A method for real-time dynamic fleet mission planning for autonomous mining

Abstract

This paper introduces a method for dynamic fleet mission planning for autonomous mining (in loop-free maps), in which a dynamic fleet mission is defined as a sequence of static fleet missions, each generated using a modified genetic algorithm. For the case of static fleet mission planning (where each vehicle completes just one mission), the proposed method is able to reliably generate, within a short optimization time, feasible fleet missions with short total duration and as few stops as possible. For the dynamic case, in simulations involving a realistic mine map, the proposed method is able to generate efficient dynamic plans such that the number of completed missions per vehicle is only slightly reduced as the number of vehicles is increased, demonstrating the favorable scaling properties of the method as well as its applicability in real-world cases.

A Two-Stage MIQP-Based Optimization Approach for Coordinating Automated Electric Vehicles in Confined Sites

“…Dynamic fleet planning method for autonomous mining is proposed in [27]. The method focuses on solving conflicts efficiently while attempting to minimize delays and waiting times by using a modified genetic algorithm.…”

“…With the obtained integer solution, the state and control trajectories are calculated by solving the "fixed-order coordination" NLP, i.e., Problem 2 with fixed crossing orders O I , O MS , O CS . 1) Crossing Order Heuristic: The MIQP that is assembled for obtaining the crossing order is formed as a quadratic approximation of (27). The way we form the quadratic approximation is similar to how the QP sub-problems are formed in Sequential Quadratic Programming (SQP) methods [38].…”

“…The way we form the quadratic approximation is similar to how the QP sub-problems are formed in Sequential Quadratic Programming (SQP) methods [38]. In essence, we can reformulate (27) as:…”