Optimizing for Transfers in a Multi-vehicle Collection and Delivery Problem

Abstract

We address the Collection and Delivery Problem (CDP) with multiple vehicles, such that each collects a set of items at different locations and delivers them to a dropoff point. The goal is to minimize either delivery time or the total distance traveled. We introduce an extension to the CDP: what if a vehicle can transfer items to another vehicle before making the final delivery? By dividing the labor among multiple vehicles, the delivery time and cost may be reduced. However, introducing transfers increases the number of feasible schedules exponentially. In this paper, we investigate this Collection and Delivery Problem with Transfers (CDP-T), discuss its theoretical underpinnings, and introduce a two-approximate polynomial time algorithm to minimize total distance travelled. Furthermore, we show that allowing transfers to take place at any location for the CDP-T results in at most a factor of two improvement. We demonstrate our approximation algorithms on large simulated problem instances. Finally, we deploy our algorithms on robots that transfer and deliver items autonomously in an office building.

Constraint Manifold Subsearch for multirobot path planning with cooperative tasks

“…Coltin and Veloso [10] and Kamio and Iba [16] have investigated problems where multiple robots independently deliver packages, exchanging packages only if the cost of delivery will be reduced or if the layout of the workspace necessitates it. However, the actual interactions between robots were brief, limited to the exchange of packages, and the planning did not account for multirobot collisions.…”

“…Handling the dynamic formation and dissolution of these robot teams requires solving two problems: scheduling and assigning robots to teams [10,11,18], and coordinating the motions of many robots, both in and out of teams. In this paper, the assignment of robots to tasks and the order in which tasks must be executed are assumed to be provided a priori, and the focus will be on finding high-quality, collision-free paths for large numbers of robots operating both individually and as members of teams.…”

Integrating Planning and Execution for a Team of Heterogeneous Robots with Time and Communication Constraints

“…Several architectures and paradigms have been proposed to deal with mission planning and execution in presence of communication limits. A common approach is to consider that each agent performs its own plan computation while some goals can be exchanged between agents within communication windows [4], [5], [6], [7]. In this approach, no global plan exists and the system is almost unpredictable.…”

Decentralized allocation of tasks with temporal and precedence constraints to a team of robots

“…Minimizing the distance traveled is common (e.g. [99,94,100]) but some instead minimize a time measure over robot paths (e.g. [56,95]).…”