Wolfgang Merkt
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Wolfgang Merkt has a total of 55 co-authors. Here are the top co-authors by number of shared publications:
- Ioannis Havoutis
Oxford Robotics Institute, University of Oxford, Oxford, U.K.
11 shared publications - Sethu Vijayakumar
University of Edinburgh
8 shared publications - Carlo Tiseo
Rehabilitation Research Institute of Singapore 50 Nanyang Avenue N3-01a-01 639798 Singapore Singapore
4 shared publications - Carlos Mastalli
School of Engineering and Physical Sciences, Heriot-Watt University, Edinburgh, U.K.
4 shared publications - Henrique Ferrolho
School of Informatics University of Edinburgh United Kingdom
3 shared publications - Siddhant Gangapurwala
Oxford Robotics Institute, University of Oxford, Oxford, U.K.
3 shared publications
Latest Publications
- Open Access
- 4 April 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 13, 69062-69080
Perceptive Locomotion Through Whole-Body MPC and Optimal Region Selection
Real-time synthesis of legged locomotion maneuvers in challenging industrial settings is still an open problem, requiring simultaneous determination of footsteps locations several steps ahead while generating whole-body motions close to the robot’s limits. State estimation and perception errors impose the practical constraint of fast re-planning motions in a model predictive control (MPC) framework. We first observe that the computational limitation of perceptive locomotion pipelines lies in the combinatorics of contact surface selection. Re-planning contact locations on selected surfaces can be accomplished at MPC frequencies (50-100 Hz). Then, whole-body motion generation typically follows a reference trajectory for the robot base to facilitate convergence. We propose removing this constraint to robustly address unforeseen events such as contact slipping, by leveraging a state-of-the-art whole-body MPC (Croccodyl). Our contributions are integrated into a complete framework for perceptive locomotion, validated under diverse terrain conditions, and demonstrated in challenging trials that push the robot’s actuation limits, as well as in the ICRA 2023 quadruped challenge simulation.
- 4 April 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 13, 69062-69080
Open AccessPerceptive Locomotion Through Whole-Body MPC and Optimal Region Selection
Real-time synthesis of legged locomotion maneuvers in challenging industrial settings is still an open problem, requiring simultaneous determination of footsteps locations several steps ahead while generating whole-body motions close to the robot’s limits. State estimation and perception errors impose the practical constraint of fast re-planning motions in a model predictive control (MPC) framework. We first observe that the computational limitation of perceptive locomotion pipelines lies in the combinatorics of contact surface selection. Re-planning contact locations on selected surfaces can be accomplished at MPC frequencies (50-100 Hz). Then, whole-body motion generation typically follows a reference trajectory for the robot base to facilitate convergence. We propose removing this constraint to robustly address unforeseen events such as contact slipping, by leveraging a state-of-the-art whole-body MPC (Croccodyl). Our contributions are integrated into a complete framework for perceptive locomotion, validated under diverse terrain conditions, and demonstrated in challenging trials that push the robot’s actuation limits, as well as in the ICRA 2023 quadruped challenge simulation.
