Applications of Successive Convexification in Autonomous VehiclePlanning and Control

TL;DR

This paper adapts successive convexification techniques from aerospace to autonomous vehicle planning and control, demonstrating their effectiveness in speed planning and obstacle avoidance with detailed formulations and simulations.

Contribution

First systematic application of aerospace-derived successive convexification methods to autonomous vehicle planning and control problems.

Findings

Successfully formulated and solved speed planning and MPC problems with constraints

Demonstrated obstacle avoidance and evasion maneuvers in simulations

Provided detailed problem formulation and implementation insights

Abstract

In this paper, we present the application of successive convexification methods to autonomous driving problems borrowed from recent aerospace literature. We formulate two optimization problems within the successive convexification framework. Using arc-length parametrization in the vehicle kinematic model, we solve the speed planning and model predictive control problems with a range of constraints and obstacle configurations. This paper is the first systematic application of successive convexification methods from the aerospace literature to the autonomous driving problems. In addition, we show a simple application of logical state-trigger constraints in a continuous formulation of the optimization by including an evasion maneuver in the simulations section. We give details of the problem formulation and implementation and present and discuss the results.