Marco Faroni

Marco Faroni

Robotics researcher

Research

Human-aware decision making for collaborative robotics

Poor robot planning hinders the effectiveness of human-robot collaboration. This includes conservative implementation of safety rules, planning of trajectories that interfere with human movements, and suboptimal design and execution of complex tasks. I am interested in developing all-around planning algorithms for effective and safe human-robot collaboration, spanning path planning and re-planning, trajectory control, and task-and-motion planning.

Sampling-based motion planning and re-planning

New challenges continously arise in motion planning, e.g., planning with complex dynamics or black-box model, manipulation planning with time-critical requirements or under uncertainty. I am interested in designing new sampling-based motion planning algorithms for challenging problems such as kinodynamic motion planning and high-dimensional problems.

MPC for path following of redundant manipulators

When a robot has more degrees of freedom than those strictly necessary to carry out a task, it is said to be kinematically redundant. The additional degrees of freedom allow for posture optimization (e.g. to avoid obstacles or improve dexterity). I’ve designed model-predictive-control methods to exploit the robot redundancy at best and achieve optimal path following, even when the desired tasks drive the robot outside its joint physical limits.

Real-time trajectory scaling for optimal path following

Robot often need to adapt the trajectory on the fly to prevent collisions or tracking errors. The physical limits of the robot joints do not allow for error-free trajectory scaling or stopping, which is why online trajectory scaling is often needed to reduce the robot path error. I’ve worked on the formulation of real-time (<1ms) trajectory scaling methods using model predictive control (MPC) and look-ahead techniques to reduce the robot path error in the case of redundant and non-redundant robots, and joint-space and Cartesian-space tasks.