基于深度强化学习的区域化视觉导航方法

doi:10.16183/j.cnki.jsjtu.2019.277

上海交通大学学报 ›› 2021, Vol. 55 ›› Issue (5): 575-585.doi: 10.16183/j.cnki.jsjtu.2019.277

所属专题：《上海交通大学学报》2021年12期专题汇总专辑；《上海交通大学学报》2021年“自动化技术、计算机技术”专题

基于深度强化学习的区域化视觉导航方法

李鹏, 阮晓钢, 朱晓庆(), 柴洁, 任顶奇, 刘鹏飞

北京工业大学信息学部,北京 100124

收稿日期:2019-09-26 出版日期:2021-05-28 发布日期:2021-06-01
通讯作者: 朱晓庆 E-mail:alex.zhuxq@bjut.edu.cn
作者简介:李鹏(1992-),男,河北省廊坊市人,博士生,主要研究方向为机器人导航.
基金资助:
国家自然科学基金(61773027);北京市自然科学基金(4202005)

A Regionalization Vision Navigation Method Based on Deep Reinforcement Learning

LI Peng, RUAN Xiaogang, ZHU Xiaoqing(), CHAI Jie, REN Dingqi, LIU Pengfei

Faculty of Information Technology, Beijing University of Technology, Beijing 100124, China

Received:2019-09-26 Online:2021-05-28 Published:2021-06-01
Contact: ZHU Xiaoqing E-mail:alex.zhuxq@bjut.edu.cn

摘要/Abstract

摘要：

针对移动机器人在分布式环境中的导航问题,提出一种基于深度强化学习的区域化视觉导航方法.首先,根据分布式环境特征,在不同区域内独立学习控制策略,同时构建区域化模型, 实现导航过程中控制策略的切换和结合.然后,为使机器人具有更好的目标导向行为,在区域导航子模块中增加奖励预测任务,并结合经验池回放奖励序列.最后,在原有探索策略的基础上添加景深约束,防止因碰撞导致的遍历停滞.结果表明: 奖励预测和景深避障的应用有助于提升导航性能.在多区域环境测试过程中,区域化模型在训练时间和所获奖励上展现出单一模型不具备的优势,表明其能更好地应对大范围导航.此外,实验在第一人称视角的3D环境下进行,状态是部分可观察的,利于实际应用.

关键词: 深度强化学习, 分布式环境, 区域化模型, 奖励预测, 景深避障

Abstract:

Aimed at the problems of navigation in distributed environment of a mobile robot, a regionalization vision navigation method based on deep reinforcement learning is proposed. First, considering the characteristics of distributed environment, the independent submodule learning control strategy is used in different regions and the regionalization model is built to switch and combine navigation control strategies. Then, in order to make the robot have a better goal-oriented behavior, reward prediction task is integrated into the submodule, and reward sequence is played back in combination with the experience pool. Finally, depth limitation is added to the primitive exploration strategy to prevent the traversal stagnation caused by collision. The results show that the application of reward prediction and depth obstacle avoidance is helpful to improve navigation performance. In the process of multi-area environment test, the regionalization model shows the advantages that the single model does not have in terms of training time and rewards, indicating that it can better deal with large-scale navigation. In addition, the experiment is conducted in the first-person 3D environment, and the state is partially observable, which is conducive to practical application.

Key words: deep reinforcement learning, distributed environment, regionalization model, reward prediction, depth obstacle avoidance

中图分类号:

TP242

李鹏, 阮晓钢, 朱晓庆, 柴洁, 任顶奇, 刘鹏飞. 基于深度强化学习的区域化视觉导航方法[J]. 上海交通大学学报, 2021, 55(5): 575-585.

LI Peng, RUAN Xiaogang, ZHU Xiaoqing, CHAI Jie, REN Dingqi, LIU Pengfei. A Regionalization Vision Navigation Method Based on Deep Reinforcement Learning[J]. Journal of Shanghai Jiao Tong University, 2021, 55(5): 575-585.

