cars is an increased driving safety.1,2,3 Safe motion control is consists of: • Assessing the collision probability while driving along a given trajectory. • Selecting control signals that minimize collision probability. Introduction Problem statement System overview Experimental results Conclusion 3  C. D. Harper, C. T. Hendrickson, and C. Samaras, “Cost and benefit estimates of partially-automated vehicle collision avoidance technologies,” Accident Analysis & Prevention, vol. 95, pp. 104–115, 2016.  P. Liu, R. Yang, and Z. Xu, “How safe is safe enough for self-driving vehicles?” Risk analysis, 2018.  H. Wang, Y. Huang, A. Khajepour, Y. Zhang, Y. Rasekhipour, and D. Cao, “Crash mitigation in motion planning for autonomous vehicles,” IEEE Transactions on Intelligent Transportation Systems, vol. 20, no. 9, pp. 3313–3323, 2019. Why is safety in autonomous vehicles so important?
the prior vehicle state-space probability Gaussian distribution propagation along the predefined trajectory;3,4 • Monte Carlo Motion Planning algorithms that estimate collision probability by repeatedly simulating the vehicle movement along the desired trajectory.5,6 4 Methods for assessing collision probability  X. Hu, L. Chen, B. Tang, D. Cao, and H. He, “Dynamic path planning for autonomous driving on various roads with avoidance of static and moving obstacles,” Mechanical Systems and Signal Processing, vol. 100, pp. 482–500, 2018.  J. J. Kuffner Jr and S. M. LaValle, “Rrt-connect: An efficient approach to single-query path planning,” in ICRA, vol. 2, 2000.  A. Houenou, P. Bonnifait, and V. Cherfaoui, “Risk assessment for collision avoidance systems,” in 17th International IEEE Conference on Intelligent Transportation Systems (ITSC). IEEE, 2014, pp. 386–391.  S. Patil, J. Van Den Berg, and R. Alterovitz, “Estimating probability of collision for safe motion planning under gaussian motion and sensing uncertainty,” in Robotics and Automation (ICRA), 2012 IEEE International Conference on. IEEE, 2012, pp. 238–3244.  L. Janson, E. Schmerling, and M. Pavone, “Monte carlo motion planning for robot trajectory optimization under uncertainty,” in Robotics Research. Springer, 2018, pp. 343–361.  E. Schmerling and M. Pavone, “Evaluating trajectory collision probability through adaptive importance sampling for safe motion planning,” arXiv preprint arXiv:1609.05399, 2016  S. D. Bopardikar, B. Englot, and A. Speranzon, “Multiobjective path planning: Localization constraints and collision probability,” IEEE Transactions on Robotics, vol. 31, no. 3, pp. 562–577, 2015.  N. E. Du Toit and J. W. Burdick, “Probabilistic collision checking with chance constraints,” IEEE Transactions on Robotics, vol. 27, no. 4, pp. 809–815, 2011. Introduction Problem statement System overview Experimental results Conclusion The task of assessing the safety of the motion trajectory under conditions of ego-pose uncertainty is of high priority.7,8
Problem statement System overview Experimental results Conclusion O. Shipitko, V. Kibalov and M. Abramov, "Linear Features Observation Model for Autonomous Vehicle Localization," 2020 16th International Conference on Control, Automation, Robotics and Vision (ICARCV), Shenzhen, China, 2020, pp. 1360-1365, doi: 10.1109/ICARCV50220.2020.9305434.
