HTP Graphics

RAIN PROGRESS // A method has been developed which includes the training of a neural network with a dataset from the RAIN simulated glovebox and obtaining the cost collision from the neural network. This cost of collision is used as constraint in the inverse kinematics solver which optimises the joint poses to place the end-effector to the goal pose to avoid collision with obstacles. This method has been tested in the simulated glovebox environment with different obstacles. However, the collision also should be detected in end-effector trajectory along with joints. Till now, this work has progressed to find collision cost for joints in the arm. The future avenue is to progress the work to include both end effector and joints pose trajectories. REMOTE HANDLING 53 FUTURE ASPIRATIONS // In future work multiple collision will be considered, this can be done through multiple naive approaches such as calculating the estimate for multiple obstacles and optimising against the minimum. Also, the work can progress towards motion planning to include both end effector and joints pose trajectories. The proposed algorithm should be tested in real-world scenario. For this purpose, collaboration can be done with industries which uses gloveboxes, including industries other than nuclear. This solution can be used in other industries than gloveboxes like kitchen robotics. Collaboration and demonstration in non-nuclear settings could provide useful findings to enhance nuclear-specific challenges.

• Cover