed for social interaction, limiting their suitability and acceptance by seniors. Our work introduces a senior-oriented system for quadruped robots that allows for more intuitive user input and provides more socially expressive output. For user input, we implemented a MediaPipe-based module for hand gesture and head movement recognition, enabling control without a remote. For output, we designed and trained robotic dog gestures using curriculum-based reinforcement learning in Isaac Gym, progressing from simple standing to three-legged balancing and leg extensions, and more. The final tests achieved over 95\% success on average in simulation, and we validated a key social gesture (the paw-lift) on a Unitree robot. Real-world tests demonstrated the feasibility and social expressiveness of this framework, while also revealing sim-to-real challenges in joint compliance, load distribution, and balance control. These contributions advance the development of practical quadruped robots as social companions for the senior and outline pathways for sim-to-real adaptation and inform future user studies.

cs.RO↗

arXiv2025

Conservative Bias in Multi-Teacher Learning: Why Agents Prefer Low-Reward Advisors

Interactive reinforcement learning (IRL) has shown promise in enabling autonomous agents and robots to learn complex behaviours from human teachers, yet the dynamics of teacher selection remain poorly understood. This paper reveals an unexpected phenomenon in IRL: when given a choice between teachers with different reward structures, learning agents overwhelmingly prefer conservative, low-reward teachers (93.16% selection rate) over those offering 20x higher rewards. Through 1,250 experimental runs in navigation tasks with multiple expert teachers, we discovered: (1) Conservative bias dominates teacher selection: agents systematically choose the lowest-reward teacher, prioritising consistency over optimality; (2) Critical performance thresholds exist at teacher availability rho >= 0.6 and accuracy omega >= 0.6, below which the framework fails catastrophically; (3) The framework achieves 159% improvement over baseline Q-learning under concept drift. These findings challenge fundamental assumptions about optimal teaching in RL and suggest potential implications for human-robot collaboration, where human preferences for safety and consistency may align with the observed agent selection behaviour, potentially informing training paradigms for safety-critical robotic applications.

cs.RO↗

arXiv2025

LAPX: Lightweight Hourglass Network with Global Context

Human pose estimation is a crucial task in computer vision. Methods that have SOTA (State-of-the-Art) accuracy, often involve a large number of parameters and incur substantial computational cost. Many lightweight variants have been proposed to reduce the model size and computational cost of them. However, several of these methods still contain components that are not well suited for efficient deployment on edge devices. Moreover, models that primarily emphasize inference speed on edge devices often suffer from limited accuracy due to their overly simplified designs. To address these limitations, we propose LAPX, an Hourglass network with self-attention that captures global contextual information, based on previous work, LAP. In addition to adopting the self-attention module, LAPX advances the stage design and refine the lightweight attention modules. It achieves competitive results on two benchmark datasets, MPII and COCO, with only 2.3M parameters, and demonstrates real-time performance, confirming its edge-device suitability.

cs.CV↗

arXiv2025

Towards Closing the Domain Gap with Event Cameras

Although traditional cameras are the primary sensor for end-to-end driving, their performance suffers greatly when the conditions of the data they were trained on does not match the deployment environment, a problem known as the domain gap. In this work, we consider the day-night lighting difference domain gap. Instead of traditional cameras we propose event cameras as a potential alternative which can maintain performance across lighting condition domain gaps without requiring additional adjustments. Our results show that event cameras maintain more consistent performance across lighting conditions, exhibiting domain-shift penalties that are generally comparable to or smaller than grayscale frames and provide superior baseline performance in cross-domain scenarios.

cs.CV↗

arXiv2025

SWIFT-Nav: Stability-Aware Waypoint-Level TD3 with Fuzzy Arbitration for UAV Navigation in Cluttered Environments

Efficient and reliable UAV navigation in cluttered and dynamic environments remains challenging. We propose SWIFT-Nav: Stability-aware Waypoint-level Integration of Fuzzy arbitration and TD3 for Navigation, a TD3-based navigation framework that achieves fast, stable convergence to obstacle-aware paths. The system couples a sensor-driven perception front end with a TD3 waypoint policy: the perception module converts LiDAR ranges into a confidence-weighted safety map and goal cues, while the TD3 policy is trained with Prioritised Experience Replay to focus on high-error transitions and a decaying epsilon-greedy exploration schedule that gradually shifts from exploration to exploitation. A lightweight fuzzy-logic layer computes a safety score from radial measurements and near obstacles, gates mode switching and clamps unsafe actions; in parallel, task-aligned reward shaping combining goal progress, clearance, and switch-economy terms provides dense, well-scaled feedback that accelerates learning. Implemented in Webots with proximity-based collision checking, our approach consistently outperforms baselines in trajectory smoothness and generalization to unseen layouts, while preserving real-time responsiveness. These results show that combining TD3 with replay prioritisation, calibrated exploration, and fuzzy-safety rules yields a robust and deployable solution for UAV navigation in cluttered scenes.