图/表 15

图1

图2

图3

图4

图5

图6

图7

图8

表1

图9

图10

图11

图12

图13

图14

参考文献 25

[1]	徐德. 室内移动式服务机器人的感知、定位与控制[M]. 北京: 科学出版社, 2008.
	XU De. Perception, positioning and control of indoor mobile service robot[M]. Beijing: Science Press, 2008.
[2]	STEFFENACH H A, WITTER M, MOSER M B, et al. Spatial memory in the rat requires the dorsola-teral band of the entorhinal cortex[J]. Neuron, 2005, 45(2):301-313. doi: 10.1016/j.neuron.2004.12.044 URL
[3]	ARULKUMARAN K, DEISENROTH M, BRUNDAGE M, et al. A brief survey of deep reinforcement learning, (2017-09-28)[2019-08-20]. https://arxiv.org/pdf/1708.05866.pdf
[4]	MIROWSKI P, PASCANU R, VIOLA F, et al. Learning to navigate in complex environments,(2017-01-13)[2019-08-13]. https://arxiv.org/pdf/1611.03673.pdf
[5]	ZHU Y, MOTTAGHI P, KOLVE E, et al. Target-driven visual navigation in indoor scenes using deep reinforcement learning [C]//IEEE International Conference on Robotics and Automation. Singapore: IEEE, 2017, 3357-3364.
[6]	JADERBERG M, MNIH V, CZARNECKI W M, et al. Reinforcement learning with unsupervised auxiliary tasks[EB/OL]. (2016-11-16)[2019-04-20]. https://arxiv.org/pdf/1611.05397.pdf
[7]	OH J, CHOCKALINGAM V, SATINDER P, et al. Control of memory, active perception, and action in minecraft[EB/OL]. (2016-05-30)[2019-05-07]. https://arxiv.org/pdf/1605.09128.pdf
[8]	黄健, 严胜刚. 基于区域划分自适应粒子群优化的超短基线定位算法[J]. 控制与决策, 2019, 9(34):2023-2030.
	HUANG Jian, YAN Shenggang. Ultra-short baseline positioning algorithm based on region-division adaptive particle swarm optimization[J]. Control and Decision, 2019, 9(34):2023-2030.
[9]	张俊, 田慧敏. 一种基于边指针搜索及区域划分的三角剖分算法[J]. 自动化学报, 2021, 47(1):100-107.
	ZHANG Jun, TIAN Huimin. A triangulation algorithm based on edge-pointer search and region-division[J]. Acta Automatica Sinica, 2021, 47(1):100-107.
[10]	RUAN X G, REN D Q, ZHU X Q, et al. Optimized data association based on Gaussian mixture model[J]. IEEE Access, 2020, 8:2590-2598. doi: 10.1109/Access.6287639 URL
[11]	朱续涛, 何晓斌, 刘悦, 等. 一种简易的脑片图像的半自动区域划分及细胞计数方法[J]. 波谱学杂志, 2018, 2(35):133-140.
	ZHU Xutao, HE Xiaobin, LIU Yue, et al. A Convenient semi-automatic method for analyzing brain sections: Registration, segmentation and cell counting[J]. Chinese Journal of Magnetic Resonance, 2018, 2(35):133-140.
[12]	KULKARNI T D, SAEEDI A, GAUTAM S, et al. Deep successor reinforcement learning, (2016-06-08)[2019-05-21]. https://arxiv.org/pdf/1606.02396.pdf
[13]	MNIH V, KAVUKCUOGLU K, SILVER D, et al. Human-level control through deep reinforcement learning[J]. Nature, 2015, 518(7540):529-533. doi: 10.1038/nature14236 URL
[14]	SHIH H H, SHAO H C, PING T W, et al. Distributed deep reinforcement learning based indoor visual navigation [C]//2018 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS). Madrid, Spain: IEEE, 2018: 2531-2537.
[15]	TESSLER C, GIVONY S, ZAHAVY T, et al. A deep hierarchical approach to lifelong learning in minecraft, (2016-11-30)[2019-06-05]. https://arxiv.org/pdf/1604.07255.pdf
[16]	HAUSKNECHT M, STONE P. Deep recurrent Q-learning for partially observable MDPs[EB/OL].(2017-01-11)[2019-05-17]. https://arxiv.org/pdf/1507.06527.pdf
[17]	RIEDMILLER M. Neural fitted Q iteration—First experiences with a data efficient neural reinforcement learning method [C]//16th European Conference on Machine Learning. Berlin, Germany: Springer-Verlag, 2005: 279-292.
[18]	LANGE S, RIEDMILLER M, VOIGTLANDER A. Autonomous reinforcement learning on raw visual input data in a real world application [C]//The 2012 International Joint Conference on Neural Networks. Brisbane, Australia: IEEE, 2012: 1-8.
[19]	GU S X, LILLICRAP T, SUTSKEVER I, et al. Continuous Deep Q-learning with model-based acceleration [C]//33rd International Conference on Machine Learning. USA: International Machine Learning Society, 2016: 4135-4148.
[20]	KAELBLING L P, LITTMAN M L, CASSANDRA A R. Planning and acting in partially observable stochastic domains[J]. Artificial Intelligence, 1998, 101(1/2):99-134. doi: 10.1016/S0004-3702(98)00023-X URL
[21]	HOCHREITER S, SCHMIDHUBER J. Long short-term memory[J]. Neural Computation, 1997, 9(8):1735-1780. doi: 10.1162/neco.1997.9.8.1735 URL
[22]	BEATTIE C, LEIBO J Z, TEPLYASHIN D, et al. Deepmind lab, (2016-12-13)[2018-12-30]. https://arxiv.org/pdf/1612.03801.pdf
[23]	GERS F A, SCHMIDHUBER J, CUMMINS F. Learning to forget: Continual prediction with LSTM[J]. Neural Computation, 2000, 12(10):2451-2471. doi: 10.1162/089976600300015015 URL
[24]	MNIH V, BADIA A P, MIRZA M, et al. Asynchronous method for deep reinforcement learning [C]//33rd International Conference on Machine Learning. USA: International Machine Learning Society, 2016: 2850-2869.
[25]	LEI T, MING L. Towards cognitive exploration through deep reinforcement learning for mobile robots [EB/OL]. (2016-10-06)[2019-06-12]. https://arxiv.org/pdf/1610.01733.pdf

网络部分	动作选取	奖励预测
卷积层1	16, 8, 4	16, 8, 4
卷积层2	32, 4, 2	32, 4, 2
全连接层	256	128
LSTM	256	无

基于深度强化学习的区域化视觉导航方法

A Regionalization Vision Navigation Method Based on Deep Reinforcement Learning

RichHTML

PDF (PC)

可视化

摘要/Abstract

引用本文

使用本文

图/表 15

参考文献 25

相关文章 1

编辑推荐

Metrics

本文评价