- Open Access
- 1 January 2025
- Published by S. Karger AG in Journal of Innate Immunity
- Vol. 17 (1) , 522-538
Compartment-Specific NK Cell Phenotypes Reveal Distinct Maturation and Activation States in Inflammatory Arthritis
Introduction: Synovial natural killer (NK) cells contribute to inflammation in arthritis by secreting cytokines and modulating synovial fibroblast activation. The aim of this study was to describe systemic versus local inflammatory changes of NK cell subsets as well as their physical cell-cell interactions in arthritis patients. Methods: Spectral flow cytometry was used to compare paired peripheral blood (PB) and synovial fluid (SF) immune cells from patients with active inflammatory arthritis and healthy controls. Physical cell-cell interactions within tissues were studied by applying a recently developed cellular interaction mapping framework. Results: Our paired approach revealed significant local enrichment of immature and activated NK cells in SF, characterized by elevated markers of early differentiation, immune-checkpoint regulation, and tissue-residency, highlighting tightly controlled immune activation at inflamed sites. Single-cell analysis confirmed heterogeneity within SF-NK cells, suggesting multiple co-existing activation states and developmental stages. PB-NK cells from patients differed profoundly compared to healthy controls, showing less immature NK cell subsets and an enrichment of mature, pro-inflammatory subsets indicative of systemic immune activation. Cellular interaction mapping revealed mainly NK/neutrophil interactions of patients’ NK cells, while interactions with B-cells, T-cells, or monocytes were negligible. T-cells also displayed profound local and systemic alterations. Cellular interaction mapping revealed that next to NK/neutrophil interactions, interactions between B-cells with monocytes and T-cells with neutrophils characterize joint inflammation. Conclusion: This paired high-dimensional analysis revealed systemic and local alterations in NK cell subsets shaped by co-existing developmental stages and immune regulatory mechanisms. Cellular interaction mapping indicated that neutrophils are a main interaction-partner of NK cells in inflamed joints.
- 1 January 2025
- Published by S. Karger AG in Journal of Innate Immunity
- Vol. 17 (1) , 522-538
Open AccessCompartment-Specific NK Cell Phenotypes Reveal Distinct Maturation and Activation States in Inflammatory Arthritis
Introduction: Synovial natural killer (NK) cells contribute to inflammation in arthritis by secreting cytokines and modulating synovial fibroblast activation. The aim of this study was to describe systemic versus local inflammatory changes of NK cell subsets as well as their physical cell-cell interactions in arthritis patients. Methods: Spectral flow cytometry was used to compare paired peripheral blood (PB) and synovial fluid (SF) immune cells from patients with active inflammatory arthritis and healthy controls. Physical cell-cell interactions within tissues were studied by applying a recently developed cellular interaction mapping framework. Results: Our paired approach revealed significant local enrichment of immature and activated NK cells in SF, characterized by elevated markers of early differentiation, immune-checkpoint regulation, and tissue-residency, highlighting tightly controlled immune activation at inflamed sites. Single-cell analysis confirmed heterogeneity within SF-NK cells, suggesting multiple co-existing activation states and developmental stages. PB-NK cells from patients differed profoundly compared to healthy controls, showing less immature NK cell subsets and an enrichment of mature, pro-inflammatory subsets indicative of systemic immune activation. Cellular interaction mapping revealed mainly NK/neutrophil interactions of patients’ NK cells, while interactions with B-cells, T-cells, or monocytes were negligible. T-cells also displayed profound local and systemic alterations. Cellular interaction mapping revealed that next to NK/neutrophil interactions, interactions between B-cells with monocytes and T-cells with neutrophils characterize joint inflammation. Conclusion: This paired high-dimensional analysis revealed systemic and local alterations in NK cell subsets shaped by co-existing developmental stages and immune regulatory mechanisms. Cellular interaction mapping indicated that neutrophils are a main interaction-partner of NK cells in inflamed joints.
- Open Access
- 27 November 2023
- Published by Springer Nature in Nonlinear Dynamics
- Vol. 112 (2) , 1023-1038
Safe and compliant control of redundant robots using superimposition of passive task-space controllers
Animals are capable of robust and reliable control in unstructured environments, where they effortlessly overcome the uncertainty of interaction and are capable of exploiting singularities. These conditions are a well-known challenge for robots due to the limitations of projected dynamics, which requires accurate modelling and is susceptible to singularities. This work proposes a compliant passive control method for redundant manipulators based on a superimposition of multiple passive task-space controllers in a hierarchy without requiring any knowledge of the robot dynamics. The proposed control framework of passive controllers is inherently stable, numerically well-conditioned (as no matrix inversions are required), and computationally inexpensive (as no optimisation is used). This method leverages and introduces a novel stiffness profile for a recently proposed passive controller with smooth transitions between the divergence and convergence phases making it particularly suitable when multiple passive controllers are combined through superimposition. The experimental results demonstrate that the proposed method achieves sub-centimetre tracking performance during demanding dynamic tasks with fast-changing references, while remaining safe to interact with and robust to singularities. The data further show that the robot can fully take advantage of the redundancy to maintain the primary task accuracy while compensating for unknown environmental interactions, which is not possible from current frameworks that require accurate contact information.
