Celso M. de Melo

About

Celso M. de Melo is a multidisciplinary researcher whose work sits at the dynamic intersection of human-robot interaction, computer vision, and social decision-making. His research explores how humans and autonomous machines — including robots and self-driving vehicles — communicate, cooperate, and navigate morally complex situations together. De Melo has made notable contributions to gesture-based robot control, developing deep learning approaches that leverage synthetic data to overcome the costly challenge of annotating real-world training sets, a paper that has garnered 30 citations. His investigations into the social psychology of human-machine interaction reveal how people instinctively distinguish between human and artificial agents, influencing cooperation in measurable ways (26 citations). He has also advanced aerial video action recognition through innovative auto-zoom and temporal reasoning techniques deployable on resource-constrained UAVs, work recognized with 19 citations. Beyond perception and control, de Melo examines the ethical dimensions of autonomous systems, including how machines should navigate injury-risk moral dilemmas (17 citations) and how emotional expression can calibrate appropriate levels of human-agent cooperation. His more recent explorations into LLM-driven robotic task planning signal a forward-looking research trajectory. Collectively, his work shapes how society can thoughtfully and safely integrate autonomous machines into everyday life.

Research Focus

Key Achievements

Top Papers

1. 1
   Vision-Based Gesture Recognition in Human-Robot Teams Using Synthetic Data

   30 citations · 2020
2. 2
   Cooperation with autonomous machines through culture and emotion

   26 citations · 2019
3. 3
   AZTR: Aerial Video Action Recognition with Auto Zoom and Temporal Reasoning

   19 citations · 2023
4. 4
   Risk of Injury in Moral Dilemmas With Autonomous Vehicles

   17 citations · 2021
5. 5
   Inferring Intentions from Emotion Expressions in Social Decision Making

   11 citations · 2019
6. 6
   Real-time human action recognition from aerial videos using autozoom and synthetic data

   3 citations · 2024
7. 7
   ConceptAgent: LLM-Driven Precondition Grounding and Tree Search for Robust Task Planning and Execution

   2 citations · 2025
8. 8
   Emotional Expression Help Regulate the Appropriate Level of Cooperation with Agents

   1 citations · 2024

Key Collaborators

Contact & Links