Abstract:Deep learning has been extended to a number of new domains with critical success, though some traditional orienteering problems such as the Travelling Salesman Problem (TSP) and its variants are not commonly solved using such techniques. Deep neural networks (DNNs) are a potentially promising and under-explored solution to solve these problems due to their powerful function approximation abilities, and their fast feed-forward computation. In this paper, we outline a method for converting an orienteering problem into a classification problem, and design a customised multi-layer deep learning network to approximate traditional optimisation solutions to this problem. We test the performance of the network on a real-world parking violation dataset, and conduct a generic study that empirically shows the critical architectural components that affect network performance for this problem.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is how to use deep - learning techniques to approximately solve the Travelling Officer Problem (TOP), which is a time - dependent variant of the Travelling Salesman Problem. Specifically, the paper focuses on the problem of maximizing the number of illegally parked vehicles caught by parking attendants within a limited time. ### Problem Background Parking attendants need to patrol and issue tickets in an area, but illegally parked vehicles usually leave in a short time, and multiple illegal parking events may occur simultaneously. Therefore, attendants need to make optimal decisions in terms of time and space to catch as many illegal behaviors as possible. Although traditional optimization methods can generate good routes, they are not efficient enough in real - time applications and are difficult to meet the actual needs. ### Paper Objectives The objective of the paper is to approximately solve this problem by transforming TOP into a classification problem and designing a customized deep neural network (DNN). Specifically: 1. **Problem Transformation**: Transform TOP from an optimization problem into a classification problem in supervised learning. 2. **Data Extraction**: Extract features from spatio - temporal data and construct a training data set. 3. **Model Design**: Design a customized multi - layer deep - learning network so that it can approximate the solutions of traditional optimization algorithms. 4. **Performance Evaluation**: Verify the effectiveness of this method through experiments and compare it with traditional optimization methods and other machine - learning methods. ### Key Challenges - **Spatio - temporal Dependence**: TOP involves not only spatial information but also time information, which makes the problem more complex. - **Label Generation**: Since TOP is an optimization problem, there are no ready - made labels, and labels required for training need to be generated through optimization algorithms. - **Model Design**: How to effectively use deep neural networks to approximate the solutions of optimization algorithms, especially in combinatorial optimization problems. ### Solutions The paper proposes a general framework that combines optimization methods and deep neural networks to solve TOP. The specific steps include: 1. **Spatio - temporal Feature Extraction**: Extract spatio - temporal features from historical parking violation data. 2. **Optimization Search**: Use optimization algorithms to generate paths as training labels. 3. **Customized Neural Network**: Design a customized neural network architecture so that it can learn and approximate the solutions of optimization algorithms. Through these methods, the paper shows how to use deep - learning techniques to efficiently solve TOP in real - time scenarios, thereby improving the work efficiency of parking attendants.

Approximating Optimisation Solutions for Travelling Officer Problem with Customised Deep Learning Network

Solving Optimization Problems Through Fully Convolutional Networks: an Application to the Traveling Salesman Problem

Driving Behaviour Style Study with a Hybrid Deep Learning Framework Based on GPS Data

A Deep Reinforcement Learning Based Real-Time Solution Policy for the Traveling Salesman Problem

Solving Orienteering Problems by Hybridizing Evolutionary Algorithm and Deep Reinforcement Learning

A Scalable Deep Learning Approach for Solving High-Dimensional Dynamic Optimal Transport.

A Reinforcement Learning Approach to the Orienteering Problem with Time Windows

Spatial-Aware Deep Reinforcement Learning for the Traveling Officer Problem

Deep Learning for Data-Driven Districting-and-Routing

Large-Scale OD Matrix Estimation with A Deep Learning Method

Learning to Optimize Industry-Scale Dynamic Pickup and Delivery Problems

Effective Travel Time Estimation: when Historical Trajectories over Road Networks Matter.

Learning to Plan Maneuverable and Agile Flight Trajectory with Optimization Embedded Networks

Optimal Transshipment Route Planning Method Based on Deep Learning for Multimodal Transport Scenarios

A Deep Neural Network-Based Method for Solving Obstacle Problems

Learning 2-opt Heuristics for the Traveling Salesman Problem via Deep Reinforcement Learning

Hybridization of evolutionary algorithm and deep reinforcement learning for multi-objective orienteering optimization

Enhance the performance of navigation: A two-stage machine learning approach

Deep Learning Based Optimization in Wireless Network

Hierarchical Neural Constructive Solver for Real-world TSP Scenarios

How Good is Neural Combinatorial Optimization? A Systematic Evaluation on the Traveling Salesman Problem