Dynamic Zoning of Industrial Environments with Autonomous Mobile Robots

Russell Keith,Hung La

2024-11-12

Abstract:This paper presents a scheduling algorithm that divides a manufacturing/warehouse floor into zones that an Autonomous Mobile Robot (AMR) will occupy and complete part pick-up and drop-off tasks. Each zone is balanced so that each AMR will share each task equally. These zones change over time to accommodate fluctuations in production and to avoid overloading an AMR with tasks. A decentralized dynamic zoning (DDZ) algorithm is introduced to find the optimal zone design, eliminating the possibility of single-point failure from a centralized unit. Then a simulation is built comparing the adaptability of DDZ and other dynamic zoning algorithms from previous works. Initial results show that DDZ has a much lower throughput than other dynamic zoning algorithms but DDZ can achieve a better distribution of tasks. Initial results show that DDZ had a lower standard deviation of AMR total travel distance which was 2874.7 feet less than previous works. This 68.7\% decrease in standard deviation suggests that AMRs under DDZ travel a similar distance during production. This could be useful for real-world applications by making it easier to design charging and maintenance schedules without much downtime. Video demonstration of the system working can be seen here: \url{<a class="link-external link-https" href="https://youtu.be/yVi026oVD7U" rel="external noopener nofollow">this https URL</a>}

Robotics

What problem does this paper attempt to address?

The main problem that this paper attempts to solve is: in manufacturing and warehousing environments, how to effectively schedule autonomous mobile robots (AMR) through dynamic partitioning algorithms to ensure balanced task allocation and adapt to fluctuations in production.

Specifically, the paper focuses on the following aspects:

1. **Balanced task allocation**:
- Traditional methods may not be able to evenly distribute tasks among all AMRs, resulting in some AMRs being overloaded while others are idle. The paper introduces a decentralized dynamic zoning (DDZ) algorithm, aiming to balance the task volume of each AMR by adjusting the partitions.

2. **Adapting to production fluctuations**:
- Production requirements and task allocations in manufacturing environments are dynamically changing, and temporary delays may occur due to tool breakage, part shortages, or personnel deployment. The DDZ algorithm can re - partition areas according to current data to adapt to these changes and avoid an AMR being overwhelmed by too many tasks.

3. **Avoiding single - point failures**:
- Centralized control is prone to the single - point failure problem, that is, if the central controller has a problem, the entire system will not be able to operate normally. The DDZ algorithm adopts a decentralized approach, with each AMR operating independently. Even if an AMR fails, other AMRs can still continue to work, improving the robustness and flexibility of the system.

4. **Optimizing paths and reducing the standard deviation of travel distances**:
- Through simulation experiments, the paper found that although the DDZ algorithm is inferior to other dynamic partitioning algorithms (such as the genetic algorithm GA and simulated annealing SA) in terms of throughput, it can significantly reduce the standard deviation of the total travel distance of AMRs. This indicates that when using the DDZ algorithm, the travel distances between AMRs are more consistent, which helps to design more reasonable charging and maintenance plans and reduce downtime.

In summary, this paper is mainly devoted to developing a decentralized dynamic partitioning algorithm that can effectively schedule AMRs in manufacturing and warehousing environments to achieve balanced task allocation, adapt to production fluctuations, improve system robustness, and optimize path planning.