MAP-Elites based Hyper-Heuristic for the Resource Constrained Project Scheduling Problem

Shelvin Chand,Kousik Rajesh,Rohitash Chandra

DOI: https://doi.org/10.48550/arXiv.2204.11162

2022-04-24

Abstract:The resource constrained project scheduling problem (RCPSP) is an NP-Hard combinatorial optimization problem. The objective of RCPSP is to schedule a set of activities without violating any activity precedence or resource constraints. In recent years researchers have moved away from complex solution methodologies, such as meta heuristics and exact mathematical approaches, towards more simple intuitive solutions like priority rules. This often involves using a genetic programming based hyper-heuristic (GPHH) to discover new priority rules which can be applied to new unseen cases. A common problem affecting GPHH is diversity in evolution which often leads to poor quality output. In this paper, we present a MAP-Elites based hyper-heuristic (MEHH) for the automated discovery of efficient priority rules for RCPSP. MAP-Elites uses a quality diversity based approach which explicitly maintains an archive of diverse solutions characterised along multiple feature dimensions. In order to demonstrate the benefits of our proposed hyper-heuristic, we compare the overall performance against a traditional GPHH and priority rules proposed by human experts. Our results indicate strong improvements in both diversity and performance. In particular we see major improvements for larger instances which have been under-studied in the existing literature.

Neural and Evolutionary Computing,Artificial Intelligence

What problem does this paper attempt to address?

The problem that this paper attempts to solve is the Resource - Constrained Project Scheduling Problem (RCPSP), which is an NP - hard combinatorial optimization problem. Specifically, the goal of RCPSP is to schedule a series of activities without violating activity priorities and resource constraints. As the scale of the project increases, it becomes extremely difficult to find the optimal solution.

### Problem Background

1. **Complexity of RCPSP**:
- RCPSP involves scheduling a series of activities under two constraints: the priority relationships between activities and resource constraints.
- Priority constraint: If activity \( j \) is a successor of activity \( i \), activity \( i \) must be completed before activity \( j \) can start.
- Resource constraint: Each project is allocated a certain amount of renewable resources \( R_k \). Each activity may require one or more of these resources and cannot exceed the available amount of resources during scheduling.

2. **Limitations of Traditional Methods**:
- Traditional exact methods (such as linear programming) can only handle small - scale instances.
- Meta - heuristic algorithms (such as genetic algorithms, particle swarm optimization, etc.) can handle larger - scale instances, but have high computational costs and are difficult to quickly generate scheduling schemes.
- Priority Rules, because of their simplicity, intuitiveness, and low computational cost, are more suitable for scheduling problems in dynamic environments, but it is not easy to design effective priority rules.

### Main Contributions of the Paper

To solve the above problems, the paper proposes a MAP - Elites - based Quality - Diversity Hyper - Heuristic (MEHH). This method aims to automatically generate efficient priority rules and overcomes the diversity and premature convergence problems existing in traditional GPHH (Genetic Programming Hyper - Heuristic).

### Main Features of MEHH

1. **Quality - Diversity**:
- MAP - Elites maintains the diversity of the population by explicitly maintaining a diverse set of solutions in the feature space, avoiding premature convergence.
- Compared with traditional elite selection methods, MAP - Elites can explore a larger search space and thus find better solutions.

2. **Feature Space Construction**:
- MEHH uses three features to describe solutions: the number of nodes, the number of resource - related nodes, and Slack.
- These features include not only genotypic features (such as the structure of the tree) but also phenotypic features (such as the Slack of the generated scheduling scheme).

3. **Experimental Verification**:
- The paper conducts comparative experiments with traditional GPHH and other priority rules designed by human experts. The results show that MEHH has significant improvements in both diversity and performance, especially in large - scale instances.

### Conclusion

By introducing the MAP - Elites framework, the paper successfully solves the diversity and premature convergence problems encountered by traditional GPHH in RCPSP, providing a new and effective method for automatically discovering efficient priority rules.