cs.RO↗

arXiv2025

Ensemble Elastic DQN: A Step Dependent Ensemble Approach for Reducing Overestimation in Deep Value-Based Reinforcement Learning

Deep Q-Networks (DQN) can suffer from overestimation bias because bootstrapped targets use a maximisation operation over noisy value estimates. Ensemble-based methods and multi-step methods have each been used to improve the stability and sample efficiency of value-based reinforcement learning, but their interaction remains less well understood. This paper introduces Ensemble Elastic DQN (EEDQN), a value-based reinforcement learning algorithm that combines adaptive elastic multi-step returns with ensemble-based target aggregation. EEDQN replaces the clustering-based state similarity test used in earlier Elastic Step DQN with a lightweight Q-value difference rule, making adaptive return construction simpler to apply in discrete control settings. The method then applies horizon dependent ensemble aggregation, one-step targets use the ensemble mean, while longer elastic returns use the ensemble minimum. This design aims to reduce extreme optimistic bootstrap estimates without making every update uniformly conservative. We evaluate EEDQN on five MinAtar environments against DQN, Double DQN, Averaged DQN, MaxMin DQN, and Elastic Step DQN. EEDQN achieves the highest final return in four of the five environments and remains competitive with conservative ensemble baselines. An aggregation rule ablation shows that the best degree of conservatism is environment dependent, suggesting that adaptive return length and ensemble aggregation interact in non-trivial ways.

cs.LG↗

arXiv2025

When Large Language Models are More PersuasiveThan Incentivized Humans, and Why

Large Language Models (LLMs) have been shown to be highly persuasive, but when and why they outperform humans is still an open question. We compare the persuasiveness of two LLMs (Claude 3.5 Sonnet and DeepSeek v3) against humans who had incentives to persuade, using an interactive, real-time conversational setting. We demonstrate that LLMs persuasive superiority is context-dependent: it depends on whether the persuasion attempt is truthful (towards the right answer) or deceptive (towards the wrong answer) and on the LLM model, and wanes over repeated interactions (unlike human persuasiveness). In our first large-scale experiment, humans vs LLMs (Claude 3.5 Sonnet) interacted with other humans who were completing an online quiz for a reward, attempting to persuade them toward a given (either correct or incorrect) answer. Claude was more persuasive than incentivized human persuaders both in truthful and deceptive contexts and it significantly increased accuracy if persuasion was truthful, but decreased it if persuasion was deceptive. In a follow-up experiment with Deepseek v3, we replicated the findings about accuracy but found greater LLM persuasiveness only if the persuasion was deceptive. Linguistic analyses of the persuaders texts suggest that these effects may be due to LLMs expressing higher conviction than humans.

cs.CL↗

arXiv2025

PERCY: Personal Emotional Robotic Conversational System

Traditional rule-based conversational robots, constrained by predefined scripts and static response mappings, fundamentally lack adaptability for personalized, long-term human interaction. While Large Language Models (LLMs) like GPT-4 have revolutionized conversational AI through open-domain capabilities, current social robots implementing LLMs still lack emotional awareness and continuous personalization. This dual limitation hinders their ability to sustain engagement across multiple interaction sessions. We bridge this gap with PERCY (Personal Emotional Robotic Conversational sYstem), a system designed to enable open-domain, multi-turn dialogues by dynamically analyzing users' real-time facial expressions and vocabulary to tailor responses based on their emotional state. Built on a ROS-based multimodal framework, PERCY integrates a fine-tuned GPT-4 reasoning engine, combining textual sentiment analysis with visual emotional cues to accurately assess and respond to user emotions. We evaluated PERCY's performance through various dialogue quality metrics, showing strong coherence, relevance, and diversity. Human evaluations revealed PERCY's superior personalization and comparable naturalness to other models. This work highlights the potential for integrating advanced multimodal perception and personalization in social robot dialogue systems.

