Some elements of model engineering for optimal control

Abstract¶

This survey presentation will cover three topics relevant to motion control and state observation: (1) Model reduction: how we can compute forward and reverse mappings between whole-body and reduced dynamic models, carrying both the equations of motion and system constraints. (2) Divergent components of motion: how the analysis of nonlinear systems can turn predictive control into linear feedback of virtual states, and how we can leverage this for hierarchical behaviors. (3) Trajectory optimization: a major item in the control toolbox, whose current limitations include feasibility, recursive feasibility, and real-time performance. We review new features brought by an upcoming generation of quadratic-programming and nonlinear solvers, including differentiable QP layers and constrained trajectory optimization.

Content¶

Slides

Trajectory optimization¶

	Aligator
	PROXDDP: Proximal Constrained Trajectory Optimization

Differentiable quadratic programming¶

	ProxSuite (includes PROXQP and QPLayer)
	PROXQP: an Efficient and Versatile Quadratic Programming Solver for Real-Time Robotics Applications and Beyond
	QPLayer: efficient differentiation of convex quadratic optimization

Bio¶

Stéphane is a research scientist at Inria. He received his M.Sc. in computer science from the École Normale Supérieure in 2012, and his Ph.D. in mechano-informatics from the University of Tokyo in 2016. Stéphane has worked at CNRS as tenured researcher and at ANYbotics AG as locomotion team lead before joining Inria where he is currently (having a blast) bridging motion control and computer vision. Stéphane is a proponent of open source robotics and contributes to projects like Upkie wheeled bipeds, robot_descriptions.py and qpbenchmark.