Model Predictive Control (also known as Receding Horizon Control (RHC)) has been highly successful in process control applications. Its use for aerospace applications has been hindered by its high computational requirements. In the present paper, we propose using enhanced primal-dual interior-point optimization techniques in the convex-quadratic-program-based RHC control of a rotorcraft. Our enhancements include a previously proposed “constraint-reduction” scheme that takes advantage of the very large number of inequality constraints (compared to the number of variables), and the ensuing redundancy of a large majority of these constraints. Other enhancements include the use of a penalty function, with automatic adaptation of the penalty parameter (also previously analyzed in the context of constraint-reduction), allowing for the use of “infeasible” “warm starts”, and of the partition of the constraints into “hard” to be imperatively satisfied, and “soft” whose violations can possibly be traded-off, with appropriate supporting algorithm. The heart of the paper is the application of all these techniques to an aggressive-trajectory-following problem for a model of a utility-class helicopter. The results are encouraging, and demonstrate that RHC control of rotorcraft should soon become a mature technology
Citation
Working paper, University of Maryland, College Park, MD 20742