- 27 November 2023
- Published by Springer Nature in Nonlinear Dynamics
- Vol. 112 (2) , 1023-1038
Open AccessSafe and compliant control of redundant robots using superimposition of passive task-space controllers
Animals are capable of robust and reliable control in unstructured environments, where they effortlessly overcome the uncertainty of interaction and are capable of exploiting singularities. These conditions are a well-known challenge for robots due to the limitations of projected dynamics, which requires accurate modelling and is susceptible to singularities. This work proposes a compliant passive control method for redundant manipulators based on a superimposition of multiple passive task-space controllers in a hierarchy without requiring any knowledge of the robot dynamics. The proposed control framework of passive controllers is inherently stable, numerically well-conditioned (as no matrix inversions are required), and computationally inexpensive (as no optimisation is used). This method leverages and introduces a novel stiffness profile for a recently proposed passive controller with smooth transitions between the divergence and convergence phases making it particularly suitable when multiple passive controllers are combined through superimposition. The experimental results demonstrate that the proposed method achieves sub-centimetre tracking performance during demanding dynamic tasks with fast-changing references, while remaining safe to interact with and robust to singularities. The data further show that the robot can fully take advantage of the redundancy to maintain the primary task accuracy while compensating for unknown environmental interactions, which is not possible from current frameworks that require accurate contact information.
- 31 July 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Robotics
- Vol. 39 (5) , 3805-3820
VAE-Loco: Versatile Quadruped Locomotion by Learning a Disentangled Gait Representation
Quadruped locomotion is rapidly maturing to a degree where robots are able to realize highly dynamic maneuvers. However, current planners are unable to vary key gait parameters of the in-swing feet midair . In this article, we address this limitation and show that it is pivotal in increasing controller robustness by learning a latent space capturing the key stance phases constituting a particular gait. This is achieved via a generative model trained on a single trot style, which encourages disentanglement such that application of a drive signal to a single dimension of the latent state induces holistic plans synthesizing a continuous variety of trot styles. We demonstrate that specific properties of the drive signal map directly to gait parameters, such as cadence, footstep height, and full-stance duration. Due to the nature of our approach, these synthesized gaits are continuously variable online during robot operation. The use of a generative model facilitates the detection and mitigation of disturbances to provide a versatile and robust planning framework. We evaluate our approach on two versions of the real ANYmal quadruped robots and demonstrate that our method achieves a continuous blend of dynamic trot styles while being robust and reactive to external perturbations.
- 31 July 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Robotics
- Vol. 39 (5) , 3805-3820
VAE-Loco: Versatile Quadruped Locomotion by Learning a Disentangled Gait Representation
Quadruped locomotion is rapidly maturing to a degree where robots are able to realize highly dynamic maneuvers. However, current planners are unable to vary key gait parameters of the in-swing feet midair . In this article, we address this limitation and show that it is pivotal in increasing controller robustness by learning a latent space capturing the key stance phases constituting a particular gait. This is achieved via a generative model trained on a single trot style, which encourages disentanglement such that application of a drive signal to a single dimension of the latent state induces holistic plans synthesizing a continuous variety of trot styles. We demonstrate that specific properties of the drive signal map directly to gait parameters, such as cadence, footstep height, and full-stance duration. Due to the nature of our approach, these synthesized gaits are continuously variable online during robot operation. The use of a generative model facilitates the detection and mitigation of disturbances to provide a versatile and robust planning framework. We evaluate our approach on two versions of the real ANYmal quadruped robots and demonstrate that our method achieves a continuous blend of dynamic trot styles while being robust and reactive to external perturbations.