possible vehicle future trajectories. It depends on: • ego-pose uncertainty; • how this uncertainty will change over time. Since it is computationally intractable to generate so many predictions of trajectories with sufficient discretization, several simplifications are proposed. reference trajectory T probability distribution of vehicle pose estimated vehicle pose predicted trajectory for possible future trajectories 6 How to estimate collision probability within prediction horizon Problem statement Introduction System overview Experimental results Conclusion
generation1. 2) Speed control, which minimizes the probability of collision (further referred as the safe speed)2. 3) Combination of (1) and (2)3. 7 How to minimize collision probability?  Hu, X., Chen, L., Tang, B., Cao, D., & He, H. Dynamic path planning for autonomous driving on various roads with avoidance of static and moving obstacles. Mechanical Systems and Signal Processing, 100, 482–500, 2018.  Carvalho, Ashwin Mark. Predictive Control under Uncertainty for Safe Autonomous Driving: Integrating Data-Driven Forecasts with Control Design, UC Berkeley Electronic Theses and Dissertations, 2016.  Chu, K., Lee, M., & Sunwoo, M. Local Path Planning for Off-Road Autonomous Driving With Avoidance of Static Obstacles. IEEE Transactions on Intelligent Transportation Systems, 2012 System overview Experimental results Conclusion Introduction Problem statement
: Problem statement maximum possible speed (restricted by road properties) estimated vehicle pose probability distribution of vehicle pose current vehicle speed reference trajectory static occupancy grid map detected dynamic obstacles prediction duration for the prediction horizon collision probability threshold function speed limit for prediction horizon (the maximum allowed speed on the predicted trajectory) conditional collision probability for given within the prediction horizon . further referred as input data 8 Introduction Problem statement System overview Experimental results Conclusion
periodically. • The calculation is performed in cycle. • Аt each cycle iteration is estimated for given . • The maximum for which < is considered to be . 9 Introduction Problem statement System overview Experimental results Conclusion
one that controls the vehicle. • Dynamic vehicle model should dynamically and kinematically respond to control signals similar to a real vehicle. 10 Introduction Problem statement System overview Experimental results Conclusion additional longitudinal force for possible external factors (slope, wind, human bodies, etc) Introduction Problem statement System overview Experimental results Conclusion
Monte Carlo method sdfdsf , - is the probability of i-th pose hypothesis to be the current vehicle pose vehicle projection to the road plane predicted trajectory transferred to i-th hypothesis 11 B C D A Introduction Problem statement System overview Experimental results Conclusion
with few exceptions: • Dynamic obstacles are defined in the vehicle reference frame. • for each the trajectory is predicted separately (assuming = ). • Motion of dynamic obstacles in prediction horizon is neglected. 12 Introduction Problem statement System overview Experimental results Conclusion
severely limited: • in front of difficult sections on the route (narrow entrances, turns etc.); • where the pose estimation accuracy was low. Speed recovers after: • passing difficult sections; • clarifying vehicle estimated pose. for single estimated vehicle pose for every pose hypothesis Future trajectories collision inspection with dynamic obstacles Advantages of proposed method in comparison with single prediction deterministic method: • Vehicle slows down in advance when detected obstacles are approaching; • Vehicle motion in the proximity of obstacles becomes smoother and less jerky. Disadvantages: • High computational complexity. 14 Introduction Problem statement System overview Experimental results Conclusion
• run on Intel Core i7-8700 CPU in single-thread mode • performance-affecting parameters: ◦ number of pose hypothesis — 1000 for static, 100 for dynamic ◦ prediction horizon — 15 seconds (with a sampling step of 0.1 seconds) ◦ number of speed estimation iterations — 10 (that allows to achieve a precision of 1/1024 of the maximum speed) • Speed estimation rate — 11.2 Hz • Total computational time proportions: ◦ collision probability estimation for static obstacles — 46% ◦ collision probability estimation for dynamic obstacles — 42% ◦ other operations — 12% The proposed system is used on several self-driving cargo platforms with operating speeds up to 25 km/h. 15 Computational performance and applications Introduction Problem statement System overview Experimental results Conclusion Introduction Problem statement System overview Experimental results Conclusion
real-time calculation of safe speed limit was developed. • The method for collision probability estimation with both static and dynamic obstacles takes into account the ego-pose uncertainty. • Experiments showed that the system allows avoiding collisions by decreasing vehicle speed. • The proposed method can also be used to control the quality of the localization system • Account the error model of dynamic obstacle detection. • Expand the method by the motion prediction of dynamically detected obstacles. • Research the behavior of the system at higher speeds. 16 Conclusion Future work Introduction Problem statement System overview Experimental results Conclusion