DANiLab

Embodied Interpretability: Linking Causal Understanding to Generalization in Vision-Language-Action Models

Fri, 03 Jul 2026 00:00:00 +0000

Intent-Handover: Grounding Language in Human-Usage Regions for Trustworthy Robot-to-Human Handovers

Thu, 02 Jul 2026 00:00:00 +0000

R2HandoverSim: A Simulation Framework and Benchmark for Robot-to-Human Object Handovers

Wed, 01 Jul 2026 00:00:00 +0000

Our Embodied Interpretability Paper Got Accepted by ICML 2026

Thu, 07 May 2026 00:00:00 +0000

Our latest work has been accepted at ICML 2026. As VLA models become a key route towards general-purpose robot policies, we should not only ask whether a robot completes a task, but also why it chooses an action. Sometimes a policy can appear to work while relying on the wrong visual cues, such as background texture, lighting, or shortcuts in the scene. Our work provides a way to test this more directly: which parts of the image actually change the robot’s action? This helps researchers diagnose VLA failure modes and better understand why some policies generalise to new environments while others do not.

Embodied Interpretability: Linking Causal Understanding to Generalization in Vision-Language-Action Models

Project page: robot-future.github.io/vla-explain
Preprint: available here
Code: available soon for plug-and-play use

Many congratulations to my PhD students Hanxin Zhang, Mingshuo Xu and Abdulqader Dhafer, and to our co-authors Shigang Yue and Hongbiao Dong. We are also grateful to the reviewers for their constructive and helpful feedback.

Robots + Hardware

Tue, 10 Mar 2026 00:00:00 +0000

Agibot Visits the University of Leicester

Wed, 18 Feb 2026 00:00:00 +0000

We recently welcomed visitors from Agibot to the University of Leicester.

During the visit, Agibot brought several robots and demonstrated their capabilities to researchers and students. The team also had the opportunity to explore our robotics platforms in the lab, including robotic arms, quadruped robots, and aerial systems.

It was a pleasure to host the team and share our research environment. We look forward to staying in touch and exploring opportunities for future engagement.

PhD Opportunity in World Models for Space Robotics

Sun, 01 Feb 2026 00:00:00 +0000

Project Highlights

Develop next-generation embodied AI methods, including world models, uncertainty-aware learning, and safe decision-making, for long-horizon robotic autonomy under extreme uncertainty.

Combine theory with hands-on experimentation through validation on real robotic platforms (e.g., robotic arms) performing contact-rich inspection, repair, assembly, and servicing tasks in simulation, laboratory or space-analogue environments.

Enable trustworthy and certifiable autonomous robotics for future space infrastructure, with broader impact across aerospace, nuclear, offshore, and other safety-critical domains.

Project Summary

Robotics is recognised as a key enabler for future in-space servicing, inspection, and assembly, yet current systems remain largely dependent on teleoperation, scripted behaviours, or short-horizon control. Learning-based methods improve flexibility but often fail to generalise, lack predictive reasoning about contact dynamics, and provide limited safety guarantees, challenges that are exacerbated in extreme environments with degraded sensing, altered physics, and high failure costs.

This PhD will develop world-model-based and uncertainty-aware approaches for embodied intelligence, enabling robots to predict, plan, and adapt under severe uncertainty. The research focuses on learning predictive world models of environment and contact dynamics to support simulation-driven planning, risk-sensitive control, and online adaptation for improved robustness and safety.

All methods will be validated experimentally on physical robotic platforms performing contact-rich servicing tasks in laboratory and space-analogue environments, ensuring real-world performance beyond simulation.

The student will

develop world-model-based and learning-based algorithms for perception, prediction, planning, and control
implement and evaluate methods on real robotic hardware and simulators
design rigorous experiments and quantitative benchmarks
publish in leading AI and robotics venues
contribute to open-source and collaborative research projects

The project will deliver fundamental advances in embodied intelligence and practical autonomy technologies for ISAM and other safety-critical applications.

Required Skills and Qualifications

First-class or high 2:1 degree (or international equivalent), ideally with a Master’s, in Robotics, Computer Science, Aerospace, Electrical Engineering, or a closely related discipline
Strong foundations in robotics and control (kinematics, dynamics, state estimation, feedback control)
Experience with robotic simulation and experimentation (e.g., ROS/ROS2, Gazebo/Mujoco, or HIL systems)
Proficiency in Python and/or C++ for algorithm development and system integration

Desirable

Knowledge of machine learning methods for robotics (e.g., reinforcement or model-based learning) and ability to conduct experimental evaluation
Background in space robotics, GNC, or safety-critical systems
Prior research outputs (e.g., publications)

Applicants are strongly encouraged to contact the supervisors of the GTA projects to discuss their suitability and eligibility before they apply.

