Research

Robots that perceive people, understand instructions and act safely.

Four connected themes, six research contributions, and a clear line between what has been validated on hardware and what is still in development.

Theme A

Human–Robot Interaction & Multimodal Control

This is the core of my research. I build pipelines that track a person's hands and body in real time, recognise intent and map it onto continuous control of a 7-DoF Franka Emika Panda. The engineering challenge is less about recognition accuracy in isolation and more about the full loop: latency, calibration, dead-man safety gates, compliant motion and graceful failure when tracking drops out.

Established work

Real-time hand-gesture recognition mapped to multistep manipulation tasks
Body-pose teleoperation through MoveIt Servo with calibration and dead-man control
Integration of perception with ROS 2, MoveIt 2 and the Franka control stack

Current & emerging

Fusing gesture and speech into one interaction channel
User studies of operator workload and trust

AI-RTGM case study Body-pose control code

Theme B

Language Models & Embodied AI

Language models are useful to a robot only when their output is constrained to actions the robot can actually perform. My work uses LLMs as task interpreters that emit structured commands for a validated skill library — first with multilingual (English/German) instructions for domestic tasks, and in simulation as high-level planners on top of learned low-level controllers.

Established work

Multilingual instruction-to-command pipeline on a physical Franka Panda (IEEE ICARM 2025)
Review of large language and vision models for robotic manipulation (SN Computer Science, 2025)

Current & emerging

Local small language models for on-premise, privacy-preserving HRI
Vision–language–action models and safety-aware execution constraints

Multilingual manipulation RL + LLM hybrid

Theme C

Robot Learning, Manipulation & Motion Planning

I am interested in combining what learning does well — perception, task selection, adaptation — with what classical robotics already does reliably: kinematics, trajectory generation and collision-aware motion planning. Recent work explores structured action grammars in which each skill carries explicit preconditions and effects, so that a failure can be repaired locally rather than by replanning the whole task.

Established work

Forward and inverse kinematics, Jacobian-based numerical IK
MoveIt 2 motion planning and planning-scene manipulation
RL (PPO, SAC) in PyBullet simulation

Current & emerging

Physics-grounded action grammars with dependency-local repair
Imitation learning from teleoperated demonstrations

Action grammars Teaching repositories

Theme D

Sustainable & Intelligent Robotics

Robots that run all day should not waste energy. I have worked on energy-aware trajectory generation for high-DoF arms, and I am developing directions in robotic disassembly for circular manufacturing, autonomous agriculture and digital twins for intelligent automation. These are emerging directions and I welcome collaborators in them.

Established work

Energy-efficient trajectory planning for high-DoF robots (German Robotics Conference, 2025)

Current & emerging

Robotic disassembly for circular manufacturing
Autonomous systems for smart farming (review in preparation)
Digital twins for intelligent automation

Energy-efficient planning

Expertise

Research areas & technical foundations

The robotics, learning and control methods underpinning my research and teaching.

Robotics fundamentals

Robot kinematics
Robot dynamics
Control systems
Robot simulation
Trajectory planning

AI for robotics

Large language models (LLMs)
Vision–language models (VLMs)
Reinforcement learning (RL)
Multimodal perception
Embodied AI

Manipulation & interaction

Robot manipulation
Human–robot interaction
Robot grasping
Motion planning
ROS 2 & MoveIt 2

Research contributions

Six contributions, from published systems to active prototypes

Each case study sets out the problem, the approach, the evidence and the limits of the work.

Research 01Published · physical robot

AI-RTGM: Real-Time Gesture-Based Human–Robot Interaction

An end-to-end framework that maps human hand gestures to complete manipulation tasks on a 7-DoF Franka Emika Panda.

Concurrency and Computation: Practice and Experience, 2026

Read case study

01Instruction (EN/DE)

02LLM interpreter

03Command vocabulary

04Robot skill library

05Franka Panda

Research 02Published · physical robot

Multilingual Language-Driven Manipulation

English and German instructions, interpreted by an LLM, become executable robot commands for preparing a hot drink.

IEEE ICARM 2025, Portsmouth

057-DoF arm (PyBullet)

Research 03Published · simulation

Energy-Efficient Trajectory Planning for High-DoF Robots

Sinusoidal trajectories with velocity scaling to reduce energy use without sacrificing smoothness or accuracy.

First German Robotics Conference, Nuremberg, 2025 (abstract)

Read case study

01Task in natural language

02LLM planner

03Sub-goals

04RL policy (PPO/SAC)

05Panda in PyBullet

Research 04Published · simulation

Hybrid RL and LLM Framework for Robotic Manipulation

Language models plan the task; reinforcement-learning policies carry out the motion.

IEEE EmergIN 2025 · published conference paper

Read case study

Simulated Franka Panda stacking three coloured blocks in PyBullet

Research 05Active research · prototype

Physics-Grounded Action Grammars

Robot skills as typed atoms with preconditions and effects — so a failed step can be repaired locally.

Ongoing research · public prototype on GitHub

Research 06Doctoral milestone

PhD: Machine-Learning-Driven Multimodal Control for HRI

A doctoral programme connecting perception, language and control into one framework for human–robot interaction.

University of Brighton · awarded June 2026

Read case study

Talks & academic activities

Selected talks and conference activity

Invited Speaker

8–9 December 2025

Abu Dhabi AI-Robotics Conference (AIRoC) 2025 — Dialogue on Humanoids

“Multimodal Human–Robot Interaction Using Gesture and Language Models”

Invited talk on combining gesture recognition and language models for intuitive control of collaborative manipulators, at a conference organised by the Robotics Department of Mohamed bin Zayed University of Artificial Intelligence (MBZUAI).

Abu Dhabi, United Arab Emirates

Speaker list Conference

Attendee1–4 July 2026
18th International Conference on Social Robotics (ICSR + ART 2026)
University of London, London, United Kingdom
Attended the international conference on social robotics, human–robot interaction and creative robotics in London.
Conference (opens in a new tab)
Paper1–3 August 2025
IEEE ICARM 2025
Portsmouth, United Kingdom
Language-driven beverage preparation on a Franka Emika Panda from English and German instructions.
Paper
Conference abstractMarch 2025
First German Robotics Conference
Nuremberg, Germany
Sinusoidal trajectory generation with velocity scaling for energy-aware motion.
Abstract
Paper23–26 September 2024
ICRA@40
Rotterdam, Netherlands
Early results of gesture-to-task control on the Franka Emika Panda, the precursor to AI-RTGM.
Paper

Research & collaboration

Where the work happens

My research is based in the AI Robotics (AIR) Lab at the University of Brighton, where I have worked on multimodal control for robotic devices. I am interested in research partnerships that connect human-centred interfaces, robot learning and safe physical manipulation.

I am keen to work with academic and industrial partners on human–robot interaction, embodied AI, robotic manipulation and intelligent automation — including student projects, joint proposals and applied research.

AIR Lab Start a conversation

Collaborate on robotics research

From ideas to robot demonstrations.

I welcome conversations about joint research, applied projects and teaching collaborations involving intelligent robotic systems.

Get in touch