The article presents a new neural network structure called PPNet, designed for end-to-end near-optimal path planning. Traditional path planners, such as sampling-based ones, often have slow convergence to the optimal solution and are sensitive to the initial solution. The PPNet overcomes these challenges by dividing the path planning problem into two subproblems: path space segmentation and waypoints generation in the given path space. The authors also introduce an efficient data generation method for path planning named EDaGe-PP, which significantly improves the computation time and success rate of PPNet. The study concludes that PPNet can find a near-optimal solution in a much shorter time than the state-of-the-art path planners.
Publication date: 19 Jan 2024
Project Page: Journal of Latex Class Files, Vol. 14, No. 8
Paper: https://arxiv.org/pdf/2401.09819