Apply

Find the project listing on FindAPhD.

Apply on the University of Leicester Computer Science GTA page.

Touch-Ex: A Region-Level, Force-Annotated Visuo-Tactile Dataset

Thu, 01 Jan 2026 00:00:00 +0000

We received a £485k prestigious UKSA grant to develop the first robotic system for space welding.

Thu, 04 Dec 2025 00:00:00 +0000

See news at:

UK Space Agency: £17 million to drive the next wave of space innovation

UK Space Agency: Scottish space innovation secures investment

University and SPL news:

Space Park Leicester: new funding for the first robots to weld in space

University of Leicester news

Meet Porygon - Our AI Workstation

Fri, 31 Oct 2025 00:00:00 +0000

Team finally welcomed our AI workstation, powered by dual RTX Pro 6000 Blackwell GPUs, giving us roughly 200 GB of combined VRAM for our Vision–Language Model (VLM) and World-Model research.

New Lab Member GO2-EDU

Thu, 28 Aug 2025 00:00:00 +0000

🐾 Say hello to our newest lab member! 🐾

We’re thrilled to welcome the Unitree GO2-EDU quadruped robot — or as we like to call it, our new robotic dog 🐕🤖.

The GO2 brings incredible agility, smart sensors, and a lot of personality into the lab. It’s going to help us explore new frontiers in robotics, AI, and human–robot interaction. From testing advanced navigation to experimenting with autonomous behaviours, this little powerhouse is ready for some big adventures.

We can’t wait to share updates, demos, and maybe even a few fun tricks as our robotic pup settles in. Stay tuned — the journey’s just beginning! 🚀

Deep Transformers for Analyzing BOF Steelmaking Data

Sun, 01 Jun 2025 00:00:00 +0000

vSTMD: Visual Motion Detection for Extremely Tiny Target at Various Velocities

Thu, 01 May 2025 00:00:00 +0000

EPSRC funded PhD Open

Mon, 20 Jan 2025 00:00:00 +0000

Application Deadline:

University of Leicester — EPSRC DLA funded opportunity

Project title: Embodied Intelligence for Trustworthy Robotic Autonomy in Unconstructed Environments

Background: This PhD project addresses the challenge of achieving trustworthy robotic autonomy in dynamic, unstructured environments—such as hazardous manufacturing facilities, orbital and planetary space robotic operations, or deep-sea installations. While today’s robots can perform impressive tasks, they often lack the safety, reliability, and transparency required to earn human trust. This research focuses on Embodied Intelligence, emphasising accurate perception and spatial understanding to guide robots’ decision-making and adaptability amid real-world uncertainties.

State-of-the-art methods integrate Vision-Language Models (VLMs) and multi-agent architectures, encoding sensor and control inputs into shared representations that inform actions. However, these approaches have rarely been tested in highly unstructured scenarios, often struggling with the Sim2Real gap and limited generalisation beyond their initial training conditions. To address these gaps, this project will combine Multimodal Large Language Models (MLLMs), reinforcement learning, and synthetic scenario generation. Industrial datasets and generative AI techniques will support the creation of diverse, photorealistic training conditions that better reflect complex operational environments.

The application link and project details and further project details are available online.

Application deadline: 31 January 2025 Start date: 22 September 2025

The Conversation - How AI can help you make a computer game without knowing anything about coding

Mon, 14 Oct 2024 00:00:00 +0000

I published an article discussing how AI is completely reshaping coding—even kids can code now.

Read the article on The Conversation.

UK in Space Festival 2024

Fri, 27 Sep 2024 00:00:00 +0000

Read more on the University of Leicester news page.

The technology behind self-driving cars is playing a key role in future space missions, according to a team of scientists who have been demonstrating the importance of AI and robotics in space exploration.

The University of Leicester’s DriverLeics Team from the School of Computing and Mathematical Sciences proudly represented the institution at the UK in Space Festival held at the National Space Centre. The event, which featured major UK space industry leaders such as Airbus and Astroscale, celebrated the nation’s achievements and advancements in space exploration.

Led by Dr Daniel Z. Hao and accompanied by Dr Neslihan Suzen, the team included Computer Science students Abdulqader A.M. Dhafer and Ansel M.J. Ong. They showcased the University’s cutting-edge research in computer science and its significant contributions to UK space education and research.

Dr Hao said: “We are privileged to be invited to such an important event to celebrate the UK’s presence in space. It’s an excellent opportunity to explain to the general public and families how self-driving cars are relevant to Mars Rovers, and how we are planning to deploy AI and robotic dogs for space exploration.”

