D6.1 - Resilient navigation and adaptive control of underactuated robots:

The Lancaster team has developed a system that can be paired with a drone to accurately locate and map a contaminated area when large external disturbances are experienced, for instance during bad weather.

The system uses a cascaded two-stage modified fast simultaneous localisation and mapping system (SLAM). The technology was developed for resilient and autonomous navigation by a single drone in an unknown and environment where GPS does not work.

This new navigation system allows the robot to operate reliably over a long period of time when it is sent to a sensory degraded environment, such as a highly contaminated radioactive environment. It has been evaluated and tested in a realistic simulation environment using the ROS platform (TRL 2), and implemented on an open-architecture drone system (designed and built fully at Lancaster University) for laboratory testing (TRL 3). Preliminary work to extend the results to cooperative robots is ongoing with the support of the National Nuclear Lab.

System key points include:

• High-performance manipulator positioning controllers based on data-driven, stochastic state-dependent parameter models.
• Addresses uncertainty arising from sensor degradation, material inconsistencies, device nonlinearities, etc.
• Illustrative case study – dual manipulator, semi-autonomous pipe cutting with reciprocating saw.
• Suite of widely applicable algorithms for adaptive control, inverse kinematics, planning and robot parameter estimation.
• Additional illustrative case study
  • identification and control of aerial vehicles with unknown inertia parameters.