Skip to main content

Work Package 1 – Components

Improving the reliability of subsystems to ensure mission success in resilient robot autonomy.

WP1 aims to develop reliable robotic system components that are tolerant to faults, self-aware, can provide confidence metrics about their performance, and are transparent to the wider robotic system. Developing these new components will enable safe decision-making and action by higher-level layers of autonomous robotic systems.

MODELLING AND CONTROL OF COMPLEX ROBOTICS SYSTEMS

New environments and tasks require more complex robot morphologies and systems, such as mobile manipulators and underwater multi-robot systems. The aim is to research holistic frameworks comprising modelling, control and estimation theory, and planning that can exploit the benefits of redundant robotic systems to allow robots to function even if failures happen, that are agnostic to the systems’ morphologies, and with formal guarantees to prove their reliability.

ROBUST SENSOR FUSION

Text

Sensors are used to estimate the environment’s state and the robot’s condition. Sensor information must be reliable for the robot to decide actions and inform operators and other systems. Fusing sensor data makes the system resilient to mis-readings, degradation or sensor failure and provides more precise estimates than each sensor alone. The aim is to research fusion and state estimation algorithms that account for failures explicitly and leverage sensor redundancy or heterogeneous sensor modalities to provide more robust and precise estimations.

RISK-AWARE MOTION PLANNING AND CONTROL

Robots moving in an environment involves risks due to the workspace, degradation or failure of the robot’s actuators. The aim is to study riskaware, chance-constrained motion planning and control techniques that dynamically integrate those risks, informed and estimated by other system components to develop safer robot motions. Developed techniques should be computationally efficient and provide formal guarantees, ensuring the theoretical properties required by higherlevel layers across the robot architecture.

HUMAN-UNDERSTANDABLE CONFIDENCE METRICS

Uncertainty is intrinsic to robotic systems. Moreover, sensors are noisy, actuators have mechanical imperfections, and the models used for planning and control are simplifications of reality. Confidence metrics provide means to communicate how sensor data or actions taken are so other systems or humans can make decisions and understand under which circumstances those measurements were taken. This research focuses on providing transparency and human-understandable information about the system’s state.

Updates

Research Spotlight – Dr Mohamed Atia

Mohamed’s research aims to develop a modelling framework for underwater vehicle-manipulator systems, accounting for complex nonlinear hydrodynamics effects. Computational fluid dynamics simulations have been done to analyse these nonlinear behaviours, helping develop accurate analytical models. Those analytical models under development are computationally efficient and can be computed online, which is crucial for model-based control frameworks. This technology is widely applicable to inspection, maintenance and monitoring of vessels and underwater structures related to maritime infrastructure or nuclear decommissioning.

Research Spotlight – Seyonne Leslie-Dalley

Seyonne’s research focuses on developing a control framework for the cooperative manipulation of two independent underwater vehicle-manipulator systems. The framework will enable the intuitive description of a general bimanual manipulation task while also considering not only the system’s intrinsic constraints and environmental constraints but also constraints specific to that manipulation task. With direct applications in underwater industries such as the inspection and maintenance of offshore structures, nuclear pond decommissioning, and marine exploration, the framework’s versatility extends beyond these environments offering potential use in any industry where cooperative manipulation could be used.

    Explore Work Packages

    WP 1 Components
    WP2 Architectures
    WP3 Interactions
    WP4 Assurance
    WP5 Demonstrators