The DriverLeics Team demonstrated how advancements in autonomous vehicle technology can be applied to space missions. By drawing parallels between self-driving cars and Mars Rovers, they highlighted the role of artificial intelligence and robotics in navigating and exploring extraterrestrial terrains.

Their demonstration includes showing a robotic dog, emphasizing the potential of these technologies to enhance mission capabilities and safety for future space exploration. The initiative underscores the University of Leicester’s commitment to innovation and excellence in the field of computer science.

Dr Neslihan Suzen added: “Engaging with the public at events like the UK in Space Festival allows us to inspire the next generation of scientists and engineers. It’s a fantastic platform to showcase how our school’s research contributes to the broader space exploration efforts of the UK.”

Credit: University of Leicester

The Conversation - AI could be the breakthrough that allows humanoid robots to jump from science fiction to reality

Fri, 09 Aug 2024 00:00:00 +0000

I published an article discussing the current and near-future development of humanoids. Read the article on The Conversation.

AI and robotics scientists spark inspiration at major school STEM fair

Wed, 19 Jun 2024 00:00:00 +0000

We had the opportunity to walk through an exciting future of robots and artificial intelligence as we attended the largest school-based STEM Fair, Bright Sparks, at Leicester Grammar School with Team DriverLeics and their robot dog.

Leicester Grammar School hosted the Bright Sparks STEM Fair 2024 on 15 June, recognized as the largest school-based STEM Fair in the country, which usually draws thousands of visitors from throughout the county and beyond. The fair was a vibrant hub of scientific exploration, featuring over 70 exhibitors from local businesses, universities, and STEM organisations.

As invited exhibitors, we joined the DriverLeics team, comprised of academics from the University of Leicester College of Science and Engineering.

Learn more on the University of Leicester news page.

Dr Hao is giving a talk for PINT OF SCIENCE FESTIVAL14 May 2024

Wed, 08 May 2024 00:00:00 +0000

Get your ticket on the Pint of Science event page.

How AI powered robots will change our life in the next 5 years.

Dr. Daniel Hao (Lecturer in AI and Robotics)

In the next five years, robots that think and learn like us are going to change the way we live in some pretty exciting ways. Imagine having a robot buddy that can help out with everything from making dinner to keeping our streets clean. This talk is all about how these smart robots are stepping out of sci-fi movies and into our homes and jobs. We’ll chat about how they’re already starting to help doctors perform surgeries with laser precision, take over the boring tasks in factories, and even do chores around the house so we have more time for fun. But it’s not all smooth sailing; we’ll also touch on some tricky questions, like what happens to jobs when robots can do them too, and how we make sure our robotic helpers respect our privacy. Get ready to peek into a future where robots are part of the family, making our lives easier and more interesting.

We have a funded Robotics and AI PhD post!

Wed, 08 May 2024 00:00:00 +0000

Find the details on the project listing on FindAPhD.

About the Project

GTA funded PhD studentship in Computing

Highlights

Advance State-of-the-Art in Imitation Learning: Elevate existing imitation learning methodologies (such as GAIL, Q-learning and Meta-learning) to enhance robotic capabilities in complex manufacturing processes.
Develop a Robotic Imitation Learning Framework for Manufacturing: Create a pioneering imitation learning framework tailored for robotic applications in manufacturing operations, including precision tasks like welding and machining.
Curate a Digital Manufacturing Dataset: Support the curation of a vast digital dataset derived from the DigitalMetal consortium’s findings, aimed at supporting and advancing robotic manufacturing research.

Project

Robotic technology plays a pivotal role in the UK’s digital strategy, enhancing the digital and data roadmap for the transition from traditional manufacturing. This shift aims to boost the sector’s competitiveness, sustainability, and innovation.

Despite advances, experienced engineers remain crucial in manufacturing processes like welding and machining. Our goal is to equip robots with the ability to learn from human manufacturing experts, enabling them to perform complex tasks intelligently. While there have been successes in applying artificial intelligence, such as imitation learning, to teach robots, challenges remain in refining learning approaches (from data collection to learning algorithms) and improving the performance accuracy and robustness of autonomous robotic operations.

Project Objectives

This project aims to address these challenges by:

Modelling the robotic manufacturing process and applying techniques like domain randomisation and adaptation to narrow the gap between simulated environments and real-world applications.
Collaborating with relevant EPSRC funded projects to organize and collect manufacturing data.
Developing few-shot learning or meta-learning approaches to enhance the practicality of imitation learning.
Advancing state-of-the-art imitation learning methods, such as Inverse Reinforcement Learning (IRL) and Generative Adversarial Imitation Learning (GAIL), and exploring integrated approaches with reinforcement learning for processes like welding.

