Overall Goal
Development of a high performance humanoid robot: Integrate suitable technologies to develop a humanoid robot capable of walking inside human oriented infrastructures, manipulating human tools and interfaces
Objectives and Technologies to Advance
Soft robot design for a robust and powerful humanoid
To explore, identify and implement the complex and adaptive interactions and relationships among the robot physical body, control system, motion generation and interaction and combine and tune physical capacities in the robot design (including passive and active compliance control of the robot body) with a novel and highly integrated model-based control approach considering in detail the overall body motion, the control and interaction dynamics during locomotion and manipulation.
Agile whole body locomotion
To develop reactive and versatile leg skills to permit humanoids to walk and balance against external disturbances. To advance from the slowly adaptive pre-planned bipedal gait generators and balancing planners towards more rapidly modulated and planning-free cyclic pattern generators combined with reflexive behaviours that will allow the robot to cope with uneven terrains and rapid start and stop gait transitions.
Robust and powerful manipulation
To develop hands that can perform powerful grasping and manipulation of human designed tools and interfaces, exhibiting mechanical robustness and power performance while maintaining significant capabilities to adapt to features of the environment. To integrate intrinsic flexibility in the hand design and combine it with controlled reactive impedance regulation during task execution to improve grasping stability and manipulation tolerance to object and interaction uncertainties. To exploit the weight and dynamics of the whole-body, using not only the body extremities (hands and feet) but also the inner parts of the body, to empower stronger and more robust manipulation of the environment.
Environment perception and whole-body affordances
To develop methods for multi-modal environment perception and active exploration combining the robot’s haptic sensing capacity during interactions and contacts, multiple inertial measurement sensors and exteroceptive vision data, to create 3D model representations of the environment and to associate affordance to environmental elements. Environment models, their affordances and the task constraints build the basis for the generation of whole-body locomotion and manipulation as they provide the required input for motion planning and action execution
Motion planning and control
To develop efficient algorithms for anytime planning and control of loco-manipulation, exploiting available primitives for locomotion, grasping, and manipulation when possible, and designing ad-hoc motions (or new primitives) as needed. The planning approach will ensure safety of the robot, by disallowing motions from which the robot cannot safely recover using, e.g., pre-defined “emergency” primitives, or by “rolling back” to previously known safe conditions.
Validation using realistic scenarios
To integrate results of objectives from 1 to 5 and validate the robot locomotion and manipulation capabilities in real task scenarios. To ensure concreteness and realism of our results in challenging, independently proposed scenarios. To ensure relevance to real world application, and ultimately readiness in case of need for use by Civil Defence Corps.