- Open Access
VAE-Loco: Versatile Quadruped Locomotion by Learning a Disentangled Gait Representation
Quadruped locomotion is rapidly maturing to a degree where robots are able to realise highly dynamic manoeuvres. However, current planners are unable to vary key gait parameters of the in-swing feet midair. In this work we address this limitation and show that it is pivotal in increasing controller robustness by learning a latent space capturing the key stance phases constituting a particular gait. This is achieved via a generative model trained on a single trot style, which encourages disentanglement such that application of a drive signal to a single dimension of the latent state induces holistic plans synthesising a continuous variety of trot styles. We demonstrate that specific properties of the drive signal map directly to gait parameters such as cadence, footstep height and full stance duration. Due to the nature of our approach these synthesised gaits are continuously variable online during robot operation. The use of a generative model facilitates the detection and mitigation of disturbances to provide a versatile and robust planning framework. We evaluate our approach on two versions of the real ANYmal quadruped robots and demonstrate that our method achieves a continuous blend of dynamic trot styles whilst being robust and reactive to external perturbations.
Open AccessVAE-Loco: Versatile Quadruped Locomotion by Learning a Disentangled Gait Representation
Quadruped locomotion is rapidly maturing to a degree where robots are able to realise highly dynamic manoeuvres. However, current planners are unable to vary key gait parameters of the in-swing feet midair. In this work we address this limitation and show that it is pivotal in increasing controller robustness by learning a latent space capturing the key stance phases constituting a particular gait. This is achieved via a generative model trained on a single trot style, which encourages disentanglement such that application of a drive signal to a single dimension of the latent state induces holistic plans synthesising a continuous variety of trot styles. We demonstrate that specific properties of the drive signal map directly to gait parameters such as cadence, footstep height and full stance duration. Due to the nature of our approach these synthesised gaits are continuously variable online during robot operation. The use of a generative model facilitates the detection and mitigation of disturbances to provide a versatile and robust planning framework. We evaluate our approach on two versions of the real ANYmal quadruped robots and demonstrate that our method achieves a continuous blend of dynamic trot styles whilst being robust and reactive to external perturbations.
- Open Access
- 15 May 2023
- Published by Elsevier in Robotics and Autonomous Systems
- Vol. 166
Motion planning in dynamic environments using context-aware human trajectory prediction
- 15 May 2023
- Published by Elsevier in Robotics and Autonomous Systems
- Vol. 166
Open AccessMotion planning in dynamic environments using context-aware human trajectory prediction
- Open Access
Roll-Drop: accounting for observation noise with a single parameter
This paper proposes a simple strategy for sim-to-real in Deep-Reinforcement Learning (DRL) -- called Roll-Drop -- that uses dropout during simulation to account for observation noise during deployment without explicitly modelling its distribution for each state. DRL is a promising approach to control robots for highly dynamic and feedback-based manoeuvres, and accurate simulators are crucial to providing cheap and abundant data to learn the desired behaviour. Nevertheless, the simulated data are noiseless and generally show a distributional shift that challenges the deployment on real machines where sensor readings are affected by noise. The standard solution is modelling the latter and injecting it during training; while this requires a thorough system identification, Roll-Drop enhances the robustness to sensor noise by tuning only a single parameter. We demonstrate an 80% success rate when up to 25% noise is injected in the observations, with twice higher robustness than the baselines. We deploy the controller trained in simulation on a Unitree A1 platform and assess this improved robustness on the physical system.