Requirements for Candidates

In addition to meeting the University’s PhD degree entry requirements, potential candidates should possess a relevant degree and/or experience in robotics and artificial intelligence. Preference will be given to applicants with research experience in imitation learning or reinforcement learning and familiarity with the Robot Operating System (ROS).

Opportunities for the Successful Candidate

The selected candidate will have access to facilities including the Universal Robotics UR5e arm, 3D printers, high-performance computing, and more. They will join the University of Leicester’s extensive digital metal and robotics network, benefiting from additional resources and support. The project will be jointly supervised by the School of Computing and Mathematical Sciences and the School of Engineering. The candidate may access the data resources from the UK’s DigitalMetal Consortium.

PhD start date: 23 September 2024

Enquiries

Contact the project supervisor, Dr Daniel Z. Hao, at d(dot)hao(at)leicester(dot)ac(dot)uk or cmspgr(at)le(dot)ac(dot)uk.

Further details and application advice are on the University of Leicester CMS GTA page.

Funding Notes

The 4 year GTA funded studentships provide:

A combined teaching and stipend payment, currently £19,237 per year for 2024/25
Tuition fees at UK rates

International students are welcome to apply but must be able to demonstrate that they can fund the difference between UK and International fees for the duration of their study. The difference in fees payable by the student will be £18,864 (2024/25) per year of study.

Our work on UKSA PLATOR Project Published

Fri, 26 Apr 2024 00:00:00 +0000

This joint work between University of Surrey and University of Leicester has been presented on the International Symposium on Space Flight Dynamics (ISSFD) at ESOC Darmstadt.

The paper title: Trajectory Design and Navigation Analysis of a PLAsma TOrch Rocket Propelled Space Tug

PLATOR is a new electrothermal thruster for space logistics applications, developed by the University of Surrey and the University of Leicester. This paper describes the technology behind the development of the thruster and presents a mission scenario where a PLATOR-propelled spacecraft is used to capture and de-orbit the European Space Agency (ESA)’s Envisat satellite. The orbital transfer trajectory is designed using a time-optimal control approach, and the spacecraft’s state vector’s uncertainties are assessed through a covariance analysis. A navigation analysis is then performed to evaluate the spacecraft’s capability to autonomously track its motion during the transfer using GPS measurements. Finally, a target proximity phase is then simulated to demonstrate the spacecraft’s capability to rendezvous and dock with Envisat, using the uncertainties obtained from the covariance analysis, showing the potential of the PLATOR thruster for in-orbit servicing and active debris removal applications.

CDT in DigitalMetal (£18m) Funded! We are part of it!

Tue, 02 Apr 2024 00:00:00 +0000

A new training centre at the University of Leicester aims to boost the metals industry with the skills in data and artificial intelligence to take on the global competition of the future.

Dr Hao is the Co-I of this CDT.

The new £18 million Centre for Doctoral Training (CDT) in Digital Transformation of Metals Industry (DigitalMetal) has been funded by the Engineering and Physical Sciences Research Council (EPSRC), who announced £7m funding today (12 March), five partner universities (Birmingham, Leicester, Loughborough, Nottingham and Warwick) and industry.

More information is on the University of Leicester news page.

2024 DriverLeics Cohort!

Mon, 04 Mar 2024 00:00:00 +0000

Today, we had the first official kick-off meeting for the 2024 DriverLeics cohort! Looking forward to seeing the student team create some big buzz in 2024!

Giving a talk for MedRIFT

Tue, 13 Feb 2024 00:00:00 +0000

I am giving a talk about AI and Robotics for MedRIFT society tonight. Feel free to pop-in and have a conversation.

How robot making an espresso

Mon, 05 Feb 2024 00:00:00 +0000

We did some cool demonstration of using UR5e to make an Nespresso. Imitation learning + LLM way to go!

Welcome our new lab member - UR5e robot arm

Wed, 24 Jan 2024 00:00:00 +0000

Installed the new lab member with the DANiLab members.

Unitree Robotics Visit (B2 dog + H1 Humanoid)

Tue, 23 Jan 2024 00:00:00 +0000

Staff and students really enjoyed it! B2 robot dog is a great and robust platform and H1 will be an interesting research platform for RLs.

Join the lab

Mon, 24 Oct 2022 00:00:00 +0000

Team

Mon, 24 Oct 2022 00:00:00 +0000

Mon, 01 Jan 0001 00:00:00 +0000