arXiv subjects

Ronald Cumbal

Publications and source records attributed to Ronald Cumbal.

5 recordsLinked to original sources

arXiv2026

Lightweight Visual Reasoning for Socially-Aware Robots

Robots operating in shared human environments must not only navigate, interact, and detect their surroundings, they must also interpret and respond to dynamic, and often unpredictable, human behaviours. Although recent advances have shown promise in enhancing robotic perception and instruction-following using Vision-Language Models (VLMs), they remain limited in addressing the complexities of multimodal human-robot interactions (HRI). Motivated by this challenge, we introduce a lightweight language-to-vision feedback module that closes the loop between an LLM and the vision encoder in VLMs. The module projects image-token hidden states through a gated Multi-Layer Perceptron (MLP) back into the encoder input, prompting a second pass that reinterprets the scene under text context. We evaluate this approach on three robotics-centred tasks: navigation in a simulated environment (Habitat), sequential scene description (Mementos-Robotics), and human-intention recognition (our HRI dataset). Results show that our method improves Qwen 2.5 (7B) by $3.3\%$ (less distance), $+0.057$ description score, and $+2.93\%$ accuracy, with less than $3\%$ extra parameters; Gemma 3 (4B) and LLaVA OV 1.5 (4B) show mixed navigation results but gains $+0.111,+0.055$ and $+10.81\%,+4.79\%$ on the latter two tasks. Code is available at https://github.com/alessioGalatolo/VLM-Reasoning-for-Robotics

cs.RO↗

arXiv2026

A Collaborative Crowdsourcing Method for Designing External Interfaces for Autonomous Vehicles

Participatory design effectively engages stakeholders in technology development but is often constrained by small, resource-intensive activities. This study explores a scalable complementary method, enabling broad pattern identification in the design for interfaces in autonomous vehicles. We implemented a human-centered, iterative process that combined crowd creativity, structured participatory principles, and expert feedback. Across iterations, participant concepts evolved from simple cues to multimodal systems. Novel suggestions ranged from personalized features, like tracking lights, to inclusive elements like haptic feedback, progressively refining designs toward greater contextual awareness. To assess outcomes, we compared representative designs: a popular-design, reflecting the most frequently proposed ideas, and an innovative-design, merging participant innovations with expert input. Both were evaluated against a benchmark through video-based simulations. Results show that the popular-design outperformed the alternatives on both interpretability and user experience, with expert-validated innovations performing second best. These findings highlight the potential of scalable participatory methods for shaping emerging technologies.

cs.HC↗

arXiv2025

Why Report Failed Interactions With Robots?! Towards Vignette-based Interaction Quality

Although the quality of human-robot interactions has improved with the advent of LLMs, there are still various factors that cause systems to be sub-optimal when compared to human-human interactions. The nature and criticality of failures are often dependent on the context of the interaction and so cannot be generalized across the wide range of scenarios and experiments which have been implemented in HRI research. In this work we propose the use of a technique overlooked in the field of HRI, ethnographic vignettes, to clearly highlight these failures, particularly those that are rarely documented. We describe the methodology behind the process of writing vignettes and create our own based on our personal experiences with failures in HRI systems. We emphasize the strength of vignettes as the ability to communicate failures from a multi-disciplinary perspective, promote transparency about the capabilities of robots, and document unexpected behaviours which would otherwise be omitted from research reports. We encourage the use of vignettes to augment existing interaction evaluation methods.

cs.RO↗

arXiv2025

Crowdsourcing eHMI Designs: A Participatory Approach to Autonomous Vehicle-Pedestrian Communication

As autonomous vehicles become more integrated into shared human environments, effective communication with road users is essential for ensuring safety. While previous research has focused on developing external Human-Machine Interfaces (eHMIs) to facilitate these interactions, we argue that involving users in the early creative stages can help address key challenges in the development of this technology. To explore this, our study adopts a participatory, crowd-sourced approach to gather user-generated ideas for eHMI designs. Participants were first introduced to fundamental eHMI concepts, equipping them to sketch their own design ideas in response to scenarios with varying levels of perceived risk. An initial pre-study with 29 participants showed that while they actively engaged in the process, there was a need to refine task objectives and encourage deeper reflection. To address these challenges, a follow-up study with 50 participants was conducted. The results revealed a strong preference for autonomous vehicles to communicate their awareness and intentions using lights (LEDs and projections), symbols, and text. Participants' sketches prioritized multi-modal communication, directionality, and adaptability to enhance clarity, consistently integrating familiar vehicle elements to improve intuitiveness.

cs.HC↗

arXiv2024

You don't understand me!: Comparing ASR results for L1 and L2 speakers of Swedish

The performance of Automatic Speech Recognition (ASR) systems has constantly increased in state-of-the-art development. However, performance tends to decrease considerably in more challenging conditions (e.g., background noise, multiple speaker social conversations) and with more atypical speakers (e.g., children, non-native speakers or people with speech disorders), which signifies that general improvements do not necessarily transfer to applications that rely on ASR, e.g., educational software for younger students or language learners. In this study, we focus on the gap in performance between recognition results for native and non-native, read and spontaneous, Swedish utterances transcribed by different ASR services. We compare the recognition results using Word Error Rate and analyze the linguistic factors that may generate the observed transcription errors.

cs.CL↗