Open AccessRoll-Drop: accounting for observation noise with a single parameter
This paper proposes a simple strategy for sim-to-real in Deep-Reinforcement Learning (DRL) -- called Roll-Drop -- that uses dropout during simulation to account for observation noise during deployment without explicitly modelling its distribution for each state. DRL is a promising approach to control robots for highly dynamic and feedback-based manoeuvres, and accurate simulators are crucial to providing cheap and abundant data to learn the desired behaviour. Nevertheless, the simulated data are noiseless and generally show a distributional shift that challenges the deployment on real machines where sensor readings are affected by noise. The standard solution is modelling the latter and injecting it during training; while this requires a thorough system identification, Roll-Drop enhances the robustness to sensor noise by tuning only a single parameter. We demonstrate an 80% success rate when up to 25% noise is injected in the observations, with twice higher robustness than the baselines. We deploy the controller trained in simulation on a Unitree A1 platform and assess this improved robustness on the physical system.
- Open Access
Learning and Deploying Robust Locomotion Policies with Minimal Dynamics Randomization
Training deep reinforcement learning (DRL) locomotion policies often require massive amounts of data to converge to the desired behaviour. In this regard, simulators provide a cheap and abundant source. For successful sim-to-real transfer, exhaustively engineered approaches such as system identification, dynamics randomization, and domain adaptation are generally employed. As an alternative, we investigate a simple strategy of random force injection (RFI) to perturb system dynamics during training. We show that the application of random forces enables us to emulate dynamics randomization. This allows us to obtain locomotion policies that are robust to variations in system dynamics. We further extend RFI, referred to as extended random force injection (ERFI), by introducing an episodic actuation offset. We demonstrate that ERFI provides additional robustness for variations in system mass offering on average a 53% improved performance over RFI. We also show that ERFI is sufficient to perform a successful sim-to-real transfer on two different quadrupedal platforms, ANYmal C and Unitree A1, even for perceptive locomotion over uneven terrain in outdoor environments.
Open AccessLearning and Deploying Robust Locomotion Policies with Minimal Dynamics Randomization
Training deep reinforcement learning (DRL) locomotion policies often require massive amounts of data to converge to the desired behaviour. In this regard, simulators provide a cheap and abundant source. For successful sim-to-real transfer, exhaustively engineered approaches such as system identification, dynamics randomization, and domain adaptation are generally employed. As an alternative, we investigate a simple strategy of random force injection (RFI) to perturb system dynamics during training. We show that the application of random forces enables us to emulate dynamics randomization. This allows us to obtain locomotion policies that are robust to variations in system dynamics. We further extend RFI, referred to as extended random force injection (ERFI), by introducing an episodic actuation offset. We demonstrate that ERFI provides additional robustness for variations in system mass offering on average a 53% improved performance over RFI. We also show that ERFI is sufficient to perform a successful sim-to-real transfer on two different quadrupedal platforms, ANYmal C and Unitree A1, even for perceptive locomotion over uneven terrain in outdoor environments.
- Top Cited PapersOpen Access
- 21 September 2022
- Published by Springer Nature in Autonomous Robots
- Vol. 46 (8) , 985-1005
A feasibility-driven approach to control-limited DDP
Differential dynamic programming (DDP) is a direct single shooting method for trajectory optimization. Its efficiency derives from the exploitation of temporal structure (inherent to optimal control problems) and explicit roll-out/integration of the system dynamics. However, it suffers from numerical instability and, when compared to direct multiple shooting methods, it has limited initialization options (allows initialization of controls, but not of states) and lacks proper handling of control constraints. In this work, we tackle these issues with a feasibility-driven approach that regulates the dynamic feasibility during the numerical optimization and ensures control limits. Our feasibility search emulates the numerical resolution of a direct multiple shooting problem with only dynamics constraints. We show that our approach (named Box-FDDP) has better numerical convergence than Box-DDP $$^+$$ (a single shooting method), and that its convergence rate and runtime performance are competitive with state-of-the-art direct transcription formulations solved using the interior point and active set algorithms available in Knitro. We further show that Box-FDDP decreases the dynamic feasibility error monotonically—as in state-of-the-art nonlinear programming algorithms. We demonstrate the benefits of our approach by generating complex and athletic motions for quadruped and humanoid robots. Finally, we highlight that Box-FDDP is suitable for model predictive control in legged robots.- 21 September 2022
