PhD Position in Robust and Control-Aware Motion Planning

*Teams*: Rainbow (IRISA/Inria Rennes) and Robotics and Mechatronics group (Univ. Twente, Netherlands)

*Advisors of PhD*: Paolo Robuffo Giordano, Antonio Franchi, Quentin Delamare

*How to apply*: Interested candidates are requested to apply via this form:
The position will remain open until a satisfactory candidate is found

An effective way of dealing with the complexity of robots operating in real (uncertain) environments is the paradigm of “feedforward/feedback” or “planning/control”: in a first step a suitable nominal trajectory (feedforward) for the robot states/controls is planned exploiting the available information (e.g., a model of the robot and of the environment). This step is usually executed offline and can take into account presence of constraints (e.g., collision avoidance, limited actuation) and optimality w.r.t. metrics of interest (e.g., time, energy). An open-loop execution of this planned trajectory would, however, fail in most practical cases because of the unavoidable uncertainties affecting the information used for planning. Therefore, the planned trajectory is in practice robustly executed by making use of a motion controller that “closes the loop” (feedback) between planned and actual motion, providing robustness against all the effects that could not be considered at the planning stage. While there has been an effort in proposing “robust planners” or more “global controllers” (e.g., Model Predictive Control (MPC)), a truly unified approach that fully exploits the techniques of the motion planning and control/estimation communities is still missing and the existing state-of-the-art has several important limitations, namely (1) lack of generality, (2) lack of computational efficiency, and (3) poor robustness. All these shortcomings are a major limiting factor in the autonomy and decision-making capabilities of robots operating in all those complex scenarios (real-world conditions, non-negligible effects of the uncertainties, fast dynamics) which are instead the typical conditions in which future robots are expected to operate.

In this context, the goal of this PhD will be to contribute to the development of a general and unified “intrinsically-robust and control-aware motion planning framework” for addressing the above-mentioned issues and demonstrating the applicability of this new framework to real robots in real-world challenging tasks (see below for more details on the planned experimental validations).
This PhD is part of a 4-year research project recently funded by ANR (the french funding agency). The project is in cooperation with the RIS team at LAAS-CNRS and the Robotics and Mechatronics group at the Univ. of Twente, Netherlands.

*PhD Subject*
The work will start from our previous contributions to these topics (in particular see this paper and this paper) where we introduced the notion of “closed-loop state sensitivity” and showed how this quantity can be used as an effective “parametric uncertainty metric” for generating motion plans that are intrinsically robust to variations in the robot model parameters. The PhD student is then expected to give contributions along these points:
* development of additional uncertainty metrics derived from the state sensitivity introduced here, and coupling with more classical notions (e.g., state covariance matrix) for those uncertainties that cannot be accurately modeled in parametric form (e.g., complex sensing/actuation uncertainties)
coupling with observability metrics for addressing the problem of optimal state estimation together with that of intrinsically-robust motion planning
* efficient estimation of worst-case uncertainty tubes given a range of uncertainty (and a given controller)
optimization of the controller parameters (or even control structure) for further increasing the closed-loop robustness of the motion plans
use of modern efficient numerical optimization algorithms for producing locally optimal motion plans with the possibility of a fast online replanning
* use of modern multi-objective motion planning algorithms for including the various sensitivity/uncertainty metrics within a more general and global motion planning framework (this activity will be done in close cooperation with the RIS team at LAAS-CNRS, where a second PhD student will be hired to work on advancements on the motion planning side)
* experimental validation of the whole approach in (1) an indoor pick- and-place/assembly task involving a 7-dof torque-controlled and (2) an outdoor cooperative mobile manipulation task involving an aerial manipulator (a quadrotor UAV equipped with an onboard arm) and a skid-steering mobile robot with an onboard arm. These validations will be done in cooperation with the other partners of the project

* M.Sc. degree in computer science, robotics, engineering, applied mathematics (or related fields)
* Good experience in C/C++ , ROS
* Previous experience in some of these areas: nonlinear control, optimization, planning for robotics
* Familiarity with Matlab/Simulink is a plus
* Scientific curiosity, large autonomy and ability to work independently

The position is full-time for 3 years and will be paid according to the French salary regulations for PhD students.
The starting date is flexible and can be agreed with the successful candidate (but the PhD should start by November 2021 at the latest).

*How to apply*
Interested candidates are requested to apply via this form:
The position will remain open until a satisfactory candidate is found