Runtime assurance method for eVTOL intelligent obstacle avoidance system toward safety of intended functionality

doi:10.16265/j.cnki.issn1003-3033.2024.07.0198

Abstract

Abstract:

To ensure the SOTIF of eVTOL vehicles in UAM and reduce the verification difficulty of artificial intelligence algorithms, an obstacle avoidance model was proposed based on RTA method. Firstly, SAC (soft actor-critic) algorithm improved by the artificial potential field method was used as the complex function of the eVTOL intelligent obstacle avoidance system. Then, dynamic response planning (DRP) was used as a backup function of the intelligent avionics system to mitigate SOTIF hazards. Moreover, monitoring and decision-making modules were adopted to obtain environmental conditions and develop an RTA architecture. Finally, the simulated obstacle avoidance performance was compared between the two systems using complex function and RTA. The results showed that both methods can achieve obstacle avoidance, but the traditional obstacle avoidance system using complex functions may impose SOTIF risk. The RAT architecture design increased safe flight time from 78.4% to 98.15%, with the total route length only increasing by 0.95%, reducing risks in operational scenarios while ensuring efficiency.

Key words: safety of the intended functionality (SOTIF), electric vertical take-off and landing (eVTOL), runtime assurance (RTA), urban air mobility (UAM), intelligent obstacle avoidance system

CLC Number:

X949

DONG Lei, SONG Wenjia, CHEN Xi, LIU Jiachen, LIANG Boyao. Runtime assurance method for eVTOL intelligent obstacle avoidance system toward safety of intended functionality[J]. China Safety Science Journal, 2024, 34(7): 105-112.

Figures/Tables 15

Fig.1

Fig.2

Fig.3

Table 1

SAC algorithm pseudocode improved by APF

伪代码1:APF改进SAC算法

初始化参数

θ, ψ 1, ψ 2, ψ - 1, ψ - 2

初始化经验池
for轮次=1到M
接收初始状态

s 1

for t = 1到T
依据策略

π (θ)

和状态

s t

生成动作

a t

APF优化

a t

→

a' t

执行动作

a' t

,观察奖励

r t

和新状态

s t + 1

将

(s t, a t, r t, s t + 1)

添加到经验池D
for j = 1 to更新次数do
经验池D随机采样一批数据
计算评论家目标网络的Q值
更新评论家网络参数

ψ 1

和

ψ 2

更新演员网络参数

θ

软更新评论家目标网络参数

ψ - 1

和

ψ - 2

end for
令

s t ← s t + 1

end for
end for

Table 1

Fig.4

Fig.5

Fig.6

Table 2

SOTIF scenario classification using DRP

SOTIF场景	DRP类型	运行状态
已知安全	$ϕ ∈ ϕ s$	安全
已知不安全	$ϕ ∈ ϕ u *$	非常危险
未知安全	$ϕ ∈ ϕ u 且 ϕ ∉ ϕ u *$	危险
未知不安全	$ϕ ∈ ϕ u 且 ϕ ∉ ϕ u *$	危险

