NaviFormer: A Deep Reinforcement Learning Transformer-like Model to Holistically Solve the Navigation Problem

TL;DR

NaviFormer is a Transformer-based deep reinforcement learning model that simultaneously addresses high-level route planning and low-level trajectory prediction for efficient, real-time navigation.

Contribution

It introduces a holistic approach combining route and path planning in a single model, improving accuracy and speed over existing methods.

Findings

NaviFormer achieves competitive accuracy in navigation tasks.

It demonstrates superior computation speed suitable for real-time applications.

Experiments show effective understanding of subproblem constraints.

Abstract

Path planning is usually solved by addressing either the (high-level) route planning problem (waypoint sequencing to achieve the final goal) or the (low-level) path planning problem (trajectory prediction between two waypoints avoiding collisions). However, real-world problems usually require simultaneous solutions to the route and path planning subproblems with a holistic and efficient approach. In this paper, we introduce NaviFormer, a deep reinforcement learning model based on a Transformer architecture that solves the global navigation problem by predicting both high-level routes and low-level trajectories. To evaluate NaviFormer, several experiments have been conducted, including comparisons with other algorithms. Results show competitive accuracy from NaviFormer since it can understand the constraints and difficulties of each subproblem and act consequently to improve performance. Moreover, its superior computation speed proves its suitability for real-time missions.