Objective
Develop and analyse a combined technique implementing Pseudospectral Collocation and Multiple-Shooting Sequential Convex Programming to solve challenging Optimal Control Problems. The study is carried forward in collaboration with UC Berkeley’s Autonomous Controls Lab.
Approach
The objective is achieved by integrating ARGOS lab’s expertise in Pseudospectral Optimal Control with the existing ACL’s trajopt package, implementing the AutoSCVx algorithm for Multiple-Shooting Sequential Convex Programming. The developed approach is applied to challenging problems in quadrotor guidance and low-thrust trajectory optimization.
Why It Matters
Solving problems using Multiple-Shooting techniques, even when combined with the robust numerical properties of convex optimization, can be challenging in dynamically unstable systems and in scenarios where adequate initial guesses cannot be generated. Pseudospectral Collocation provides inherently smooth trajectories and allows for post-solution optimality assessments, enhancing the overall capability of the tool.