Table 2

Table 3

Parameters for deep reinforcement learning

参数	数值
熵正则化系数 $α$	0.2
训练批次大小	128
经验池大小	10⁶
探索函数	随机动作

Table 3

Table 4

Fig.7

Fig.8

Fig.9

Fig.10

Table 5

References 17

[1]	MENOUAR H, GUVENC I, AKKAYA K, et al. UAV-enabled intelligent transportation systems for the smart city: applications and challenges[J]. IEEE Communications Magazine, 2017, 55(3): 22-28.
[2]	李诚龙, 屈文秋, 李彦冬, 等. 面向eVTOL航空器的城市空中运输交通管理综述[J]. 交通运输工程学报, 2020, 20(4): 35-54.
	LI Chenglong, QU Wenqiu, LI Yandong, et al. Overview of traffic management of urban air mobility (UAM) with eVTOL aircraft[J]. Journal of Traffic and Transportation Engineering, 2020, 20(4): 35-54.
[3]	绿色航空制造业发展纲要(2023—2035年)[EB/OL].(2023-10-10). https://www.gov.cn/zhengce/zhengceku/202310/content_6908243.htm.
[4]	SCHWEIGER A, ANNIGHOEFER B, REICH M, et al. Classification for avionics capabilities enabled by artificial intelligence[C]. 2021 IEEE/AIAA 40^th Digital Avionics Systems Conference (DASC), 2021 :1-10.
[5]	DEY S, LEE S W. Multilayered review of safety approaches for machine learning-based systems in the days of AI[J]. Journal of Systems and Software, 2021, 176: DOI:10.1016/j.jss.2021.110941.
[6]	董磊, 刘嘉琛, 陈曦, 等. 面向适航符合性的智能航电系统认证研究进展[J]. 航空工程进展, 2023, 14(3): 26-40.
	DONG Lei, LIU Jiachen, CHEN Xi, et al. Research progress of AI-based avionics system certification for airworthiness compliance[J]. Advances in Aeronautical Science and Engineering, 2023, 14(3): 26-40.
[7]	HE Yuning, SCHUMANN J, YU Huafeng. Toward runtime assurance of complex systems with AI components[C]. PHM Society European Conference, 2022: 166-174.
[8]	EASA. Concepts of design assurance for neural networks (CoDANN) II[R], 2021.
[9]	SCHIRMER S, TORENS C, DAUER J C, et al. A hierarchy of monitoring properties for autonomous systems[C]. AIAA SCITECH 2023 Forum, 2023: 2588-2601.
[10]	COFER D, AMUNDSON I, SATTIGERI R, et al. Run-time assurance for learning-based aircraft taxiing[C]. 2020 AIAA/IEEE 39^th Digital Avionics Systems Conference (DASC), 2020: 1-9.
[11]	GHORI S, KHAMVILAI T, FERON E, et al. Runtime assurance for distributed avionics architecture[C]. 2022 IEEE/AIAA 41^st Digital Avionics Systems Conference (DASC), 2022: 1-6.
[12]	KIRKENDOLL Z. Automatic ground collision avoidance and loss of control prevention systems for general aviation using run-time assurance[C]. 2022 IEEE/AIAA 41^st Digital Avionics Systems Conference (DASC), 2022: 1-10.
[13]	ISO/PAS 21448-2019, Safety of the intend-ed functionality[S].
[14]	罗崎瑞, 张道文, 周华, 等. 面向智能汽车预期功能安全的驾驶场景评价[J]. 中国安全科学学报, 2022, 32(8): 140-145. doi: 10.16265/j.cnki.issn1003-3033.2022.08.1768
	LUO Qirui, ZHANG Daowen, ZHOU Hua, et al. Evaluation on driving scenarios for safety of intended functionality of intelligent vehicles[J]. China Safety Science Journal, 2022, 32(8): 140-145. doi: 10.16265/j.cnki.issn1003-3033.2022.08.1768
[15]	董陇军, 王加闯. 智能安全人机工程学学科建设及内容体系研究[J]. 中国安全科学学报, 2022, 32(3): 1-8. doi: 10.16265/j.cnki.issn1003-3033.2022.03.001
	DONG Longjun, WANG Jiachuang. Research on discipline construction and content system of intelligent safety ergonomics[J] China Safety Science Journal, 2022, 32(3): 1-8. doi: 10.16265/j.cnki.issn1003-3033.2022.03.001
[16]	HAARNOJA T, ZHOU A, HARTIKAINEN K, et al. Soft actor-critic algorithms and applications[J]. arXiv Preprint, 2018:DOI:10.48550/arXiv.1812.05905.
[17]	COOPER J K, SCHIERMAN J D. A sense and avoid system for unmanned aircraft in formation flight[C]. AIAA Guidance, Navigation, and Control Conference, 2014: 967-986.

编号	起点	终点	状态
1	(330,0,225)	(0,225,225)	已知安全
2	(330,0,105)	(0,225,180)	未知安全/ 未知不安全
3	(330,75,225)	(0,75,75)	已知安全
4	(330,75,105)	(0,75,180)	已知安全
5	(330,105,75)	(0,0,225)	已知安全
6	(330,225,225)	(0,0,75)	已知安全
7	(330,225,75)	(0,0,225)	已知安全
8	(330,225,180)	(0,0,180)	已知安全

方法	航路长度/m	安全时长占比/%
仅使用强化学习	324.98	98.15
RTA避障	328.08	78.4