Autonomous Navigation Trajectory Evaluation Method Based on Uncertainty Analysis
YAN Yan1,2, TANG Zhenmin2, LIU Jiayin2
1. R&D Center, China Academy of Launch Vehicle Technology, Beijing 100076, China;
2. Institute of Pattern Recognition and Artificial Intelligence, Nanjing University of Science and Technology, Nanjing 210094, China
In order to achieve qualitative and quantitative evaluation of autonomous navigation trajectory, a trajectory analysis method based on uncertainty cloud model is proposed. In this method, robot's trajectory features are extracted from autonomous driving and avoiding actions, and corresponding feature cloud models for position warp, direction warp and obstacle-avoidance safety distance trajectories are generated. In cloud model, expectation is the basic metric of trajectories, and the fuzziness and randomness of features are expressed by entropy and hyper-entropy. By taking advantages of uncertainty metric of cloud models, the transient state and stability in autonomous navigation can be calculated. Experiment results show that this method can effectively be used in trajectory evaluation of autonomous navigation, and it can compensate shortcomings of reinforcement learning methods in stability evaluation.
[1] O'Day S, Steinberg M, Yglesias C, et al. Metrics for intelligent autonomy[C]//Performance Metrics for Intelligent Systems Workshop. Gaithersburg, USA: National Institute of Standards and Technology, 2004.[2] Huang H M, Albus J, Messina E, et al. Specifying autonomy levels for unmanned systems: Interim report[C]//Proceedings of SPIE, vol.5422. Bellingham, USA: SPIE, 2004: 386-397.[3] Huang H M, Pavek K, Ragon M, et al. Characterizing unmanned system autonomy: Contextual autonomous capability and level of autonomy analyses[C]//Proceedings of SPIE, vol.6561. Bellingham, USA: SPIE, 2007: 1-9.[4] Hadsell R, Sermanet P, Ben J, et al. Learning long-range vision for autonomous off-road driving[J]. Journal of Field Robotics, 2009, 26(2): 120-144. [5] Procopio M J, Mulligan J, Grudic G. Learning terrain segmentation with classifier ensembles for autonomous robot navigation in unstructured environments[J]. Journal of Field Robotics, 2009, 26(2): 145-175. [6] Evans J M. Definitions and measures of intelligence in deep blue and the army XUV[C]//Performance Metrics for Intelligent Systems Workshop. Gaithersburg, USA: National Institute of Standards and Technology, 2007: 148-151.[7] 杨哲,张汝波.无人系统自主等级模糊评价方法[J].小型微型计算机系统,2009,30(10):2043-2047. Yang Z, Zhang R B. Fuzzy evaluated method for the autonomy levels of unmanned systems[J]. Journal of Chinese Computer Systems, 2009, 30(10): 2043-2047.[8] 王勇鑫,钱徽,金卓军,等.基于轨迹分析的自主导航性能评估方法[J].计算机工程,2011,37(6):141-144. Wang Y X, Qian H, Jin Z J, et al. Autonomous navigation performance evaluation method based on trajectory analysis[J]. Computer Engineering, 2011, 37(6): 141-144.[9] 金卓军,钱徽,朱淼良.基于倾向性分析的轨迹评测技术[J].浙江大学学报:工学版,2011,45(10):1732-1737. Jin Z J, Qian H, Zhu M L. Trajectory evaluation method based on intention analysis[J]. Journal of Zhejiang University: Engineering Science, 2011, 45(10): 1732-1737.[10] 李德毅,刘常昱,杜鹢,等.不确定性人工智能[J].软件学报,2004,15(11):1583-1594. Li D Y, Liu C Y, Du Y, et al. Artificial intelligence with uncertainty[J]. Journal of Software, 2004, 15(11): 1583-1594.