cs.HC↗

arXiv2024

AI Apology: A Critical Review of Apology in AI Systems

Apologies are a powerful tool used in human-human interactions to provide affective support, regulate social processes, and exchange information following a trust violation. The emerging field of AI apology investigates the use of apologies by artificially intelligent systems, with recent research suggesting how this tool may provide similar value in human-machine interactions. Until recently, contributions to this area were sparse, and these works have yet to be synthesised into a cohesive body of knowledge. This article provides the first synthesis and critical analysis of the state of AI apology research, focusing on studies published between 2020 and 2023. We derive a framework of attributes to describe five core elements of apology: outcome, interaction, offence, recipient, and offender. With this framework as the basis for our critique, we show how apologies can be used to recover from misalignment in human-AI interactions, and examine trends and inconsistencies within the field. Among the observations, we outline the importance of curating a human-aligned and cross-disciplinary perspective in this research, with consideration for improved system capabilities and long-term outcomes.

cs.CY↗

arXiv2024

How Can LLMs and Knowledge Graphs Contribute to Robot Safety? A Few-Shot Learning Approach

Large Language Models (LLMs) are transforming the robotics domain by enabling robots to comprehend and execute natural language instructions. The cornerstone benefits of LLM include processing textual data from technical manuals, instructions, academic papers, and user queries based on the knowledge provided. However, deploying LLM-generated code in robotic systems without safety verification poses significant risks. This paper outlines a safety layer that verifies the code generated by ChatGPT before executing it to control a drone in a simulated environment. The safety layer consists of a fine-tuned GPT-4o model using Few-Shot learning, supported by knowledge graph prompting (KGP). Our approach improves the safety and compliance of robotic actions, ensuring that they adhere to the regulations of drone operations.

cs.RO↗

arXiv2024

Attention-Enhanced Lightweight Hourglass Network for Human Pose Estimation

Pose estimation is a critical task in computer vision with a wide range of applications from activity monitoring to human-robot interaction. However,most of the existing methods are computationally expensive or have complex architecture. Here we propose a lightweight attention based pose estimation network that utilizes depthwise separable convolution and Convolutional Block Attention Module on an hourglass backbone. The network significantly reduces the computational complexity (floating point operations) and the model size (number of parameters) containing only about 10% of parameters of original eight stack Hourglass network. Experiments were conducted on COCO and MPII datasets using a two stack hourglass backbone. The results showed that our model performs well in comparison to six other lightweight pose estimation models with an average precision of 72.07. The model achieves this performance with only 2.3M parameters and 3.7G FLOPs.

cs.CV↗

arXiv2024

MERCI: Multimodal Emotional and peRsonal Conversational Interactions Dataset

The integration of conversational agents into our daily lives has become increasingly common, yet many of these agents cannot engage in deep interactions with humans. Despite this, there is a noticeable shortage of datasets that capture multimodal information from human-robot interaction dialogues. To address this gap, we have recorded a novel multimodal dataset (MERCI) that encompasses rich embodied interaction data. The process involved asking participants to complete a questionnaire and gathering their profiles on ten topics, such as hobbies and favorite music. Subsequently, we initiated conversations between the robot and the participants, leveraging GPT-4 to generate contextually appropriate responses based on the participant's profile and emotional state, as determined by facial expression recognition and sentiment analysis. Automatic and user evaluations were conducted to assess the overall quality of the collected data. The results of both evaluations indicated a high level of naturalness, engagement, fluency, consistency, and relevance in the conversation, as well as the robot's ability to provide empathetic responses. It is worth noting that the dataset is derived from genuine interactions with the robot, involving participants who provided personal information and conveyed actual emotions.

cs.HC↗

arXiv2024

Adaptive Alignment: Dynamic Preference Adjustments via Multi-Objective Reinforcement Learning for Pluralistic AI

Emerging research in Pluralistic Artificial Intelligence (AI) alignment seeks to address how intelligent systems can be designed and deployed in accordance with diverse human needs and values. We contribute to this pursuit with a dynamic approach for aligning AI with diverse and shifting user preferences through Multi Objective Reinforcement Learning (MORL), via post-learning policy selection adjustment. In this paper, we introduce the proposed framework for this approach, outline its anticipated advantages and assumptions, and discuss technical details about the implementation. We also examine the broader implications of adopting a retroactive alignment approach through the sociotechnical systems perspective.

cs.LG↗