- Published by Springer Nature in Autonomous Robots
- Vol. 46 (8) , 985-1005
Top Cited PapersOpen AccessA feasibility-driven approach to control-limited DDP
Differential dynamic programming (DDP) is a direct single shooting method for trajectory optimization. Its efficiency derives from the exploitation of temporal structure (inherent to optimal control problems) and explicit roll-out/integration of the system dynamics. However, it suffers from numerical instability and, when compared to direct multiple shooting methods, it has limited initialization options (allows initialization of controls, but not of states) and lacks proper handling of control constraints. In this work, we tackle these issues with a feasibility-driven approach that regulates the dynamic feasibility during the numerical optimization and ensures control limits. Our feasibility search emulates the numerical resolution of a direct multiple shooting problem with only dynamics constraints. We show that our approach (named Box-FDDP) has better numerical convergence than Box-DDP $$^+$$ (a single shooting method), and that its convergence rate and runtime performance are competitive with state-of-the-art direct transcription formulations solved using the interior point and active set algorithms available in Knitro. We further show that Box-FDDP decreases the dynamic feasibility error monotonically—as in state-of-the-art nonlinear programming algorithms. We demonstrate the benefits of our approach by generating complex and athletic motions for quadruped and humanoid robots. Finally, we highlight that Box-FDDP is suitable for model predictive control in legged robots. - Open Access
A Feasibility-Driven Approach to Control-Limited DDP
Differential dynamic programming (DDP) is a direct single shooting method for trajectory optimization. Its efficiency derives from the exploitation of temporal structure (inherent to optimal control problems) and explicit roll-out/integration of the system dynamics. However, it suffers from numerical instability and, when compared to direct multiple shooting methods, it has limited initialization options (allows initialization of controls, but not of states) and lacks proper handling of control constraints. In this work, we tackle these issues with a feasibility-driven approach that regulates the dynamic feasibility during the numerical optimization and ensures control limits. Our feasibility search emulates the numerical resolution of a direct multiple shooting problem with only dynamics constraints. We show that our approach (named BOX-FDDP) has better numerical convergence than BOX-DDP+ (a single shooting method), and that its convergence rate and runtime performance are competitive with state-of-the-art direct transcription formulations solved using the interior point and active set algorithms available in KNITRO. We further show that BOX-FDDP decreases the dynamic feasibility error monotonically--as in state-of-the-art nonlinear programming algorithms. We demonstrate the benefits of our approach by generating complex and athletic motions for quadruped and humanoid robots. Finally, we highlight that BOX-FDDP is suitable for model predictive control in legged robots.
Open AccessA Feasibility-Driven Approach to Control-Limited DDP
Differential dynamic programming (DDP) is a direct single shooting method for trajectory optimization. Its efficiency derives from the exploitation of temporal structure (inherent to optimal control problems) and explicit roll-out/integration of the system dynamics. However, it suffers from numerical instability and, when compared to direct multiple shooting methods, it has limited initialization options (allows initialization of controls, but not of states) and lacks proper handling of control constraints. In this work, we tackle these issues with a feasibility-driven approach that regulates the dynamic feasibility during the numerical optimization and ensures control limits. Our feasibility search emulates the numerical resolution of a direct multiple shooting problem with only dynamics constraints. We show that our approach (named BOX-FDDP) has better numerical convergence than BOX-DDP+ (a single shooting method), and that its convergence rate and runtime performance are competitive with state-of-the-art direct transcription formulations solved using the interior point and active set algorithms available in KNITRO. We further show that BOX-FDDP decreases the dynamic feasibility error monotonically--as in state-of-the-art nonlinear programming algorithms. We demonstrate the benefits of our approach by generating complex and athletic motions for quadruped and humanoid robots. Finally, we highlight that BOX-FDDP is suitable for model predictive control in legged robots.